Version history

A living proposal
needs a public memory.

Every material release records what changed, why the package changed, and what remains unfinished. Future evidence-based corrections will be preserved here.

Download the consolidated NRR research archive (v0.1–v0.26) ↓

Version10.0

Community-contained profiles and clearer cutline cards

What changed

  • Moved profile access into a compact Community account utility, keeping the public proposal free of account gates.
  • Rebuilt the teams-outside-the-cutline rows as fully clickable résumé selectors with NRR rank, record, and Index aligned on the right; removed the unexplained score-gap pills and Show card labels.
  • Connected Clerk-managed email-or-username accounts so passwords, verification, sessions, and recovery are not stored in a home-grown database; Google is an optional convenience rather than a requirement.
  • Connected bettercollegefootball.com and durable inbound contact aliases.

Still open

  • Complete final production sign-in and recovery tests after the identity-provider keys are promoted.
  • The .org defensive registration remains optional.
Version9.9

Community account clarity and synchronized proposal backups

What changed

  • Removed the complete-proposal download from the global header. The website is now stated as the primary version; synchronized backup documents remain only at the bottom of Proposal and in Evidence.
  • Verified the complete My profile sign-in, load, save, consent, and persisted-default flow in production. Profiles remain optional unless a reader wants to post; the proposal, rankings, evidence, and approved discussions remain public without an account.
  • Renamed the private account destination My community profile, strengthened its return path to Community, and added a direct profile entry inside the Community hero.
  • Updated the downloadable proposal’s NRR section to v0.48, its current 30 / 30 / 20 / 15 / 5 public balance, and the twelve time-locked 2024–25 weekly checkpoints.
  • Tightened the Rankings switchboard, exposed the full conference-championship label, and moved the disagreement review immediately after the team and résumé workspace.

Still open

  • A custom domain and provider-backed email aliases still require owner purchase and account setup.
  • The public website controls if a later revision and a downloaded backup ever diverge; future material releases should update both in the same change.
  • Institutional and bowl marks remain in the launch rights register pending final legal and licensing decisions.
Version9.8

Launch operations, rights register, and incident readiness

What changed

  • Added a private Operations & rights control room with live queue counts, a DMCA-readiness matrix, a rights-risk register, durable-address plan, incident-response runbook, and a concrete owner-only pre-launch list.
  • Published a standard security.txt contact record and added a daily owner queue-review routine without allowing automation to approve, decline, remove, or publish reader content.
  • Expanded pre-storage community screening to reject exposed credentials, government identifiers, payment-card details, and direct threats of violence while retaining private human review for every publish decision.
  • Added direct source links for every actively credited non-personal photograph that was missing one and re-ran the photo-reuse ledger audit.
  • Removed the owner’s personal home address from public footers, contact, rights, and unsubscribe pages while the registered-agent and sender-address decisions remain open.
  • Recorded the 138 cached school marks and six bowl marks as the principal unresolved launch-rights risk instead of treating an image-source URL or no-endorsement disclaimer as permission.

Still open

  • Alec must purchase the domain, choose public contact details, create any provider accounts, accept their terms, and complete any Copyright Office filing in his own legal identity.
  • Email aliases, outbound update delivery, Turnstile, and broader identity providers cannot be connected until a domain and the relevant account credentials exist.
  • A qualified U.S. attorney—not the website or an automated scan—must give the final answer on §512 eligibility, project-name clearance, and whether the planned institutional and bowl mark uses are defensible.
Version9.7

Layered community safety and long-run operations

What changed

  • Added server-side pre-storage screening for common profanity, slurs, public email addresses, phone numbers, street addresses, repeated-character spam, and excessive links. Automated screening assists rather than replaces human review.
  • Added a bot-trap field to community submissions while retaining accountable sign-in, one-contribution-per-day limits, and private moderation before publication.
  • Added a private safety-report link to every published thread and reply, plus a dedicated community-safety contact category and prefilled report subject.
  • Published clearer community-safety disclosures, emergency guidance, and prohibited-content terms without claiming that automated filters are complete or infallible.
  • Added counter-notice instructions, a repeat-infringer policy, a configurable account-blocking control, and a private notice-record practice without claiming safe-harbor eligibility.
  • Expanded the private launch center with daily, monthly, quarterly, incident-response, and annual operating checks for moderation, backups, deletion requests, access, dependencies, rights, and vendor changes.

Still open

  • The owner must consistently monitor and act on the private moderation and reporting queues; code cannot supply judgment or emergency response.
  • A designated DMCA-agent registration, legal review, domain purchase, independent trademark clearance, and permission decisions remain owner or counsel actions.
  • If participation grows materially, the project should reassess manual moderation capacity and whether a specialist abuse-detection service is warranted.
Version9.6

Launch rights, profile, and accessibility controls

What changed

  • Added a public copyright, trademark, photography, and DMCA-status page with a complete notice checklist, rights-reporting channel, and the project mailing address.
  • Added a visible My profile link in the global header and restored phone as an optional private profile field. Existing values are retained; saving the field blank removes a value. Phone is never required to sign in or post.
  • Added a public one-step update-list unsubscribe page before any email provider is connected, plus the sender mailing address in the contact area and global footer.
  • Added a public accessibility statement and a dedicated accessibility-report category; completed a targeted keyboard, focus, semantic, mobile-menu, and screen-reader-oriented review.
  • Added compact third-party-mark notices to the logo-heavy Rankings and Bracket Lab tools and linked them to the full rights record.

Still open

  • At this release, the hosting platform still supplied identity; version 10.0 superseded that flow with Clerk-managed accounts.
  • The public DMCA reporting channel is not a U.S. Copyright Office designated-agent registration. Registration remains an owner filing and must publish the registered agent’s complete contact information.
  • A preliminary exact-phrase search is not a full federal, state, and common-law trademark clearance or legal opinion.
Version9.5

Pre-launch privacy, security, and rights review

What changed

  • Added public Terms & Rights, copyright-reporting, privacy-removal, and accessibility contact paths; linked the terms from the footer and sitemap.
  • Stopped requesting or returning phone numbers in new community profiles. ChatGPT sign-in email and the private account identifier now provide the accountability layer with less collected data.
  • Added same-origin JSON request checks, payload limits, a fixed canonical origin, and baseline browser security headers including content-type protection, frame blocking, referrer limits, and a Content Security Policy.
  • Replaced three photographs without verified source metadata and the official championship trophy image with licensed archive photographs and an original CSS trophy illustration.
  • Updated React, the server-component runtime, vinext, Vite, Cloudflare tooling, and Wrangler; the deployed dependency audit reports no production vulnerabilities.
  • Rebuilt the ranking switchboard status and published-weight banner for readable checkpoint context and five percentage-led model blocks.

Still open

  • Team and bowl marks remain documentary identifiers and still require a final rights-and-permissions review before a custom-domain launch.
  • Any legacy phone values already stored in the database were not deleted; deletion requires an explicit owner-authorized data cleanup.
  • Before sending update-list email, the project still needs a compliant unsubscribe flow and valid postal address.
Version9.4

NRR v0.48 loss hierarchy and shared playoff field

What changed

  • Kept the public 30% Quality-Win Depth, 30% Schedule-Adjusted Record, 15% Opponent-Adjusted Performance, 20% Loss Integrity, and 5% Win Control balance unchanged.
  • Recalibrated direct loss severity around official-checkpoint-majority Opponent Value. Margin damage now fades continuously as opponent quality rises, while a smooth uncapped softplus term gives weak-opponent losses the additional damage they warrant. No hard damage cap, opponent tier, school rule, or conference rule enters the calculation.
  • Added regression comparisons for the reported 2024 cases. BYU’s home loss to 5–7 Kansas now carries more than 1.75 times the direct severity and final Index allocation of Tennessee’s road loss to elite Georgia; Arizona State’s sustained road loss at Cincinnati remains materially worse than Tennessee’s late road loss at Arkansas.
  • Added a verified 2024 Selection Sunday replay to Bracket Lab, including all five conference champions and a results-informed starting path that remains fully editable.
  • Created one shared 18-team field rule for Bracket Lab and Rankings. The weekly ranking table now marks double byes, bowl-stage byes, campus hosts, road entrants, projected conference champions, and the required non-Power Four champion.

Still open

  • The opponent-conditioned margin rate, smooth weak-opponent scale, and other Loss Integrity curves remain transparent calibration choices requiring held-out seasons and the live 2026 weekly run.
  • Weekly conference-champion markers are projections until a champion is known; Selection Sunday uses recorded championship results.
  • The compact playoff markers are a same-day field translation of NRR, not an additional input into NRR or a post-formula ranking adjustment.
Version9.3

NRR v0.47 additive loss attribution

What changed

  • Kept every v0.46 ranking, component score, public weight, lower-tail calibration, schedule rule, and title-game policy unchanged.
  • Removed the leave-one-loss-out display calculation. That counterfactual could make a worse game show a smaller marginal Index number because the remaining loss inherited shared frequency and weakest-loss penalties.
  • The full-résumé Loss Integrity result is still calculated once with the same nonlinear formula. Its exact deficit from 100 is now allocated additively across losses in proportion to each game’s direct severity, which already includes opponent value, margin, Q4 / 6:00 / 2:00 state, venue, repeat-opponent damage, and conference-title protection.
  • Each expanded loss card now publishes the game’s share of the team’s total loss cost, its Loss Integrity allocation, and its exact published-Index allocation after the record foundation. Those allocations add back to the team’s complete Loss Integrity penalty.
  • Added a cross-team regression using the reported 2024 comparison: Arizona State’s road loss at Cincinnati must show a larger game severity and Index allocation than Tennessee’s late road loss at Arkansas.

Still open

  • The additive figures explain how the scored Loss Integrity component affects the published Index; they do not claim that removing a game would leave every other nonlinear résumé input unchanged.
  • The allocation uses the same direct severity equation for every loss and contains no school, conference, poll-position override, or manual reorder.
  • The loss formula still requires additional held-out seasons and the live 2026 weekly run; this release corrects attribution, not the calibrated ranking order.
Version9.2

NRR v0.46 readable lower-tail calibration

What changed

  • Kept the public 30% Quality-Win Depth, 30% Schedule-Adjusted Record, 15% Opponent-Adjusted Performance, 20% Loss Integrity, and 5% Win Control balance unchanged.
  • Replaced the saturating weakest-loss reduction with a continuous 1.2-rate exponential decay applied to the same 0.45-power opponent-value shortfall. A smooth two-point lower-tail calibration then preserves useful separation among damaged Loss Integrity résumés without imposing a minimum score.
  • Applied a smaller 1.5-point endpoint-anchored lower-tail calibration to Quality-Win Depth, so thin but real opponent evidence remains distinguishable longer. Zero evidence remains exactly zero and 100 remains exactly 100.
  • Regenerated and audited all twelve 2024–25 checkpoints. At 2025 Selection Sunday, Louisville’s Loss Integrity moves from 2.61 to about 9.1 and Illinois moves from 2.27 to about 9.3, while truly empty Quality-Win Depth still reaches zero.
  • Preserved the calibrated loss comparisons and title-game policy: Ohio State’s 2024 home Michigan loss remains about 1.61 times as damaging as Texas’s ordinary home Georgia loss, while the Oregon road loss and Georgia title-game loss remain in the same low-risk severity tier.

Still open

  • The two lower-tail lift values and the 1.2 exponential rate are transparent calibration choices requiring additional held-out seasons and the live 2026 weekly run.
  • The transformation is monotonic and has no hard floor, cap, team rule, conference rule, or manual placement; it changes spacing, not the meaning of zero or 100.
  • Lower-tail readability does not turn a weak win or a loss into positive quality credit. Résumés can still approach zero as evidence deteriorates.
Version9.1

NRR v0.45 bad-loss separation and explicit title-game context

What changed

  • Kept the public 30% Quality-Win Depth, 30% Schedule-Adjusted Record, 15% Opponent-Adjusted Performance, 20% Loss Integrity, and 5% Win Control balance unchanged.
  • Changed the continuous weakest-loss credibility exponent from 1.25 to 0.45. The universal curve now lets one truly damaging loss dominate the loss profile more clearly without adding a team, conference, opponent-tier, or manual-placement rule.
  • Changed every conference-title loss to one-third ordinary game severity while retaining half ordinary record and loss-frequency treatment. Repeat-opponent damage remains, so a rematch is not erased.
  • Regenerated all twelve 2024–25 checkpoints. At 2024 Selection Sunday, Ohio State’s home Michigan loss produces a 10.99-point final Index hit versus 6.81 for Texas’s ordinary home loss to Georgia—a 1.61-to-1 relationship. Ohio State’s one-point road loss at Oregon and Texas’s title-game rematch sit in the same low-risk severity tier at 10.16 and 10.19.
  • Added explicit Conference championship labels to compact résumé rows, expanded win and loss proof, opponent résumé overlays, and archived game views. Ordinary neutral-site games remain labeled Neutral.

