Editorial standard · Model 2026-07-28

How 20-0 Works

Historical facts, 20-0 ratings, projected records, and simulated results are different things. This page explains each one, publishes the current game constants, and records material corrections.

1. Historical facts

A player's displayed season identifies the season used as editorial context for the game. The legacy dataset's detailed statistics and accolade snippets are currently hidden because it has no complete source/version manifest and an audit confirmed numeric errors. We do not present those fields as verified facts while the dataset is being rebuilt.

New or restored historical fields must identify a specific source, version or retrieval date, covered fields, license or terms, local transformations, and reviewer. Established league, team, Hall of Fame, and statistical references may be used for manual cross-checking. A link is a fact-checking reference, not a sponsorship or blanket reuse license.

Unverified fields are omitted rather than estimated. If a displayed fact differs from a published record, readers can use the correction process below. We review material corrections and record confirmed changes.

2. Player ratings are editorial

The 0-100 score beside a player is a 20-0 game value, not an NFL, team, Hall of Fame, or third-party rating. Current scores are legacy editorial inputs used by the game model; they are not independently verified historical grades. The intended framework considers position-relative dominance, era context, production, awards, and postseason impact, then compares players within a position so the number supports a draft decision.

Cross-era comparison is inherently uncertain. Rules, schedules, stat definitions, and positional responsibilities changed over time. We therefore treat a one-point difference as a game-balancing choice, not proof that one player had a better career. Scores remain subject to review as the player source manifest is rebuilt.

3. Roster and season calculation

The lineup has six offensive and six defensive slots. Each side is the arithmetic mean of its six player contributions. An empty or skipped slot contributes 50. If malformed shared data places a player on the wrong side, only 70% of that player's rating counts.

A regular-season opponent is sampled between 74 and 90. The three modeled postseason games add 6, 12, and 18 opponent-rating points. Week 18 combines the Wild Card and Divisional labels into one game; the model is not a complete NFL bracket replay. Win probability compares the opponent to the average of offense and defense, with no separate side-specific matchup or balance modifier:

1 / (1 + exp(-(our rating - opponent rating) / 6))

The draft preview averages 200 deterministic scenario runs to estimate wins, then approximates perfect-season odds by raising the estimated average win probability to the twentieth power. It does not count perfect runs directly, so treat that value as a planning heuristic. The final 20-game record is a separate seeded simulation: the same saved lineup and seed reproduce the same schedule and result. Generated opponent names and box-score flavor are fictional and do not represent actual NFL games.

Worked example: one roster from slots to score

Consider a completed roster with offensive contributions of 96, 92, 88, 84, 82, and 50. Those six slots total 492, so offense is 82.0. The defensive contributions are 94, 90, 86, 84, 80, and 50: 484 in total, or 80.7 after rounding. The combined rating used against an opponent is therefore (82.0 + 80.7) / 2 = 81.35.

Against a regular opponent rated 82, the logistic formula gives this roster roughly a 47% win chance for that game. If the final seeded schedule produces 17 wins and 3 losses, the grade is B+ under the published bands. The leaderboard score is then approximately 17 × 50 + 82.0 × 1.5 + 80.7 × 1.5 = 1,094 points. This is a worked example, not a claim about a real NFL team or a universal expected result.

The empty slot is the important comparison. Replacing the defensive 50 with an 80 changes defense to 85.7 and the combined rating to 83.85. Against the same 82-point regular opponent, the chance rises to about 58%. A single legal fit can therefore matter more than a small upgrade to a side that was already strong.

Sensitivity and model limits

• A one-point rating change is a game-input adjustment, not proof of a real-world one-point difference in player ability.
• A five-point change can move win probability noticeably near an even matchup, but it still does not remove schedule variance or the playoff boosts.
• The same contribution gain to either side has the same effect after averaging. Compare each player with the value of the legal slot it would replace; the model adds no separate bonus or penalty for balance.
• The 200 quick-season preview and the final seeded 20-game run answer different questions. A projection is a planning signal; the saved record is one reproducible outcome.
• Opponent names, weekly box scores, and the 20-game schedule are fictional game output. They are not historical NFL results or forecasts of a future season.

The current implementation is identified as model 2026-07-28. When a constant, rating input, or correction changes, the affected page and correction log should say what changed and whether old share links remain readable. Replaying an old link can preserve its saved lineup while still displaying the current explanatory policy; it does not turn a fictional result into a historical record.

4. Grades and leaderboard score

The leaderboard score is rounded to the nearest whole number:

wins × 50 + offense × 1.5 + defense × 1.5

For example, an 18-2 roster with offense 91 and defense 84 scores 1,163, while an 18-2 roster with offense 87 and defense 90 scores 1,166. The second ranks higher because the additive score uses the sum of both side ratings. Exact score ties are ordered by earlier submission time. The grade, in contrast, depends only on wins. Classic and Gridiron IQ use the same season model but maintain separate leaderboards.

5. Gridiron IQ Scout Report

The Scout Report does not compare raw player scores. For every round with at least two eligible choices, it measures where the selected player ranked among the available candidates. Best available earns 100% for that decision, worst earns 0%, and the final Scout Rating is the average. Forced choices and skipped rounds are excluded.

• PERFECT EYE: 100+
• ELITE EYE: 90+
• SHARP SCOUT: 78+
• GUT CALLER: 60+
• COIN FLIP: 0+

6. Sources, names, and brand boundaries

20-0 uses player, team, and league names as plain text to identify the historical subjects being discussed. We do not use official team or league logos, player photographs, or claim affiliation, sponsorship, or endorsement. Generated avatars are fictional game artwork, are not intended to depict real player likenesses, and use no official visual assets. Simulated opponents are fictional as well.

Historical facts can be independently stated, while a publisher's original prose, images, selection, arrangement, and database access terms remain separate rights. Our operating rule is to verify and restate facts, link readers to references, and avoid bulk copying or scraping restricted sources.

• Feist Publications, Inc. v. Rural Telephone Service Co. (Cornell Legal Information Institute) — U.S. Supreme Court decision distinguishing facts from protectable selection and arrangement.
• NFL.com Terms and Conditions (National Football League) — NFL.com's content, database presentation, logos, and trademarks remain subject to its terms and rights.
• Trademark scope of protection (United States Patent and Trademark Office) — Background on trademarks as source identifiers for particular goods and services.
• Terms of Use (Sports Reference) — Explains permitted page-level data use and restrictions on scraping or recreating a competing database.
• C.B.C. Distribution and Marketing v. MLB Advanced Media (U.S. Court of Appeals via GovInfo) — Fantasy-sports decision involving public player names and statistics; relevant context, not a universal license.

7. Corrections

Send a suspected error to [email protected] with the player or article, season, field in question, your proposed correction, and a public reference. We check material corrections against more than one source where possible, update the affected page, and record the change here.

What to read next

See the product-specific analysis in our guides, or contact us with a correction.