Still open

  • The 0.45 weakest-loss exponent and one-third title-game severity rate are transparent calibration choices requiring additional held-out seasons and the live 2026 weekly run.
  • A title-game loss still lowers the résumé and remains fully visible; the policy reduces disproportionate risk rather than converting the result into a reward.
  • The adjustment is universal: no Ohio State, Texas, Michigan, Georgia, school, conference, or manual placement rule enters the calculation.
Version9.0

NRR v0.44 loss-integrity and title-game consistency

What changed

  • Kept the public 30% Quality-Win Depth, 30% Schedule-Adjusted Record, 15% Opponent-Adjusted Performance, 20% Loss Integrity, and 5% Win Control balance unchanged, along with the v0.43 Schedule-Adjusted Record calibration.
  • Changed the continuous weakest-loss credibility exponent from 2 to 1.25. A damaging loss now registers sooner below Opponent Value 60, but there is still no hard opponent tier, cliff, conference rule, or quality-loss credit.
  • Applied half ordinary record, loss-frequency, and game-severity treatment to every conference-title loss. A title-game rematch still retains the repeat-opponent damage and eight-point Schedule-Adjusted Record deduction.
  • Regenerated all twelve 2024–25 checkpoints. At 2024 Selection Sunday, Texas posts 51.3 Loss Integrity and Ohio State 40.1. Ohio State’s home loss to Michigan becomes its dominant final Index hit at 7.88 points versus 2.65 for the Oregon loss; Texas’s title-game rematch hit is 2.36 versus 6.44 for the ordinary Georgia loss.

Still open

  • The 1.25 weakest-loss exponent and half ordinary title-game treatment remain public calibration choices requiring additional held-out seasons and the live 2026 weekly run.
  • The adjustment is universal: no Ohio State, Texas, Michigan, Georgia, school, conference, or manual placement rule enters the calculation.
  • A title-game loss is still visible and still lowers the résumé; repeat-opponent penalties remain so rematches are not erased.
Version8.9

NRR v0.43 equal-record schedule separation

What changed

  • Kept the public 30% Quality-Win Depth, 30% Schedule-Adjusted Record, 15% Opponent-Adjusted Performance, 20% Loss Integrity, and 5% Win Control balance unchanged.
  • Rebalanced Schedule-Adjusted Record internally from 50% achieved record / 30% fixed-benchmark performance / 20% schedule challenge to 40 / 30 / 30. Actual record remains the largest single input, while schedule context receives enough authority to distinguish similar records earned against materially different slates.
  • Added one continuous 12-point logistic contrast curve after the existing 30% full-slate / 70% challenge-weighted schedule blend. The curve creates stronger separation around the national middle without a Top-25 tier, counted-game cap, conference rule, or hard evidence ceiling.
  • Regenerated all twelve 2024–25 checkpoints. At 2024 Selection Sunday, Ohio State’s Schedule-Adjusted Record rises to 77.5 while SMU falls to 69.7—a 7.8-point separation instead of 3.1. Indiana remains No. 9, Arizona State remains No. 11, and SMU moves from No. 7 to No. 10 through the formula alone.

Still open

  • The 40 / 30 / 30 internal balance and 12-point contrast scale remain public calibration choices requiring additional held-out seasons and the live 2026 weekly run.
  • The revision is intended to improve equal-record schedule separation universally, not force Ohio State, SMU, or any conference to a selected position.
  • No post-formula review or manual score adjustment is permitted; the unrounded Index still determines every published position.
Version8.8

NRR v0.42 schedule-adjusted-record calibration

What changed

  • Kept the public 30% Quality-Win Depth, 30% Schedule-Adjusted Record, 15% Opponent-Adjusted Performance, 20% Loss Integrity, and 5% Win Control balance unchanged.
  • Rebalanced Schedule-Adjusted Record internally from 65% achieved record / 20% fixed-benchmark performance / 15% schedule caliber to 50 / 30 / 20. Actual record remains the single largest input, while performance against the same fixed benchmark and the challenge of the slate receive enough authority to distinguish superficially similar records.
  • Rebuilt schedule challenge without a school or conference rule. Ranked-opponent strength blends 90% selected official checkpoint standing with 10% national play-level strength; unranked opponents retain play-level strength. Schedule caliber blends 30% of the complete slate with 70% of one smooth challenge-weighted average.
  • Counted every opponent and retained continuous values: no Top-25-only rule, hard schedule tier, counted-game cap, conference criterion, named-team adjustment, or manual reorder enters the result.
  • Regenerated all twelve 2024–25 weekly checkpoints and the public audit from one formula path. At 2024 Selection Sunday, Ohio State posts a 74.5 Schedule-Adjusted Record versus 71.4 for SMU, 71.4 for BYU, and 70.5 for Arizona State; Texas posts 68.5.

Still open

  • The 50 / 30 / 20 internal blend, 90 / 10 ranked-opponent input, and 30 / 70 full-slate/challenge blend remain public calibration choices requiring additional held-out seasons and the live 2026 weekly run.
  • The calibration is intended to improve general schedule and benchmark behavior, not force a named team or conference to a selected rank.
  • No post-formula review or manual score adjustment is permitted; the unrounded Index still determines every published position.
Version8.7

NRR v0.41 résumé-priority calibration

What changed

  • Shifted five public weight points from Opponent-Adjusted Performance to Quality-Win Depth. The published balance is now 30% Quality-Win Depth, 30% Schedule-Adjusted Record, 20% Loss Integrity, 15% Opponent-Adjusted Performance, and 5% Win Control.
  • Kept every underlying win, record, performance, loss, record-foundation, and head-to-head equation unchanged so the release isolates one visible question: how the public priorities should balance earned victories against descriptive play-level performance.
  • Regenerated all twelve 2024–25 checkpoints from the same formula path. The experiment moves 2024 Alabama from No. 14 to No. 13, Indiana from No. 8 to No. 10, Arizona State from No. 10 to No. 8, and BYU from No. 11 to No. 9; at the 2025 checkpoint, North Texas moves from No. 24 to No. 22. Those tradeoffs are disclosed rather than manually corrected.
  • Reordered the public résumé component bars by published weight—Quality-Win Depth, Schedule-Adjusted Record, Loss Integrity, Opponent-Adjusted Performance, and Win Control—and kept the Outside Top 25 badge on one line in compact evidence rows.

Still open

  • The 30 / 30 / 15 / 20 / 5 balance remains a public calibration choice requiring held-out seasons and the live 2026 weekly run.
  • The Indiana and Arizona State movement shows that greater Quality-Win Depth emphasis is a real tradeoff, not a cost-free correction.
  • No named-team adjustment, conference criterion, post-formula reorder, or manual score change enters v0.41.
Version8.6

NRR v0.40 record and schedule-validation revision

What changed

  • Shifted the public balance from 30% Quality-Win Depth and 25% Schedule-Adjusted Record to 25% and 30%. Opponent-adjusted performance, Loss Integrity, and Win Control remain 20%, 20%, and 5%. The change gives the complete record more authority and limits the ability of one elite win to carry an otherwise weaker résumé.
  • Kept full-slate opponent caliber as the performance-credibility baseline and added one continuous validation gate using the two strongest national schedule connections. The same formula runs for every team; no school, conference, Group-of-Five flag, poll label, or manual placement enters it.
  • Changed the universal record-foundation schedule deduction to grow with loss rate to the 1.5 power. A one-loss soft-schedule résumé keeps materially more record credit than a two- or three-loss résumé against the same slate.
  • Regenerated all twelve checkpoints. At 2024 Selection Sunday, Indiana moves from No. 15 to No. 8, Arizona State from No. 7 to No. 10, and BYU from No. 8 to No. 11. At the 2025 checkpoint, James Madison is No. 17, North Texas No. 24, and three-loss South Florida is outside the Top 25 at No. 27.
  • Simplified compact win evidence by removing percentile clutter, preserving directly comparable evidence units, allowing full opponent names to wrap, and removing the unscored championship-policy bar from the public résumé card.

Still open

  • The validation gate, record-foundation exponent, and 25 / 30 / 20 / 20 / 5 weights remain public calibration choices requiring held-out seasons and the live 2026 weekly run.
  • The revision is designed to improve general record and schedule behavior, not to force any named team to a selected rank.
  • The official CFP checkpoint remains the majority input to Opponent Value, and every rank is still ordered by the unrounded formula Index.
Version8.5

NRR v0.39 soft-schedule credibility revision

What changed

  • Kept the published 30 / 25 / 20 / 20 / 5 weights and the 65 / 20 / 15 schedule-adjusted-record blend; this release changes how isolated soft-schedule performance and record credibility are calibrated, not the public priorities.
  • Replaced the symmetric 75 / 25 scored-performance blend with a one-directional soft-schedule credibility curve. The national play-level ridge remains the baseline; one continuous deduction grows below national schedule caliber 50 and fades toward zero as the slate strengthens. Strong schedules receive no positive bonus.
  • Applied the same shortfall lightly to Win Control and to the record foundation. The record deduction grows continuously with loss frequency, preserving more credit for a one-loss soft-schedule résumé than for a two- or three-loss résumé against the same slate.
  • Regenerated all twelve checkpoints. At 2025 Selection Sunday, James Madison moves from No. 16 to No. 18, North Texas from No. 19 to No. 25, and three-loss South Florida from No. 23 to No. 26. Oklahoma remains above Alabama, while the leading contenders retain their order.
  • Kept the model conference-neutral and formula-driven: no conference name, Group-of-Five flag, school, brand, poll label, hard schedule tier, or manual rank adjustment enters the correction.

Still open

  • The four-point softplus scale, 4.5-point performance factor, 40-point Win Control decay, 1.15 record curve, and loss-rate schedule deduction are public calibration choices requiring held-out seasons and the live 2026 weekly run.
  • A weak schedule does not erase an excellent record; it reduces confidence progressively when that résumé also accumulates losses.
  • The next validation step should test the same equation against additional seasons without altering it to fit individual teams.
Version8.4

NRR v0.38 schedule-connectivity revision

What changed

  • Added a checkpoint-wide percentile beside every comparable win-evidence value, so each result now carries a cross-FBS checkpoint comparison without replacing the evidence units used by the formula.
  • Rebuilt compact loss rows around two explicitly different numbers: individual game severity before résumé curves and the final marginal Index hit after loss frequency, weakest-loss credibility, record foundation, and title-game policy.
  • Changed schedule-adjusted record from 70% achieved record / 20% fixed-benchmark performance / 10% slate caliber to 65 / 20 / 15. Record remains the clear majority while a soft slate receives modestly more universal opponent context.
  • Changed scored opponent-adjusted performance to 75% national ridge performance plus 25% schedule-connected performance. The connection is continuous, uses the same full-slate-caliber rule for every team, and never reads a school, conference, brand, poll, or Group-of-Five label.
  • Regenerated all twelve checkpoints. James Madison remains No. 16 in 2025; North Texas moves from No. 18 to No. 19; Arizona moves to No. 18. Across the audited 2024 Selection Sunday field, every affected Top-25 team moves no more than two places.

Still open

  • Evidence percentiles are descriptive checkpoint context; evidence units remain the actual formula input.
  • A weak schedule should not erase an excellent record, and an excellent record should not conceal a weak opponent portfolio. Additional held-out seasons are still needed to test that balance.
  • The 65 / 20 / 15 schedule-adjusted-record blend and 75 / 25 scored-performance blend are public calibration choices that require the live 2026 weekly run.
Version8.3

NRR cross-résumé win comparison

What changed

  • Replaced the misleading compact QWD allocation with two genuinely cross-résumé measures: formula-derived Opponent Value and comparable evidence units.
  • Rebuilt expanded win cards around Opponent Value, Game Grade, and evidence units; the proportional Quality-Win Depth and Index allocations now sit inside the calculation disclosure and are explicitly labeled within-team figures.
  • Clarified that a team’s bounded QWD score is calculated from its complete win portfolio, so an individual win’s proportional allocation cannot be compared with an allocation on a different résumé.
  • Added a universal regression audit confirming that Indiana’s 2025 road win at Oregon and neutral-site win over Ohio State each produce at least 15% more evidence than Texas Tech’s wins over BYU and Utah.
  • Kept NRR v0.37, all five component weights, every checkpoint ranking, and every conference-neutral rule unchanged; this release corrects a public comparison error rather than forcing the model toward named teams.

Still open

  • Evidence units are formula inputs rather than standalone rating points; the complete résumé still determines the bounded Quality-Win Depth component.
  • The v0.37 evidence calibration remains subject to held-out seasons and the live 2026 weekly run.
  • Opponent Value, Game Grade, and evidence units should remain visible together so no single contextual number is mistaken for the complete résumé score.
Version8.2

NRR v0.37 continuous elite-win recalibration

What changed

  • Kept the published 30 / 25 / 20 / 20 / 5 weights and changed only the elite end of Quality-Win Depth evidence.
  • Replaced the unrestricted exponential acceleration above the national midpoint with a continuous, unbounded diminishing-return transform. Every stronger opponent still produces more evidence; there is no hard cap, tier cutoff, or flat maximum.
  • Preserved the existing supporting-win curve below the national midpoint, so the correction does not manufacture extra evidence for weak wins or use a school, conference, or brand label.
  • Regenerated all twelve 2024–25 checkpoints, the downloadable weekly archive, and the machine audit from the same formula implementation.
  • Reduced the ability of one elite result to overwhelm an otherwise ordinary résumé while retaining elite wins as the strongest individual evidence.

Still open

  • The 0.35 elite-evidence diminishing rate remains a public calibration choice requiring held-out seasons and the live 2026 weekly run.
  • Quality-Win Depth remains nonlinear and the published component remains smoothly bounded, even though individual opponent evidence has no hard ceiling.
  • Additional backtesting should evaluate whether the same calibration remains stable when the official Top 25 has a different distribution of elite and fringe teams.
Version8.1

NRR win-evidence hierarchy

What changed

  • Kept the published 30 / 25 / 20 / 20 / 5 formula after testing a 10% Win Control alternative; the larger share frequently elevated dominant teams with weak opponent portfolios rather than correcting them.
  • Renamed the public Win Value label Game Grade to make clear that it is a comparable 0–100 result grade, not the win’s direct rating contribution.
  • Added each win’s actual Quality-Win Depth contribution to the collapsed résumé row, so elite wins no longer appear only modestly better because their nonlinear evidence conversion was hidden.
  • Redesigned expanded win cards around two primary actions—open the opponent résumé or the box score—plus one compact impact strip. Opponent mix, checkpoint math, and the complete explanation now sit behind a Show calculation disclosure.
  • Kept every NRR formula, checkpoint score, ranking, and conference-neutral rule unchanged; this release improves the explanation and hierarchy of the existing calculation.

Still open

  • Win Control remains a deliberately small 5% separator because it adds sustained-game information but can inflate teams that dominate weak opponents if given too much weight.
  • The live 2026 weekly run remains the best next test of whether the five-measure balance generalizes without manual team or conference adjustments.
Version8.0

Saturday mosaic and sitewide editorial polish

What changed

  • Rebuilt twelve of the thirty opening photographs around genuinely different visual ideas: an 1882 team portrait, field storming, rain celebration, snowy tailgate, midnight ritual, campus hill, mountain venue, sunset, fireworks, close game action, and a campus locomotive.
  • Retained the collage rules of thirty distinct schools, thirty distinct photographs, real documentary sources, no Michigan imagery, and a public source-and-license record for every tile.
  • Changed the opening collage from an arbitrary tile grid into a deliberate editorial mosaic with selected focal photographs, brighter image treatment, and less visual competition behind the proposal headline.
  • Condensed the four playoff reward tiers into a two-by-two scoreboard so the opening feels more photographic and less like one oversized blue panel.
  • Rebuilt the global footer into clear Explore, Participate, and Project Record directories; strengthened active navigation, keyboard focus, shared card interaction, and responsive spacing across the site.

Still open

  • The opening collage remains an editorial cross-section rather than a complete catalog of college-football culture.
  • Future imagery should replace—not merely add to—the mosaic and must introduce a new school plus a genuinely different visual idea.
  • Independent legal review of reuse terms, trademarks, and depicted people remains appropriate before commercial launch or monetization.
Version7.9

NRR résumé-card impact transparency

What changed

  • Audited the 2024 Notre Dame loss to Northern Illinois against Georgia’s road loss at Ole Miss. The formula already treats the weak-opponent home upset as substantially worse in the final rating: approximately 11.33 published Index points versus 5.46, without a manual team or conference adjustment.
  • Replaced the misleading compact loss-row severity number with the loss’s marginal hit to the final published Index, where a larger negative number is worse.
  • Added a three-stage loss proof separating pre-curve game severity, Loss Integrity component impact, and final Index impact. The former 32.2 and 35.1 figures remain available only as game-severity inputs rather than being presented as the cross-résumé comparison.
  • Renamed the compact win headline Win Value, defined it as a 0–100 grade where higher is better, and added a three-stage path from Win Value through Quality-Win Depth points to final Index points.
  • Kept version 0.36’s formula, weights, rankings, checkpoint data, and conference-neutral rules unchanged; this release corrects explanation and hierarchy rather than recalibrating results.

Still open

  • Marginal effects remain nonlinear and therefore do not sum exactly across a full résumé.
  • Pre-curve severity remains useful for explaining one game, but final marginal Index effect is the proper comparison across different résumés.
  • The 2026 live weekly run and additional held-out seasons remain necessary calibration evidence.
Version7.8

Homepage collage editorial rebuild

What changed

  • Replaced ten weak or poorly cropped homepage tiles with real college-football photographs centered on packed stadiums, game action, team entrances, and student energy.
  • Removed the culturally sensitive horse-rider image, the isolated rider portrait, the firearm-bearing mascot portrait, the empty locker room, the awkward handler-and-dog frame, and several static mascot or object shots.
  • Added distinct scenes from Notre Dame, Georgia, Utah, Virginia, Wyoming, Alabama, Florida, Tulane, California, and Appalachian State without repeating a school or photograph in the thirty-tile collage.
  • Rechecked every replacement source and license, updated the public credits automatically, and resized the new assets for practical web delivery while preserving clean responsive crops.

Still open

  • The collage remains an editorial cross-section rather than a complete representation of every FBS institution.
  • Future replacements should meet the same standards: clearly football-specific, visually energetic, crop-safe, properly licensed, and free of gratuitous weapons or culturally sensitive mascot imagery.
  • Independent legal review of image reuse terms and depicted marks remains appropriate before monetization.
Version7.7

NRR v0.36 loss-accounting and championship consistency

What changed

  • Kept the published 30 / 25 / 20 / 20 / 5 headline weights while moving schedule-adjusted record internally to 70% achieved record, 20% benchmark overperformance, and 10% slate caliber.
  • Raised the multiplicative additional-loss frequency rate from 20% to 25%, making a third loss materially more damaging without awarding positive credit for a so-called quality loss.
  • Applied qualifying conference-title protection consistently: a first title-game loss to that opponent now counts as half game damage and half a loss in the frequency and record calculations. The loss remains visible and still lowers the résumé.
  • Narrowed peer head-to-head from a 2.75-point bonus across a 2.5-point band to a 0.35-point tiebreak inside a one-point band. A win over one nearby peer can no longer lift that team above unrelated teams with stronger records.
  • Regenerated all twelve checkpoints. At 2024 Selection Sunday, 11–2 Penn State rises from No. 10 to No. 6 and ranks above 9–3 Ole Miss at No. 11; Notre Dame remains No. 4. The audit again reports no school, conference, or manual placement rule.

Still open

  • Loss-frequency strength and half-loss title-game protection remain policy choices requiring additional held-out seasons and the live 2026 weekly run.
  • Penn State’s rise reflects record and loss-accounting corrections, not positive credit for losing to highly ranked teams.
  • The narrow head-to-head tiebreak remains formula-defined but intentionally cannot dominate the full résumé.
Version7.6

NRR v0.35 record-majority calibration and résumé fit

What changed

  • Shifted the published balance to 30% quality-win depth, 25% schedule-adjusted record, 20% opponent-adjusted performance, 20% loss integrity, and 5% win control. The change transfers five points from win depth to the record-based component without removing elite-win evidence.
  • Strengthened the universal record foundation to a 40% floor plus 60% actual winning percentage. Actual record is now the majority input to résumé credibility, with the existing half-loss protection preserved for a first conference-title defeat.
  • Regenerated all twelve 2024–25 checkpoints. At 2024 Selection Sunday, 11–1 Notre Dame rises from No. 6 to No. 4; at the 2025 checkpoint, 10–2 Oklahoma and 10–2 Notre Dame both clear 10–3 Alabama, while 5–7 Florida State falls to No. 42.
  • Expanded the opponent-résumé overlay for laptop widths, gave its Index a dedicated wider column, and removed numeric clipping from the score and component panels.
  • Rebuilt the public CSV and machine audit from the same formula. Every rank remains ordered by the unrounded Index with no school, conference, or manual placement rule.

Still open

  • The 30 / 25 / 20 / 20 / 5 weights and 40 / 60 foundation remain calibration choices requiring additional held-out seasons and the live 2026 weekly run.
  • The proposal protects a first conference-title loss by half inside the formula; that policy remains separate from any team-specific judgment.
  • Résumé overlays prioritize readable comparison evidence rather than complete advanced-stat or drive-level reports.
Version7.5

NRR v0.34 record foundation and résumé interaction pass

What changed

  • Added one universal record foundation after the five weighted measures: 60% floor plus 40% actual winning percentage. Strong opponent evidence remains valuable, but it can no longer erase the season’s actual result.
  • Applied the foundation inside the formula and before publication—not as a manual reorder. The same conference-neutral equation runs for every team and week, with the existing half-loss treatment preserved for a first conference-title defeat.
  • Regenerated all twelve 2024–25 checkpoints. At 2025 Selection Sunday, 5–7 Florida State moves from No. 22 to No. 33, while 10–2 Oklahoma moves above 10–3 Alabama on the formula Index.
  • Added record-foundation and pre-foundation weighted-measure columns to the public CSV and expanded the automated reconstruction audit.
  • Made opponent names inside win and loss cards directly interactive: click to expand the game proof and double-click the name to open the opponent résumé. Refined the résumé and box-score overlays for laptop viewing and replaced misleading tiny loss headlines with each loss’s total damage.

Still open

  • The 60 / 40 record-foundation balance is a public calibration choice requiring additional held-out seasons and the live 2026 weekly run.
  • The résumé model still evaluates accomplishment rather than future team strength and remains an aid—not a replacement claim for the CFP committee.
  • Game overlays currently publish line scores and late-game checkpoints, not full drives, individual statistics, injury context, or video.
Version7.4

NRR v0.33 official-checkpoint opponent value and game archive

What changed

  • Rebuilt Opponent Value around the official weekly CFP checkpoint: 60% official standing, 20% schedule-connected performance, 10% schedule caliber, and 10% schedule-adjusted record. An unranked opponent receives a formula-derived standing baseline capped below the official No. 25 floor.
  • Removed the frozen provisional NRR Index from final Opponent Value, reducing circularity between one team’s NRR score and another team’s Quality-Win Depth while retaining the separate frozen band used only by the narrow peer head-to-head rule.
  • Added large horizontal opponent-résumé overlays to every win and loss card. Readers can inspect the opponent’s Index, record, official rank, five scored measures, best wins, and losses without replacing the résumé already open.
  • Added archived game views with final score, venue, date, quarter-by-quarter scoring, overtime periods when present, late-game checkpoints, and direct links to either team’s résumé.
  • Regenerated and audited all twelve 2024–25 checkpoints. At 2025 Selection Sunday, Indiana records 93.1 Quality-Win Depth and 91.8 schedule-adjusted record; Ole Miss ranks No. 6 with a 76.3 QWD, while Notre Dame ranks No. 12 with 54.0.

Still open

  • The 60 / 20 / 10 / 10 Opponent Value blend and curved official-standing scale remain public calibration choices requiring additional held-out seasons and the live 2026 weekly run.
  • Official CFP standing is intentionally the majority Opponent Value input in the CFP-window model; preseason research previews still require a separately labeled source policy.
  • The quarter-by-quarter archive summarizes scoring flow but does not yet include drives, individual statistics, injuries, or video.
Version7.3

NRR v0.32 nonlinear elite-win evidence

What changed

  • Removed the 55 / 25 / 10 / 7 / 3 portfolio shares. Every distinct defeated opponent now receives the same evidence equation, so a fourth equivalent quality win is worth the same evidence as the first rather than being discounted by order.
  • Added an exponential opponent-evidence curve after the comparable Game Value calculation. The opponent establishes the curved base; sustained game control is limited to a 0.85–1.15 multiplier. One truly elite win can therefore outweigh two fringe Top-25 wins without letting margin manufacture opponent quality.
  • Added evidence from every distinct win before converting the additive total to a bounded 0–100 Quality-Win Depth score through one smooth saturation curve. Lesser wins remain visible but contribute very little; there is no five-win cutoff.
  • Added cumulative Top 5, Top 10, Top 15, Top 20, and Top 25 win counts to every weekly résumé card. Expanded win proof now publishes Game Value, evidence units, share of accumulated evidence, component points, raw-composite points, and final Index points.
  • Regenerated the twelve 2024–25 checkpoints. At 2025 Selection Sunday, undefeated Indiana records 91.8 schedule-adjusted record and 89.0 Quality-Win Depth; BYU records 69.9 and 55.2. Ole Miss ranks No. 6 with 67.5 QWD, above Notre Dame at No. 12 with 47.0.

Still open

  • The 18-point evidence scale, six-unit depth decay, 0.85–1.15 performance band, 40 / 25 / 25 / 10 Opponent Value blend, and peer head-to-head band remain public calibration choices requiring held-out seasons and the live 2026 weekly run.
  • Official Top 5 / 10 / 15 / 20 / 25 counts are public checkpoint context; continuous formula-derived Opponent Value—not a conference or school label—creates scoring evidence.
  • The provisional opponent graph remains frozen after one pass to prevent circular résumé inflation.
Version7.2

NRR v0.31 defining wins, record anchor, and peer result

What changed

  • Rebuilt schedule-adjusted record around 65% achieved record, 25% wins above one fixed national benchmark expectation, and 10% full-slate caliber. The 2025 undefeated Selection Sunday résumé now scores above 90 in SAR instead of being compressed near 80.
  • Reweighted the five-win portfolio to 55 / 25 / 10 / 7 / 3. The two defining wins now supply 80% of Quality-Win Depth, while the public Game Value remains directly comparable when two teams defeat the same opponent at the same venue.
  • Rebuilt Opponent Value as 40% frozen provisional NRR Index, 25% schedule-connected performance, 25% schedule caliber, and 10% schedule-adjusted record. The connection correction moves two points per schedule-caliber point around 50, reducing soft closed-slate inflation without a school or conference label.
  • Added a capped, formula-defined peer head-to-head credit: a direct winner receives 2.75 raw-composite points only when the teams have the same record and sit within 2.5 frozen provisional résumé points. The credit appears in the résumé proof and changes the Index itself; it is not a manual reorder.
  • Changed the weakest-loss credibility rule from a broad linear penalty to a curved factor below Opponent Value 60. Borderline national losses remain negative without being treated like truly catastrophic losses, which still receive the steepest penalty.

Still open

  • The 55 / 25 / 10 / 7 / 3 portfolio, 40 / 25 / 25 / 10 Opponent Value blend, two-for-one schedule connection, 2.5-point peer band, and 2.75-point credit remain public calibration choices requiring held-out seasons and the live 2026 weekly run.
  • NRR v0.31 is a résumé model, not a predictive power rating and not a proposed replacement for the CFP committee.
  • The provisional opponent graph remains frozen after one pass to prevent circular résumé inflation.
Version7.1

NRR v0.30 comparable quality-win values

What changed

  • Removed portfolio-position distortion from individual wins. The five best distinct wins now receive equal 20% shares, so defeating the same opponent at the same venue no longer changes from a first-slot contribution to a fourth-slot contribution.
  • Rebuilt final Opponent Value as 60% frozen provisional NRR Index, 25% actual record, and 15% opponent-adjusted performance. The separate schedule term was removed because schedule context already enters through the frozen provisional Index.
  • Defined every comparable game value as 75% opponent-and-venue value plus 25% sustained game performance. Final margin and Q4, 6:00, and 2:00 checkpoints may separate a dominant result from a narrow escape without allowing portfolio order to manufacture the difference.
  • Changed the résumé-card headline from position-weighted QWD points to the directly comparable Game Value, while retaining the exact 20% component contribution inside the expanded proof.
  • Regenerated all twelve 2024–25 checkpoints. At 2025 Selection Sunday, the Southern California and Oklahoma home wins over Michigan now carry Game Values of 72.0 and 68.7 and QWD contributions of 14.41 and 13.74, rather than 34.05 and 6.04.

Still open

  • The 60 / 25 / 15 Opponent Value mix and 75 / 25 game-value mix remain calibration choices requiring additional held-out seasons and the live 2026 weekly run.
  • Equal shares favor full five-win depth over a position-weighted portfolio; the effect on teams with one elite win and thinner supporting résumés remains an explicit validation question.
  • The provisional Index remains frozen after one pass to prevent circular résumé inflation.
Version7.0

NRR v0.29 elite-win and road-context correction

What changed

  • Rebuilt Quality-Win Depth so the best defeated opponent supplies 50% of the five-win portfolio, followed by 22%, 14%, 9%, and 5%. This prevents several ordinary wins from automatically outweighing one elite result while preserving meaningful depth.
  • Added a frozen first-pass NRR Index to Opponent Value. The final blend is 50% opponent-adjusted performance, 15% full-schedule caliber, and 35% provisional NRR; the one-way sequence captures the opponent’s complete résumé without circular score inflation.
  • Raised maximum road context to six points and neutral context to three, then scaled both continuously by opponent quality. An elite road win receives the full adjustment, while a soft road win cannot manufacture equivalent résumé value.
  • Reduced the separate quality-win control bonus from eight to four game-value points because Win Control already carries margin evidence. Loss Integrity remains on the independent first-pass opponent graph and is unchanged by this Quality-Win Depth revision.
  • Regenerated all twelve 2024–25 checkpoints with no team, school, conference, poll, or post-formula exception. At 2025 Selection Sunday, Ole Miss is No. 7 and Notre Dame No. 8; Ole Miss now leads Quality-Win Depth 70.5–67.5, with the Oklahoma road win contributing 44.58 points versus 37.92 from Notre Dame’s Southern California home win.

Still open

  • The 50 / 22 / 14 / 9 / 5 portfolio and 50 / 15 / 35 Opponent Value blend remain calibration choices requiring held-out seasons and the live 2026 weekly run.
  • The provisional Index is intentionally frozen after one pass; repeated iteration would create circular résumé inflation.
  • Official CFP tiers remain comparison labels, while NRR scoring remains conference-neutral and formula-derived.
Version6.9

NRR v0.28 opponent-value recalibration

What changed

  • Rebuilt Opponent Value as 70% opponent-adjusted performance, 20% full-schedule caliber, and 10% schedule-adjusted record. The formula now distinguishes how good an opponent played, how demanding its entire slate was, and what record it achieved.
  • Reduced the record-driven inflation that could make a soft-schedule opponent appear far more valuable than a difficult-schedule opponent with additional losses. No conference, school, brand, poll, or recruiting label enters the correction.
  • Recalculated every 2024 and 2025 weekly checkpoint. In the 2025 South Florida résumé, the road win at Florida now ranks ahead of the home win over Boise State and immediately behind the road win at North Texas, matching the evidence hierarchy without a team-specific override.
  • Published the v0.28 weekly CSV, audit summary, formula explanation, proposal language, and résumé cards from the same regenerated archive.

Still open

  • The 70 / 20 / 10 Opponent Value balance remains a public calibration choice and should be tested against the live 2026 season and additional held-out years.
  • A difficult schedule can identify a resilient opponent, but it cannot turn a loss-heavy team into a Top-25 win by itself; performance and record still supply 80% of Opponent Value.
  • Quality-win tiers remain time-locked national comparisons, not conference or brand judgments.
Version6.9

Cleanup batch three and year-round coordination

What changed

  • Moved the College Football–NFL Coordination Compact’s page-specific presentation rules out of the oversized global stylesheet and into a route-owned stylesheet, reducing unrelated global code without changing the site, URL, or visual system.
  • Added a six-milestone May-to-May operating cycle covering the shared calendar lock, one preseason athlete-status guide, finalist protections, the January championship handoff, evaluation and return procedures, and a public annual audit.
  • Replaced the implementation framework’s generic negotiation window with the same concrete annual operating cycle and pilot scorecard used by the compact.
  • Retained all twelve current weekly NRR snapshots, personal photographs, and live editorial images. This cleanup does not change the NRR formula, historical ranking order, or public archive.

Still open

  • The current image library remains the largest published payload. A future batch should create carefully reviewed responsive derivatives and modern-format versions without lowering source quality or deleting Alec’s originals.
  • The two companion coordination documents remain current at the principle level; their next formal document release should incorporate the six-milestone operating calendar.
  • Every milestone remains a negotiation proposal requiring agreement among the CFP, NCAA, NFL, NFLPA, conferences, athletes, bowls, venues, and media partners.
Version6.8

Cleanup batch two and same-site coordination

What changed

  • Confirmed the College Football–NFL Coordination Compact as an integrated route inside the existing College Football Reform Project, with the same primary navigation, footer, proposal links, and public domain rather than a separate website.
  • Added regression coverage for the coordination route, its two companion-document downloads, and its entry points from the homepage, proposal, implementation framework, and footer.
  • Consolidated twenty-eight superseded NRR research downloads into one verified legacy archive while keeping the current v0.27 weekly audit and summary available as direct public files.
  • Removed six verified orphaned, non-personal image assets that had no live page, component, metadata, or stylesheet caller. Personal photographs and every currently rendered editorial image were deliberately retained.

Still open

  • The legacy archive preserves the research record, but older direct file URLs are no longer promoted as current public downloads.
  • The local dependency directory remains intentionally untouched because deleting it would slow future development without reducing the published website package.
  • Additional cleanup should continue in measured batches with build, ranking, image-reuse, route, and download regression checks after every material removal.
Version6.7

College Football–NFL Coordination Compact

What changed

  • Added a formal coordination compact without creating a seventh reform card. The compact is integrated into the proposal’s calendar, player-transition, governance, and implementation chapters and published as a separate companion document.
  • Established a negotiated target of Saturday primetime during NFL Wild Card Weekend for the college national championship, with the NFL retaining earlier Saturday inventory plus its Sunday and Monday windows. The site explicitly distinguishes this proposal from the current 2027 calendar.
  • Created One Clean Doorway for professional transition: athletes may explore the draft and receive advice without losing eligibility, may return through a fixed withdrawal deadline, and cross an objective finality line only through a signed NFL player contract, NFL regular-season roster or practice compensation, or NFL regular-season participation.
  • Defined a three-year renewable council compact covering calendar publication, player transition, consent-based health-data portability, verified testing, insurance, degree completion, workload protection, and public reporting.
  • Left the National Résumé Rating formula, historical archive, and published weights unchanged.

Still open

  • No CFP, NFL, NFLPA, media-rights, bowl, venue, or athlete-representation agreement currently guarantees the proposed Saturday window.
  • The precise legal effect of draft entry, contracts, compensation, roster status, and return to college requires labor, eligibility, collective-bargaining, and antitrust review.
  • Any NFL or NFLPA contribution to health, education, insurance, or longitudinal research must be negotiated and independently audited.
Version6.6

NRR v0.27 opponent tiers and game-level contribution proof

What changed

  • Added the national Top-25 Opponent Value line at each weekly checkpoint as the public quality-win standard. Every qualifying win is now labeled Top 5, Top 10, Top 15, Top 20, or Top 25 from the same time-locked national order.
  • Allowed a modest four-point road or two-point neutral adjustment to lift only a just-outside opponent across that public threshold. Ordinary sub-Top-25 wins appear only as separately labeled supporting depth, while a borderline No. 25 result receives only borderline credit.
  • Expanded each Saturday résumé card with the opponent tier, qualification basis, depth slot and share, checkpoints, component contribution, raw NRR points, and final Index effect for every scored win.
  • Applied the same opponent tiers to every loss and published each loss's marginal integrity and Index effect, while retaining the rule that a loss can never earn positive quality credit.
  • Preserved the v0.26 continuous ranking calculation while regenerating and auditing the complete 2024–25 weekly archive under NRR v0.27 with no team, conference, poll, or post-formula adjustment.

Still open

  • The weekly Top-25 threshold is a transparent communication choice and should be tested in the held-out 2026 season.
  • Road and neutral qualification bonuses require sensitivity review so contextual credit does not create an overly broad quality-win pool.
  • Opponent tiers explain résumé context; they do not claim that the NRR should replace the CFP committee.
Version6.5

NRR v0.26 published balance and record calibration

What changed

  • Changed the published NRR balance to 35% quality-win depth, 20% schedule-adjusted record, 20% opponent-adjusted performance, 20% loss integrity, and 5% win control. The model now gives a damaged loss profile the same published weight as schedule-adjusted record and opponent-adjusted performance.
  • Made achieved record the clearer center of schedule-adjusted record. The component now blends 60% reliability-adjusted record, 25% full-slate caliber, and 15% wins above benchmark expectation, while continuing to penalize repeated losses without using school or conference labels.
  • Reframed the public model controls as a private what-if workspace. Readers may test a different weighting balance in their browser, restore the published default in one action, and cannot change the official NRR archive or proposal.
  • Regenerated the complete 2024–25 weekly archive under NRR v0.26 and retained strict Index-order, component-range, and no-override auditing across every public checkpoint.

Still open

  • The 35 / 20 / 20 / 20 / 5 balance remains a published proposal choice rather than a settled scientific constant and should be evaluated against the held-out 2026 season.
  • The public what-if workspace changes component weights only; it does not rewrite component definitions, official comparisons, or historical evidence.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version6.4

Tradition collage and stadium-entry refinement

What changed

  • Replaced remaining generic crowd, band, and historic filler with recognizable college-football traditions across thirty distinct school subjects. The atlas contains no repeated photograph, no repeated school, and no Michigan imagery.
  • Expanded the final visual range with live mascots, signature entrances, stadium rituals, championship celebrations, game action, and town traditions while retaining one source-and-license record for every tile.
  • Applied localized, soft blur directly over visible commercial marks in the real stadium panorama. The treatment adds no replacement label, invented stadium name, or promotional copy.
  • Redesigned the pre-entry tunnel screen as a focused invitation with a restrained glass panel, concise interaction cues, and one primary Step onto the field action while preserving the user-controlled panorama after entry.

Still open

  • The thirty-image atlas is an editorial cross-section rather than a complete representation of every FBS institution or tradition.
  • Mobile crop behavior and the panorama blur mask should receive a final device-level editorial review before commercial launch.
  • Independent legal review of image reuse terms and depicted trademarks remains appropriate before monetization.
Version6.3

Homepage tradition atlas and navigation consolidation

What changed

  • Rebuilt the opening collage as a thirty-school atlas with thirty distinct source photographs, no repeated school subject, no Michigan imagery, and no generic crowd-wave tile.
  • Expanded the visual range across mascots, entrances, tailgates, stadiums, student rituals, historic teams, game action, town traditions, and fan celebrations while limiting marching-band imagery to two photographs.
  • Moved the long-form Analysis collection out of primary navigation and renamed it Essays in the footer. The Evidence ledger remains a primary destination because it uniquely documents readiness, safeguards, source quality, and unresolved validation work.
  • Centralized collage subjects, crop positions, source links, creator credits, and licenses in one audited manifest, then added an automated uniqueness and category-diversity check.
  • Updated the public image-credit ledger to document the collage rule and the complete thirty-image source set.

Still open

  • The collage is an editorial cross-section rather than a complete representation of every FBS institution or tradition.
  • Future photographs must add a new school subject and a genuinely new visual idea; additions should not weaken the current source and license standard.
  • Independent legal review of reuse terms remains appropriate before any commercial launch or monetization.
Version6.2

NRR v0.25 formula lock and persistent archive

What changed

  • Moved the season and release controls into a persistent switchboard that stays with the ranking table, team card, comparison views, and formula controls as the reader scrolls. Desktop and mobile readers can change year or week without returning to the top of the page.
  • Removed the post-formula review and Index-materialization layer. One continuous raw composite now produces the published Index, rank, weekly movement, table order, team card, and downloadable archive; there are no school, conference, poll, or post-score overrides.
  • Raised component precision and retained the unrounded raw composite in the audit export, eliminating artificial one-tenth score staircases while keeping the public Index readable and exactly reconstructable.
  • Rebuilt schedule-adjusted record around 60% achieved record, 25% full-slate caliber, and 15% wins above a fixed benchmark expectation so weak schedules cannot manufacture elite record scores.
  • Published the v0.25 weekly CSV and machine-readable audit. All 12 releases and 1,620 team-checkpoints pass with zero formula-order errors, zero invalid or saturated components, and zero post-formula adjustments.

Still open

  • The public archive intentionally begins with the first CFP release and currently covers 2024 and 2025; earlier-season calculations would be research previews rather than official NRR releases.
  • The v0.25 component calibrations remain provisional and require a held-out 2026 weekly audit before adoption.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version6.1

NRR v0.24 archive and integrity audit

What changed

  • Rebuilt the Rankings page around the reader’s actual tasks: jump to the complete current Top 25, browse a combined year-and-release archive, inspect the model, or read the method record.
  • Replaced the separate season and oversized week controls with one compact archive navigator for 2024 and 2025, all six public releases per season, previous/next movement, and a mobile release menu. Every current poll now shows all 25 teams without a hidden second page.
  • Corrected Loss Integrity’s overlapping weak-loss deductions. Version 0.24 uses one bounded continuous weak-opponent adjustment, reciprocal aggregate-damage decay, a multiplicative additional-loss factor, and a bounded weakest-loss credibility factor; it preserves severe exceptional-loss consequences without forcing ordinary multi-loss résumés to zero.
  • Added a permanent release audit covering 12 checkpoints and 1,620 team-checkpoints. It rejects non-finite or saturated scored components, invalid Loss Integrity values, nonsequential ranks, and any table that is not strictly ordered by Final NRR Index.
  • Published the v0.24 weekly CSV and machine-readable audit summary while preserving the 40 / 20 / 20 / 15 / 5 proposal weights and every conference-neutral input rule.

Still open

  • The public archive intentionally begins with the first CFP release and currently covers 2024 and 2025; earlier-season calculations would be research previews rather than official NRR releases.
  • The revised loss curve needs additional held-out seasons and week-to-week stability testing before its calibration should be considered settled.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version6.0

NRR v0.23 weekly checkpoint archive

What changed

  • Preserved the published 40 / 20 / 20 / 15 / 5 formula, every scored component, and every universal review rule from v0.22.
  • Rebuilt the model at twelve time-locked checkpoints: all five Tuesday CFP releases plus Selection Sunday in both 2024 and 2025. Each run uses only games completed through the preceding Saturday, preventing future results from leaking into an earlier week.
  • Established a public release policy: NRR begins with the season’s first official CFP ranking and updates with each committee release. Any earlier-season calculation is labeled a research preview rather than an official NRR release.
  • Added season and week controls, same-date CFP comparisons, movement since the prior NRR, and a six-checkpoint path for every selected team.
  • Published a downloadable weekly audit file and added validation that rejects non-finite component values. The audit also corrected missing late-game checkpoint fields and a floating-probability boundary without changing the formula.

Still open

  • The archive currently covers two complete seasons; additional held-out seasons are needed before any calibration rule should be treated as settled.
  • Earlier-season research previews, lineup-aware tests, and formal week-to-week stability thresholds remain open.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.9

NRR v0.22 record-burden correction

What changed

  • Kept the published 40 / 20 / 20 / 15 / 5 component weights and every scored formula unchanged.
  • Added a conference-neutral record-burden review: when a team with one additional loss is within five raw-composite points, it must retain materially stronger opponent evidence to remain ahead of the better record. The evidence index uses 60% quality-win depth and 40% opponent-adjusted performance; a 14-point advantage is required to defeat the record burden.
  • Required the existing opponent-credibility gate before the catastrophic-loss review can promote a team. A soft résumé can no longer gain places merely because nearby teams carry damaged Loss Integrity.
  • Published the v0.22 five-season audit. In 2025 Arizona and Michigan move ahead of four-loss Louisville and Arizona State; North Texas falls outside the Top 25; Navy moves to No. 25. The first sixteen positions remain unchanged.
  • Applied no school, conference, poll, brand, recruiting, or roster adjustment.

Still open

  • The five-point neighborhood, 60 / 40 opponent-evidence blend, and 14-point materiality threshold are public calibration choices requiring weekly and held-out validation.
  • Record burden is intentionally narrow: it compares teams separated by exactly one loss and does not establish a rigid loss-count sorting rule.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.8

NRR v0.21 catastrophic-loss correction

What changed

  • Kept the published 40 / 20 / 20 / 15 / 5 component weights, every quality-win, schedule-adjusted-record, opponent-adjusted-performance, and win-control formula, and the v0.20 Final Index reconciliation unchanged.
  • Added a universal continuous catastrophic-loss deduction below Opponent Value 50. Each value point below 50 removes six additional loss-integrity points after the ordinary damage curve, while Loss Integrity remains bounded from zero to 100.
  • Required at least 25 Loss Integrity points before a team can receive the record-review benefit. Below 15, a nearby team may pass only with at least ten more integrity points, no more than one additional loss, and no more than a six-point raw-score gap.
  • Published the v0.21 five-season audit. In 2024 Notre Dame’s home loss to Northern Illinois reduces its Loss Integrity from 42.3 to 9.8 and moves it from No. 2 to No. 4; Oregon remains No. 1, Georgia No. 2, and Ohio State No. 3.
  • Applied no school, conference, poll, brand, recruiting, or roster adjustment.

Still open

  • Opponent Value 50, the six-for-one deduction, and the 25 / 15 review gates are public calibration choices requiring weekly and held-out validation.
  • The loss review deliberately distinguishes a truly weak opponent from an ordinary bad loss, but it does not attempt to measure injuries, officiating, or lineup strength.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.7

NRR v0.20 index integrity correction

What changed

  • Corrected the public Index contradiction identified in the 2025 Ole Miss case: a reviewed team can no longer appear above another team while displaying a lower final Index.
  • Preserved every v0.19 component formula, the published 40 / 20 / 20 / 15 / 5 weights, and the universal record, command, and head-to-head review rules.
  • Separated the pre-review Base Index from the Final NRR Index. Every review effect is now a disclosed adjustment rather than a hidden ordering rule.
  • Defined the review adjustment as the minimum one-tenth of an Index point required to express the reviewed order. The public table now sorts strictly from highest to lowest Final NRR Index.
  • Expanded the v0.20 five-season CSV with Base Index and review-adjustment fields so every final score can be independently reconstructed.

Still open

  • The one-tenth-point reconciliation is an ordering calibration, not a new football-performance component; its long-term use requires weekly and held-out validation.
  • Review thresholds remain provisional and are applied without conference, school, poll, brand, recruiting, or roster criteria.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.6

NRR v0.19 record-and-command review

What changed

  • Kept all five v0.18 component formulas and the published 40 / 20 / 20 / 15 / 5 weights unchanged.
  • Added a universal record-integrity review after the raw composite. A team with credible opponent evidence—OAP of at least 70 or quality-win depth of at least 50—can move through a nearby cluster when it owns a materially better record; a one-loss advantage is limited to a six-point raw-score neighborhood.
  • Added a same-record Command Index for teams within four raw points. A team moves ahead only with a six-point command advantage measured by 75% opponent-adjusted performance and 25% win control.
  • Kept head-to-head as the last local review and added no conference, school, poll, brand, recruiting, or roster adjustment.
  • Published the v0.19 five-season audit. In 2024 Indiana moves from No. 17 to No. 7 and Tennessee from No. 12 to No. 8; in 2025 one-loss Ole Miss moves to No. 7 while three-loss Alabama moves to No. 12.

Still open

  • The six-point record neighborhood, four-point same-record neighborhood, and six-point Command Index threshold require weekly and held-out validation.
  • The opponent-credibility gate is designed to avoid automatic soft-schedule promotion; its 70 OAP and 50 quality-depth thresholds remain public calibration choices.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.5

NRR v0.18 command-and-scale calibration

What changed

  • Preserved the published 40 / 20 / 20 / 15 / 5 weights, the 35 / 25 / 20 / 12 / 8 quality-win depth shares, and the v0.17 schedule-adjusted-record, opponent-adjusted-performance, and loss-integrity formulas.
  • Added winning-side Q4, 6:00, and 2:00 checkpoints. Final margin supplies 70% of win control and the three checkpoints supply 30%; only a qualifying quality win can receive the separate control bonus, which is capped at five component points and never penalizes a close win.
  • Introduced a public NRR Index of 20 + 0.85 × raw composite, capped at 100. The transformation is monotonic and therefore cannot change any ranking, comparison, or head-to-head policy result; undefeated 2025 Indiana now publishes at 91.6.
  • Published a new five-season v0.18 audit and retained the older v0.17 download. In the 2025 checkpoint, 20 of the official Top 25 remain in the field, 18 are within three positions, and the average rank gap is 3.3.
  • Documented the 2024 Alabama–Penn State–Tennessee ordering as an open calibration case rather than forcing a conference, brand, or multi-position head-to-head correction.

Still open

  • The 2024 model places Alabama No. 6, Penn State No. 7, and Tennessee No. 12. Alabama’s five-win portfolio currently outweighs its near-zero loss-integrity score; that balance requires weekly and out-of-sample review.
  • The checkpoint bonus is intentionally small, but its five-point cap and late-game weights remain provisional until a broader play-level validation is complete.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.4

National Résumé Rating v0.17 quality-depth correction

What changed

  • Kept the published 40 / 20 / 20 / 15 / 5 component weights and every non-quality-win formula unchanged.
  • Rebalanced the five distinct win-depth shares from 50 / 20 / 15 / 10 / 5 to 35 / 25 / 20 / 12 / 8. The best win remains the largest result, but the rest of the résumé now supplies 65% of quality-win depth.
  • Applied the rule universally without a conference, school, brand, or poll adjustment. BYU’s 2024 road win over SMU still leads its evidence, but connected Big 12 depth can no longer ride behind that result as heavily.
  • Published the v0.17 five-season audit. In 2024, BYU moves from No. 6 to No. 9, Arizona State remains directly ahead after winning the head-to-head meeting, and Boise State moves from No. 11 to No. 13.
  • The five-season comparison improves in four checkpoints on at least one of Top-25 overlap, average rank gap, or correlation; the 2024 correlation rises from 0.82 to 0.84 while the average rank gap improves from 6.1 to 6.0.

Still open

  • BYU remains eight positions above the official 2024 CFP order, so its SMU result, loss profile, and schedule network remain a disclosed calibration case rather than a solved conclusion.
  • James Madison is No. 26 in the 2025 final-selection audit. The model neither grants automatic Group of Five access nor imposes a Group of Five penalty; weekly and out-of-sample validation remain required.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.3

National Résumé Rating v0.16 marquee-win correction

What changed

  • Kept the published 40 / 20 / 20 / 15 / 5 component weights unchanged while recalibrating the evidence inside quality-win depth and loss integrity.
  • Replaced the linear quality transformation with a normalized S-curve above the universal 35-point floor. The curve is centered at Opponent Value 55 so elite wins are no longer compressed toward ordinary qualifying wins.
  • Changed the five distinct win-depth shares to 50 / 20 / 15 / 10 / 5. The best win now supplies half of the component while four additional wins continue to test depth.
  • Separated loss severity from loss frequency. The reciprocal severity score remains nonnegative, and each loss after the first removes ten integrity points.
  • Restricted head-to-head to one local neighboring-team comparison within eight composite points and one loss. The policy can settle a close pair but cannot pull a team across an unrelated cluster.
  • Published the v0.16 five-season audit. Ole Miss rises to No. 8 in 2025 with a 60.3 quality-win-depth score, while 2024 Boise State falls to No. 11 rather than benefiting from soft closed-schedule wins.

Still open

  • The S-curve midpoint, depth shares, and additional-loss deduction are public calibration choices that require weekly snapshots, additional seasons, and held-out validation.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.2

National Résumé Rating v0.15 soft-schedule correction

What changed

  • Kept the published default at 40% distinct quality-win depth, 20% schedule-adjusted record, 20% opponent-adjusted performance, 15% loss integrity, and 5% win control.
  • Rebuilt schedule-adjusted record around 55% wins above fixed benchmark expectation, 25% average opponent caliber, and 20% record rarity. A fixed 12-point-per-win scale replaces the variance divisor that could exaggerate predictable soft schedules.
  • Changed Opponent Value to 80% opponent-adjusted performance and 20% corrected schedule-adjusted record, then added a universal 35-point floor before defeated opponents create quality-win credit. No school or conference label enters either rule.
  • Simplified the public model to the five scored measures plus championship policy; schedule and away diagnostics remain in the downloadable audit but no longer clutter the formula controls.
  • Published the v0.15 five-season audit. Boise State settles near No. 9 and Tennessee rises to No. 7 in the 2024 checkpoint, while the 2025 top six remain unchanged.

Still open

  • The 35-point quality threshold, 12-point expectation scale, and 55 / 25 / 20 record blend are public calibration choices that require weekly snapshots, additional seasons, and held-out validation.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.1

National Résumé Rating v0.14 weighting and loss-integrity calibration

What changed

  • Restored the published default to 40% distinct quality-win depth, 20% strength of record, 20% opponent-adjusted performance, 15% loss integrity, and 5% win control; schedule strength and away performance remain supporting diagnostics.
  • Replaced the negative-floor loss score with a nonnegative reciprocal decay: 100 ÷ (1 + aggregate loss damage ÷ 50). This preserves opponent, margin, venue, late-game, weak-opponent, and repeat-loss evidence without collapsing ordinary multi-loss résumés to -15.
  • Kept the SOR correction isolated: 70% wins above benchmark expectation, 30% record rarity, plus SOR-only repeat and extra-loss deductions; no conference-specific adjustment was added.
  • Published the v0.14 five-season CSV and aligned the Ranking Lab, Formula Explorer, proposal copy, and home-page weights to the same default.

Still open

  • The 50-point aggregate damage decay is a public calibration choice that needs weekly snapshots, more seasons, and held-out validation.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version5.0

National Résumé Rating v0.13 strength-of-record correction

What changed

  • Kept every published weight and every component other than strength of record unchanged: 25% opponent-adjusted performance, 25% strength of record, 20% distinct quality-win depth, 25% loss integrity, and 5% win control.
  • Rebuilt strength of record around wins above a fixed benchmark’s expected record (70%) plus record rarity (30%). A team cannot create an elite SOR merely by piling up wins over opponents the benchmark was already expected to beat.
  • Restored loss integrity, opponent value, quality wins, win control, and all other non-SOR calculations to the v0.12 behavior. The repeat-opponent and extra-loss deductions stay inside SOR; loss integrity does not receive a new penalty.
  • The 2024 audit now places Oregon No. 1, Boise State No. 2, Penn State No. 3, and Notre Dame No. 4; Indiana and Tennessee remain above Clemson. The 2025 audit keeps BYU above Utah through the bounded head-to-head rule.
  • Published the v0.13 five-season CSV and kept the interactive Ranking Lab, Formula Explorer, proposal text, and backtest on the same conference-neutral logic.

Still open

  • The weak-schedule calibration is a public hypothesis that needs weekly snapshots, more seasons, and held-out validation before adoption.
  • The benchmark rating and uncertainty scale remain fixed across teams and conferences, but their sensitivity should be reported annually.
  • The NRR remains an independent résumé aid, not a claim that it should replace the CFP committee.
Version4.9

National Résumé Rating v0.12 résumé-first reset

What changed

  • Rebuilt the public default around five scored measures: 25% opponent-adjusted performance, 25% strength of record, 20% distinct quality-win depth, 25% loss integrity, and 5% win control.
  • Moved loss integrity into the score with one universal opponent-relative rule: continuous margin, late-game checkpoints, venue, repeat-loss context, and a bounded weak-opponent modifier below Opponent Value 75. No quality-loss bonus and no conference-specific adjustment are used.
  • Added a bounded head-to-head tiebreak. A direct winner can move above a close résumé only when the teams are within four ranking slots, within eight composite points, and within one loss; the rule never acts as a league or brand proxy.
  • The 2024 audit now places Oregon No. 1, Boise State No. 2, Penn State No. 3, and Notre Dame No. 4; Indiana and Tennessee remain above Clemson. The 2025 audit places BYU above Utah after accounting for their direct result.
  • Published the v0.12 five-season CSV and kept the interactive Ranking Lab on the same tiebreak and score logic.

Still open

  • The 50-point weak-opponent ceiling is a universal calibration choice, not a punishment for any conference. It requires weekly sensitivity tests, additional seasons, and held-out validation.
  • A five-season final-selection audit is directional evidence, not proof that NRR should replace the CFP committee. The largest gaps remain public for review.
  • Head-to-head is deliberately bounded: a direct result can settle a close pair, but cannot erase a materially stronger overall résumé.
Version4.8

National Résumé Rating v0.11 structural recalibration

What changed

  • Replaced the conference-anchor correction with a conference-blind national play-level model. A weighted ridge fit solves every offense and defense simultaneously from meaningful scrimmage-play EPA; conference names, committee rankings, brands, and recruiting never enter.
  • Published explicit blowout restraints: every first-half scrimmage play remains, third-quarter plays remain while the score is within 28, and fourth-quarter or overtime plays remain while it is within 21. Kneel-downs and nullified plays are excluded. Every historical extract is also locked to its selection-day cutoff so future bowl or playoff plays cannot leak into the audit.
  • Rebalanced the public formula to 45% opponent-adjusted performance, 30% strength of record, 20% five-win quality depth, and 5% win control. Strength of record now blends 60% actual win percentage with 40% fixed-benchmark schedule rarity and removes eight points for an additional loss to the same opponent.
  • Removed loss impact from the scored formula while preserving the full diagnostic. Outcomes, opponent value, margin, and game state already enter strength of record and opponent-adjusted performance; a third scored charge was distorting the result.
  • In the 2024 audit, Texas is No. 6 and SMU No. 7; Tennessee No. 8, Ole Miss No. 12, and Alabama No. 14 all clear BYU at No. 17. In 2025, Alabama is No. 9, Ole Miss No. 13, and BYU No. 14. No conference-specific rule or adjustment produced those changes.

Still open

  • Version 0.11 deliberately accepts lower agreement with the committee. The model is an independent résumé aid, not a committee-copying exercise.
  • The ridge regularization level, eight-point repeat-opponent deduction, and published score-state filters require sensitivity tests at weekly checkpoints, seasons outside 2021–2025, and true held-out samples.
  • Play-level data still cannot identify every substitution or prove precisely when each game became noncompetitive. A lineup-aware garbage-time audit remains open.
Version4.7

National Résumé Rating v0.10 loss calibration

What changed

  • Added public play-by-play game state at three checkpoints for every audited loss: score entering the fourth quarter, score with six minutes remaining, and score with two minutes remaining. The same 30 / 35 / 35 checkpoint blend applies to every team.
  • Recalibrated loss damage by opponent value without awarding quality-loss credit. Every loss retains a positive base; weak opponents add up to 25 damage points, and a bounded negative tail prevents a severe bad-loss profile from disappearing at zero.
  • The 2025 audit now grades Oregon’s loss impact at 74.3 and Texas Tech’s at 72.8. Oregon remained within one score of undefeated No. 1 Indiana through six minutes; Texas Tech was closer late but lost to a materially weaker Arizona State team.
  • Tulane moves from NRR No. 20 to No. 23 after its 48–26 loss at UTSA is fully priced: down 19 entering Q4, 15 at 6:00, and 15 at 2:00. The 2025 audit retains 24 of the CFP Top 25, with 20 teams within three positions and a 2.4-position average gap.
  • Removed every Texas Tech-specific logo override. Rankings and Bracket Lab now use the shared clean Double T-only asset with no caption embedded beneath it.

Still open

  • Three score checkpoints improve context but do not prove when starters left, which players were on the field, or every possession became noncompetitive.
  • The five-season audit is calibration evidence, not proof that v0.10 is final. Weekly snapshots, additional seasons, and out-of-sample bias checks remain required.
  • The opponent-calibrated base, 30 / 35 / 35 checkpoint blend, 25-point weak-loss cap, and −15 loss-impact floor require sensitivity testing before adoption.
Version4.6

National Résumé Rating v0.9 and Saturday selection-room redesign

What changed

  • Replaced the 35-point margin ceiling with a continuous exponential curve. Every additional point still changes the result, but its marginal value steadily shrinks, reducing incentives to chase late scores.
  • Added public quarter-by-quarter linescores to all five audited seasons. Loss-margin damage is reduced when a team led, was tied, or remained within one score entering the fourth; the base loss, venue, weak-opponent, repeat-loss, and title-game rules remain intact.
  • Recalibrated the five-season audit and published a v0.9 CSV. At the 2025 checkpoint, Oregon remains NRR No. 6 versus CFP No. 5, while its loss-impact score rises from 70.0 to 73.5 because it trailed Indiana by seven entering the fourth. The Top-25 overlap remains 24 of 25, 21 teams land within three positions, and the average gap is 2.2.
  • Rebuilt the Rankings page as a restrained Saturday selection room using three real, credited tradition photographs, a game-program visual language, a film-room discrepancy review, and a laminated résumé card.
  • Removed the Texas Tech-only row treatment that enlarged the Double T and forced its name below the mark. Texas Tech now uses the same aligned logo-and-name row as every other team.

Still open

  • Quarter scores establish fourth-quarter context but do not reveal substitutions or exact garbage-time possessions. Play-level win probability and non-garbage-time control remain the next validation layer.
  • The five-season audit remains calibration evidence, not proof that v0.9 is final. Weekly snapshots and seasons outside 2021–2025 are still required.
  • Team Strength remains a separate results-based support model and requires predictive validation before broader use.
Version4.5

National Résumé Rating v0.8 and New Year’s Six emphasis

What changed

  • Shifted the published NRR default from 40 / 40 / 10 / 10 to 35% quality wins, 40% strength of record, 15% loss impact, and 10% win control so the complete record remains primary and an additional loss is harder to offset with a strong win collection.
  • Limited conference-title rating protection to a first loss to that opponent. A repeated loss to the same opponent receives full ordinary loss damage, the five-point repeat deduction, and no SOR protection; playoff-path protection remains separate and intact.
  • Regenerated the five-season audit and v0.8 CSV. At the 2025 checkpoint, Ole Miss is No. 7, Notre Dame No. 8, and BYU No. 9; 24 of 25 committee teams appear in the NRR Top 25, 21 land within three positions, and the average gap is 2.2.
  • Replaced the institution-specific home hero collage with a neutral, CSS-built night-stadium field and retained the headline “Keep the traditions. Fix the incentives.”
  • Elevated the New Year’s Six into a named core reform and moved its complete six-bowl playoff stage directly beneath the home-page overview.

Still open

  • The five-season audit is a calibration check, not proof that v0.8 is final; weekly snapshots and additional out-of-sample seasons remain required.
  • BYU’s No. 9 versus CFP No. 12 is now within the proposal’s explanation band, but its opponent values and the cross-conference Team Strength anchor remain subjects for continued review.
  • The proposed New Year’s Six working set is not an announced CFP contract, official future site assignment, or guarantee of participation.
Version4.4

National Résumé Rating v0.7

What changed

  • Added an opponent-adjusted cross-conference anchor to 20% of Team Strength Rating so a closed conference schedule cannot validate itself entirely through internal results. Every conference uses the same evidence rule; no league receives a preset bonus or penalty.
  • Raised the quality-win road bonus from six to eight points while retaining three points for a neutral-site win and the best-three-distinct-opponents rule.
  • Added bad-loss damage when the defeated opponent’s Opponent Value falls below 75, capped that addition at 15, and added five points when a team loses to the same opponent again. Strong opponents still create no quality-loss reward.
  • Gave a required conference-title loss half effect in SOR and half ordinary loss-severity damage while keeping the loss visible in the record and preserving any repeat-loss addition.
  • Regenerated the five-season audit and v0.7 CSV. At the 2025 checkpoint, NRR now places Indiana No. 1, Ohio State No. 2, and Texas Tech No. 3; 23 of 25 committee teams appear in the NRR Top 25, 22 land within three positions, and the average gap is 2.1.

Still open

  • BYU remains NRR No. 7 versus CFP No. 12. The repeat-loss rule lowers its loss-impact score, but its SOR and distinct wins remain strong enough to keep this a disclosed audit case.
  • Win control already measures whether victories were close or convincing. It favors Texas Tech over Ohio State in the 2025 checkpoint, so adding a second consistency measure would duplicate evidence rather than explain the ordering.
  • The 20% cross-conference anchor, Opponent Value 75 bad-loss threshold, eight-point road bonus, and half title-loss SOR treatment require weekly, out-of-sample, and sensitivity testing before adoption.
Version4.3

National Résumé Rating v0.6

What changed

  • Rebuilt quality wins around an Opponent Value that blends 70% Team Strength Rating with 30% opponent record accomplishment, while retaining the road and neutral bonuses.
  • Limited the quality-win average to three distinct defeated opponents so a rematch cannot occupy two of the three evidence slots.
  • Replaced the field-relative SOR percentile with a fixed probability scale: 100 times the square root of one minus the probability that the benchmark team would match the record.
  • Set quality wins and SOR to 40% each and reduced loss impact and win control to 10% each. A title-game loss receives half normal loss-severity damage while remaining fully present in the actual record and SOR.
  • Regenerated the five-season audit and v0.6 CSV. At the 2025 checkpoint, 20 of 25 committee teams land within three positions and the average gap falls to 2.3; Indiana leads Texas Tech 91.7–87.8, while BYU remains a disclosed five-position audit case.

Still open

  • BYU’s NRR No. 7 versus CFP No. 12 requires a direct audit of the separate Team Strength Rating and schedule network; no conference-specific correction has been applied.
  • Conference-level error, opponent-value stability, weekly snapshots, and seasons outside 2021–2025 remain required tests before adoption.
  • The 70/30 opponent-value blend, square-root SOR calibration, and half title-loss severity are public starting hypotheses rather than permanent constants.
Version4.2

National Résumé Rating v0.5

What changed

  • Replaced the hard 21-point margin cap with a diminishing curve: full separation value through seven, reduced value through 21, sharply reduced value from 22 through 35, and no additional value beyond 35.
  • Moved venue value inside quality wins: a road victory adds six points and a neutral-site victory three before the best three wins are averaged. The separate away-performance line remains visible but unscored, preventing a second reward for the same result.
  • Raised win control from 10% to 15% and removed the separate 5% road-and-neutral score. The published formula is now 40% strength of record, 30% venue-adjusted quality wins, 15% loss impact, and 15% win control.
  • Clarified conference-championship handling: NRR publishes regular-season and final snapshots, the title game updates final NRR, and the participant’s pregame playoff-path tier—not the evidence—remains protected.
  • Regenerated the five-season audit and v0.5 CSV. The 2025 checkpoint matches 22 of 25 CFP teams, places 16 within three positions, and moves BYU to NRR No. 12, equal to its CFP position.

Still open

  • Diminishing margin bands reduce but cannot entirely eliminate strategic incentives around late scoring; future play-by-play data should test non-garbage-time control measures.
  • The venue bonuses and 35-point ceiling require sensitivity testing at weekly checkpoints and on seasons outside the current 2021–2025 sample.
  • The regular-season versus final NRR comparison should be added as a first-class historical view once every title-game checkpoint is separately reconstructed.
Version4.1

National Résumé Rating v0.4

What changed

  • Moved strength of record to the largest single share at 40%, while keeping quality wins as the direct evidence of whom a team actually defeated.
  • Replaced the quality-loss reward with loss impact: every loss subtracts a base, margin, and venue penalty, and no opponent-strength rating can add points for losing.
  • Added a modest win-control measure that captures capped winning margin without importing opponent quality a second time.
  • Set the published default to 40% strength of record, 30% quality wins, 15% loss impact, 10% win control, and 5% road & neutral record; schedule strength and conference title remain supporting diagnostics.
  • Regenerated the five-season audit and published the v0.4 CSV. The 2025 checkpoint now matches 23 of 25 CFP teams, with 18 within three positions and a 3.0-position average gap; BYU moves to NRR No. 13 versus CFP No. 12.

Still open

  • A five-season checkpoint is a useful audit, not proof of validity. Weekly AP and CFP snapshots, out-of-sample seasons, bias checks, and gaming tests remain required.
  • Loss impact deliberately explains severity rather than opponent quality. A future review can test whether its base and margin penalties are calibrated without reintroducing quality-loss credit.
  • The Team Strength Rating remains a separate results-based support model and should receive its own predictive validation before it is used for anything beyond context.
Version4.0

National Résumé Rating v0.3

What changed

  • Removed strength of schedule from the scored formula and retained it as a visible supporting diagnostic, eliminating a direct second reward for opponent difficulty already captured by strength of record.
  • Removed the separate conference-title bonus because the championship result already affects record, quality wins, loss profile, and game control; title status remains visible and available for access or tiebreak rules outside NRR.
  • Rebuilt strength of record around a fixed top-25 benchmark, replaced general opponent quality with the top three opponents actually defeated, and added an explicit loss profile with margin, location, multiple-loss, and repeat-opponent penalties.
  • Used BYU’s 2025 résumé as a documented failure case: the revision moves BYU from NRR No. 5 to No. 9, compared with CFP No. 12, while improving the 2025 average committee gap from 3.44 to 2.40 positions.
  • Added an audit queue for the largest NRR–CFP discrepancies, an exact team-level weighted-points calculation, the logo-only Texas Tech mark, and a downloadable v0.3 five-season backtest.

Still open

  • Version 0.3 improves average five-season rank gap, within-three agreement, and correlation but slightly reduces average Top-25 overlap; the model is intended to measure résumé accomplishment rather than copy the committee.
  • James Madison’s 2025 NRR No. 18 versus CFP No. 24 remains a prominent review case and requires opponent-level diagnosis before another formula revision.
  • Weekly checkpoints, out-of-sample seasons, lower-division opponent data, and formal stability and bias audits remain necessary before adoption.
Version3.9

NRR formula audit and explainer

What changed

  • Revised the proposed National Résumé Rating default from 30 / 25 / 15 / 10 / 10 / 5 / 5 to 25 / 30 / 15 / 10 / 10 / 5 / 5, giving wins, losses, and strength of record the largest share while reducing overlapping opponent-quality credit.
  • Tested more record-heavy alternatives across the 2021–2025 final-selection checkpoints; the narrower five-point revision slightly improved Top-25 overlap, within-three-position agreement, average rank gap, and rank correlation, while the tested 40% results version performed worse.
  • Added a selectable seven-component formula explorer that publishes the question, exact working calculation, data inputs, and guardrail behind every percentage—including the complete road-and-neutral method.
  • Compressed the Rankings introduction and formula hierarchy so the live status, proposed model, current published poll, and historical laboratory are easier to distinguish.
  • Moved the supplied Texas Tech Double T into the shared team identity directory so Rankings, Bracket Lab, and every other team display use the same correct mark.

Still open

  • The sensitivity review is a model guardrail, not an attempt to reproduce committee judgment exactly; weekly checkpoints and out-of-sample validation remain necessary.
  • Several components still share the underlying Team Strength Rating. Version 0.3 should test orthogonalized inputs and a fuller lower-division schedule model before any adoption claim.
  • Defined context remains neutral for every team until a specific, objective rule and review process are published.
Version3.8

NRR-first rankings refinement

What changed

  • Rebuilt Rankings around three clear layers: the current published Top 25, Alec Levine’s proposed National Résumé Rating formula, and the historical testing lab.
  • Made the proposal’s 30 / 25 / 15 / 10 / 10 / 5 / 5 default the visual center of the page and moved temporary reader weight changes behind an optional Adjust the model control.
  • Condensed the current poll display to a ten-team overview with the rest of the Top 25 available on demand, while retaining the automatic handoff to the official CFP order when current-season committee rankings begin.
  • Removed the oversized poll-policy grid and compressed method, limitations, and sources into reader-controlled disclosures.
  • Bundled all 138 team marks locally and made both Rankings and Bracket Lab prefer the local directory so external image failures no longer leave empty team tiles.

Still open

  • The live 2026 NRR remains intentionally empty until completed games create résumé evidence.
  • Institutional marks are used for editorial identification and remain the property of their respective owners; a commercial launch should confirm production licensing requirements.
  • Weekly historical checkpoints and an expanded predictive Team Strength model remain future validation work.
Version3.7

Rankings and National Résumé Rating release

What changed

  • Renamed the primary bracket destination Bracket Lab and promoted Rankings—with a visible NRR marker—to the main navigation.
  • Added a complete Rankings page architecture that distinguishes the AP Top 25, official CFP ranking, National Résumé Rating, and Team Strength Rating instead of treating them as interchangeable orders.
  • Published the verified August 17, 2026 AP Top 25 on the Rankings page and made AP the only pre-CFP display source; the Coaches Poll is excluded throughout the live bracket, calendar, and ranking feed.
  • Made the proposal’s 30 / 25 / 15 / 10 / 10 / 5 / 5 NRR weights visible on the home page and preserved the interactive seven-factor model and 2021–2025 selection-week backtests.
  • Separated the two proposed ratings from their supporting measures: strength of schedule, strength of record, venue-adjusted quality results, road performance, capped game control, best wins, losses, and conference results.

Still open

  • The live 2026 NRR remains intentionally empty until completed games create résumé evidence; AP is a display source and contributes zero percent to the formula.
  • Version 0.1 has five final-selection checkpoints; weekly AP and CFP backtests, bias testing, sensitivity analysis, and a fuller predictive Team Strength model remain required.
Version3.4

Complete bracket identity and venue release

What changed

  • Installed a built-in 138-program FBS identity directory so every available custom-bracket team has its identifying mark, abbreviation, primary color, and readable selected state even when the live directory is unavailable.
  • Removed the separate advance-arrow control from editable seeds: choosing a team or clicking its name now selects that team directly, matching the interaction used throughout later rounds.
  • Rebuilt the simulation strip as two compact Field and Simulation tool groups with smaller actions and a clearer reading order.
  • Added city-and-state labels to every campus-hosted game and official bowl identity art to the six proposed Round-of-12 destinations.
  • Created an original football-and-columns trophy concept for the separate National Tournament rather than imitating or recoloring the CFP championship trophy.

Still open

  • Institutional and bowl marks are used only for editorial identification and remain the property of their respective owners; a commercial launch should replace public identification feeds with licensed production assets where required.
  • Campus locations identify each program’s primary 2026 home city; temporary, alternate, or neutral-site games can require a season-specific venue override.
Version3.3

Saved and shareable custom brackets

What changed

  • Added a compact progress meter to custom CFP and National Tournament brackets without adding another panel to the default viewing experience.
  • Added on-device saving and a Resume Saved action so a reader can return to the most recent custom field and its completed picks in the same browser.
  • Added shareable bracket links that preserve the selected tournament, complete field, and winner selections directly in the URL.
  • Validated shared and saved fields before loading them, including team count, duplicate-team prevention, and the National Tournament’s Power Four exclusion.

Still open

  • Device saves remain in that browser only; cross-device accounts and cloud bracket libraries would require authenticated server storage.
  • Share links intentionally contain the bracket data rather than a tracking identifier, so they can be long but do not require a public bracket database.
Version3.2

Inline bracket builder and National Tournament lab

What changed

  • Replaced the separate seed editor with a blank bracket that is completed in place: team dropdowns now appear inside each seed’s first matchup, duplicate entrants are blocked, and downstream games accept only teams that actually advanced.
  • Added a prominent Create Your Own Bracket action and condensed the favorites-versus-upsets setting into one simulation-style menu.
  • Added a separate 16-team National Tournament simulator with current and 2025 counterfactual examples, two rounds of reseeding, campus hosting through the semifinals, and a distinct title presentation.
  • Enforced the proposal’s National Tournament eligibility boundary in the custom selector by excluding every Power Four member while leaving independents eligible by affiliation.
  • Added a cached live FBS team directory so selected programs throughout both custom brackets can display their current identifying mark, abbreviation, and color.

Still open

  • The National Tournament examples are mechanics demonstrations, not official historical selections or predictions; an adopted field would require final résumé data, champion access, win thresholds, and CFP exclusions.
  • Institutional marks are loaded from a public identification feed and remain the property of their respective owners; a licensed production data provider would be preferable for a fully commercial launch.
Version3.1

Bowl weekend and build-your-own bracket

What changed

  • Rewrote the creator page around Alec Levine’s Columbus roots, Worthington upbringing, and Ohio State education; removed the field-notes framing, reduced the photo essay, and compressed the stadium list into one quiet line.
  • Made the six-game late-December bowl round a major home-page feature and explained why a destination stage fits the winter-break point in the calendar.
  • Added a visible My Bracket mode with seed-by-seed selectors for the complete 2026 FBS directory, current Top 25 priority ordering, automatic duplicate-team swapping, and instant bracket recalculation.
  • Expanded the live ranking payload from the playoff Top 18 to the full Top 25 for selector ordering while preserving the CFP → AP → Coaches source priority and 30-minute checks.
  • Compressed the global footer and added consolidated AI-assistance, independence, marks, simulation, and affiliation disclosures.

Still open

  • The six-bowl set is an illustrative proposed working rotation, not an announcement of future CFP contracts or guaranteed sites.
  • The custom selector includes active and transitioning 2026 FBS programs; future membership changes will require a directory update.
Version3.0

Neutral home field and community forum

What changed

  • Replaced the LSU-led first screen with a four-school documentary collage so the national landing experience no longer appears to promote one program.
  • Rebuilt the Community home around real popular and newest thread feeds, published reply counts, proposal rooms, transparent sorting, and an explicit no-fake-engagement rule.
  • Embedded the relevant public thread preview and thread-creation area inside every detailed proposal chapter while retaining the complete community room for longer discussion.
  • Added required private email and phone fields to fan profiles, explicit contact-data consent, server-side profile completion checks, and public/private identity explanations throughout the posting flow.
  • Renamed the primary About destination to About the Creator and evolved the header identity into a broader College Football Reform Project wordmark.

Still open

  • Public posts and replies remain pre-moderated and limited to one contribution per account every 24 hours; this is a deliberate community-safety constraint rather than an instant-message system.
  • Phone numbers are private profile fields, not SMS-verified identities, and no marketing-text program is enabled.
Version2.9

Locked bracket path and global cleanup

What changed

  • Reordered the quarterfinal column by locked semifinal path so adjacent games feed the same semifinal without misleading connector crossings; the underlying second-reseed matchups did not change.
  • Replaced the boxed FBS badge with a restrained editorial wordmark and carried it through the header, bracket, and footer.
  • Removed the repeated five-photo rail from every page and limited the global footer to contact, press, standards, version-history, and image-credit links; primary destinations now live in the main navigation.
  • Blurred both visible Spectrum naming-rights labels in the real UCF panorama, replaced them with neutral UCF identification, and documented the modification in the image ledger.

Still open

  • The panoramic proposal remains a single user-controlled stadium node rather than a multi-node physical route.
  • The project identity can be renamed later if a custom public domain becomes the permanent brand.
Version2.8

Look-around stadium proposal

What changed

  • Removed the long scroll sequence and its repeated zooms of the same field photograph.
  • Installed a real, rights-checked panoramic college-football stadium that readers can drag, swipe, zoom, and explore directly.
  • Anchored all ten proposal positions inside the stadium view, added camera movement to each selected position, and paired the environment with an always-available position dock and conventional full index.
  • Added a focused stadium entry point to the home page, local viewer files for faster and more private loading, a still-image fallback, keyboard-accessible markers, and complete source attribution.

Still open

  • This release provides a genuine look-around panoramic environment, but it is one stadium node rather than a multi-location Street View system.
  • A future multi-node tour would require several compatible, rights-cleared 360-degree captures whose physical viewpoints form a coherent route.
Version2.7

Photographic stadium walkthrough

What changed

  • Removed the fabricated field plane, generated yard lines, floating waypoints, and center-field seal from the proposal experience.
  • Rebuilt the proposal as an eleven-stop photographic walk from the tunnel entrance, down the ramp, through the stadium opening, and across a real field.
  • Added step-by-step crossfades, scroll-linked camera movement, previous and next controls, a clickable stadium route map, visible location labels, and direct links into every proposal chapter.
  • Kept all primary imagery real and rights-checked while preserving the conventional full proposal index and reduced-motion route.

Still open

  • The current walkthrough uses cinematic transitions between real still photographs; a true user-controlled 360-degree tour would require a properly licensed panoramic capture or an external mapping provider and API access.
  • The route map is a navigation aid, not an architectural map of the photographed stadiums.
Version2.6

First-person field walk

What changed

  • Replaced the separate proposal banner and diagram-like field with one continuous first-person sequence that begins inside a real stadium tunnel, walks toward daylight, and crosses onto the field.
  • Added scroll-linked head-height camera movement, restrained pointer parallax, a lens frame, and augmented-reality-style proposal waypoints that appear in the reader’s path.
  • Preserved a direct read-without-motion route, the compact proposal index, a reduced-motion presentation, focused mobile controls, and full photo-credit access.
  • Installed the supplied Texas Tech Double T artwork in the interactive bracket and protected it from being overwritten by the live ranking feed.
  • Added a live second-reseed explanation showing why the highest remaining seed receives the lowest remaining seed and when the bracket locks.

Still open

  • The first-person sequence is an editorial simulation built with real stadium photography and browser motion—not a 360-degree recording or institutional virtual-reality product.
  • Further device testing can refine the exact camera pacing without changing the proposal’s conventional reading path.
Version2.5

The downfield journey

What changed

  • Replaced the static proposal field with a scroll-driven stadium journey: the camera remains at field level while yard lines and proposal positions advance toward the reader.
  • Added a live game-plan display that changes across all ten proposal sections, with direct links into each complete policy chapter.
  • Added an interactive ten-stop progress rail, field-position markers, a moving center route, downfield progress, and a restrained scroll cue.
  • Kept the compact proposal play sheet below the experience and added a reduced-motion mode so the concept remains accessible rather than becoming a navigation barrier.
  • Reworked the interaction for smaller screens with focused cards, simplified depth, and touch-sized section controls.

Still open

  • This is a deliberately ambitious editorial interaction; future passes can refine pacing and depth after testing across additional real-world devices.
  • The real stadium image remains documentary background material and does not imply institutional endorsement.
Version2.4

Enter the stadium

What changed

  • Rebuilt the proposal index as a field-level stadium experience that moves from a tunnel entrance into a selectable ten-part game plan on the grass.
  • Preserved a compact conventional proposal index beneath the field so the visual concept never makes the policy harder to navigate.
  • Replaced the football-and-18 icon with a restrained FBS Reform typographic mark shared by the header, footer, and bracket.
  • Rebuilt the projected champion presentation as a light championship stage with a substantially larger official CFP trophy image and clearer winning-team identity.
  • Added a direct traditions handoff and documented the new real, rights-checked visual sources.

Still open

  • The stadium field is an editorial navigation concept, not an affiliation with or endorsement by the pictured institution.
  • The projected champion remains the output of the selected simulation settings and is never presented as a prediction or official result.
Version2.3

Complete operating calendar

What changed

  • Rebuilt the calendar around a clear handoff: the current 2026–27 system finishes unchanged, and the proposed operating calendar activates January 26, 2027, immediately after the current CFP championship.
  • Made the proposed system the default view while preserving a verified current-system reference and a continuous 25-month comparison bridge.
  • Replaced the broad design-and-backtesting runway with dated football, ranking, media, academic, recovery, coaching, recruiting, transfer, roster-certification, selection, and postseason windows.
  • Added month navigation, visible-month context, phase labels, event counts, a monthly agenda, richer event explanations, operating rules, locations, source links, and a grouped detailed timeline.
  • Extended the model through the next 2028 kickoff so the reform reads as a repeatable annual rhythm rather than a one-season diagram.

Still open

  • All reform dates remain an illustrative policy model until leagues, institutions, athletes, broadcasters, bowls, and governing bodies adopt a final calendar.
  • Local academic calendars would control the precise campus exam protections inside the proposed national quiet bands.
Version2.2

Stadium identity and complete bracket sites

What changed

  • Removed the oversized bracket-lab banner so the simulator opens directly on its live field, source status, controls, and playoff board; compressed other informational page introductions into shorter, photo-backed headers.
  • Replaced the generic 18 tile with an original football-and-laces project mark and added a rotating strip of credited, real college-football places and traditions across the public site.
  • Corrected selected-team contrast throughout the bracket, including Notre Dame, and replaced the generic cup illustration with a credited photograph of the College Football Playoff National Championship Trophy.
  • Added a venue line to every matchup: campus hosts in the campus rounds, Rose, Sugar, Orange, Cotton, Peach, and Fiesta bowls in the six-game Round of 12, and a preselected neutral site for the national championship.
  • Carried the six-bowl working set into the proposal detail and calendar so the rules, simulator, and season model use the same language.

Still open

  • The six bowl identities and geographic assignments are an illustrative working rotation for this proposal, not an announcement of official CFP contracts or future sites.
  • Exact dates, host stadiums, and bowl pairings would be finalized through media, venue, athlete-welfare, and governance review.
Version2.1

Editorial reset

What changed

  • Rebuilt the home page around one clear reading path and removed repeated proposal summaries, duplicate download sections, the author biography block, and attributed slogan quotations.
  • Reduced the home-page tradition gallery from fifteen images to six stronger, higher-resolution photographs with a more deliberate editorial layout.
  • Rebuilt the About page around the graduation portrait and five selected personal photographs, removing the oversized photo wall, stadium-card repetition, and attributed quote.
  • Simplified primary navigation, renamed the essay area Analysis, removed unnecessary author references from proposal, simulator, calendar, community, moderation, analytics, and contact interfaces, and refreshed shared cards, heroes, spacing, buttons, and responsive layouts.

Still open

  • The Media page intentionally retains formal author attribution because that page exists for press use.
  • Additional personal photographs should be added only when they contribute a genuinely new chapter rather than repeat an existing game-day scene.
Version2.0

Author story release

What changed

  • Rebuilt the About page as a complete first-person author profile rooted in Alec Levine’s Worthington upbringing and four years at Ohio State.
  • Published ten personal photographs selected and supplied by Alec, including Ohio Stadium, Beaver Stadium, the Rose Bowl, the 2024–25 playoff run, the senior-year national championship, and his 2025 graduation.
  • Added a personal timeline, five-stadium passport, photographic essay, and three editorial beliefs connecting Alec’s lived experience to the policy proposal.
  • Expanded the authorship, independence, image-permission, and AI-assistance disclosures.

Still open

  • Captions can be refined with exact dates, opponent context, and the names of pictured friends if Alec chooses to publish them.
  • High-school and early Worthington photographs could add a future chapter about Alec’s life before Ohio State.
Version1.9

Verified bracket center release

What changed

  • Replaced illustrative poll orders with the verified August 4, 2026 AFCA Coaches Poll and published the source, release date, and last-checked date beside the bracket.
  • Added an explicit ranking-source ladder that prioritizes CFP, then AP, then Coaches, while showing which source is active and which have not yet been released.
  • Added the official December 7, 2025 final CFP Top 18 as a historical archive that can be loaded into the proposed format.
  • Added an optional full-tournament 2025 illustration with every generated outcome visibly labeled hypothetical rather than historical.
  • Replaced the month strip with a physically scrollable, 25-month calendar grid, direct kickoff and postseason jumps, detailed event selection, and a timeline/calendar toggle.
  • Corrected the Christmas recovery window so it ends before the January 1–2 CFP quarterfinals.

Still open

  • The ranking snapshots are manually verified; automatic updating requires a dependable licensed or official data feed.
  • The 18-team projection uses ranking order as seed order and does not yet adjudicate all automatic-access, champion-protection, and eligibility rules.
  • Exact 2027 dates remain a planning model requiring governance, academic-calendar, media, and athlete-welfare review.
Version1.8

Current snapshot and calendar release

What changed

  • Added a poll snapshot selector to the bracket lab with illustrative Coaches Poll, AP-style, and résumé-order starting fields.
  • Added a full proposed 2027 season calendar from the August 2026 design runway through the opening of the 2028 season.
  • Added calendar filters for football season, postseason, roster season, and design/testing work, with visible labels separating planning assumptions from official schedules.
  • Added the calendar to primary navigation and the public tools area so readers can move between the current bracket view and the full-year operating model.

Still open

  • The poll selector is prepared for a live source connection but currently uses clearly labeled illustrative snapshots.
  • Exact 2027 dates require future media, academic-calendar, recovery, and governance decisions.
Version1.7

Guided proposal release

What changed

  • Turned the ten-part proposal index into a set of clickable, full-length reading pages.
  • Added policy summaries, key rules, operating explanations, safeguards, unresolved testing questions, written-source references, and topic-specific discussion links for every part.
  • Added persistent section navigation and previous/next controls so readers can move through the complete framework without returning to the index.

Still open

  • Reader questions and evidence should continue to refine each section.
  • Historical simulations and independent legal, financial, and athlete-welfare review remain required before adoption.
Version1.6

Saturday Atlas release

What changed

  • Expanded the home page into a documentary Saturday Atlas spanning 17 distinct college-football places and traditions without repeating a university in the gallery.
  • Added real, rights-checked photography from Ole Miss, Texas, Wisconsin, USC, LSU, Notre Dame, Clemson, Washington, Colorado, Tennessee, Penn State, and Appalachian State.
  • Reserved Ohio State imagery for Alec’s future personal story instead of using it as generic home-page decoration.
  • Added native lazy loading to the large photo gallery and expanded the public source-and-license ledger with links to original files.
  • Checked a large set of .com and .org brand options against authoritative registry data and recorded the strongest choices in the private launch center.

Still open

  • Alec’s personal photo gallery awaits selected images and captions.
  • A custom domain remains available only until another buyer registers it; Alec must choose and purchase the final name through a registrar.
Version1.5

Real traditions release

What changed

  • Removed all AI-generated imagery from the public site and social preview.
  • Rebuilt the home page around real, rights-checked photographs of college football, campuses, crowds, bands, tailgating, mascots, and rivalry ceremonies.
  • Added an editorial traditions gallery and warmer language centered on the feeling of returning to campus.
  • Expanded the image ledger with detailed creator, institution, and reuse information.

Still open

  • Alec’s personal photo gallery awaits selected images and captions.
  • Additional traditions can be added when strong, reusable photography is verified.
Version1.4

College football identity release

What changed

  • Rebuilt the public visual system around stadium, marching-band, mascot-culture, and game-day imagery. AI-generated editorial scenes used in that release were removed in version 1.5.
  • Added persistent fan profiles with chosen public name, favorite team, and community perspective.
  • Changed contribution limits to one opening perspective or response per signed-in account every 24 hours.
  • Clarified that phone numbers are not collected without a real phone-based service.
  • Prepared the author page to incorporate Alec’s high-school, Columbus, campus, and game-day photographs.

Still open

  • Alec’s personal photo gallery awaits selected images and captions.
  • A custom domain still requires a final name choice and registration.
Version1.3

Community and accountability release

What changed

  • Created nine moderated discussion rooms with opening perspectives and threaded reader responses.
  • Added private contribution histories for signed-in readers and a private owner launch center.
  • Added explicit, consent-based first-party audience measurement and private aggregate analytics.
  • Added voluntary community priority signals and public, non-scientific aggregate counts.
  • Published a finite analytics-retention policy and clearer launch limitations.

Still open

  • Audience history begins with this release; earlier traffic cannot be reconstructed.
  • Update requests are stored, but email campaigns require an external delivery service.
  • Personal photo essay and custom domain still require Alec’s choices.
Version1.2

Evidence and participation release

What changed

  • Added evidence and validation dashboard with official source library.
  • Added interactive 18-team bracket simulator and print-ready executive summary.
  • Added Alec’s View essay section, contact workflow, and private inbox.
  • Added version history, correction channel, and expanded public navigation.
  • Fixed duplicated page titles and removed unstable client-side link prefetching.

Still open

  • Personal photo essay awaits images selected by Alec.
  • Custom domain awaits a domain choice and ownership access.
Version1.1

Public platform release

What changed

  • Expanded the original landing page into proposal, author, reader-perspective, and standards pages.
  • Added moderated reader submissions with sign-in, privacy notice, rate limits, and owner approval.
  • Published the site publicly while keeping moderation private.

Still open

  • First reader contributions not yet approved or published.
Version1.0

Initial proposal package

What changed

  • Published the integrated policy proposal, validation brief, media framework, NIL framework, communication kit, and editable workbook.
  • Established Alec Levine authorship and full AI-assistance disclosure.
  • Defined the non-Power Four-only National Tournament and no-transfer rule for eliminated CFP teams.

Still open

  • Historical data and confidential commercial inputs remain incomplete.

Corrections

Material errors should be visible—not quietly erased.

Submit a correction with the specific passage, the claimed error, and a supporting source. Verified material corrections will be dated here.

Request a correction