How My Pro Football Hall of Fame Rating (HoFW) Works
How career records become HoFW scores and induction forecasts, and where those estimates are uncertain.
What HoFW measures
HoFW (Hall of Fame Worthiness) answers a single question: how strong is this player's Hall of Fame case, on merit, compared to the average real inductee at their position? A score of 100 means "as strong a case as the typical enshrined player at that position." There is no ceiling and no floor. A handful of inner-circle greats clear 200.
Whether the committee will vote them in is a separate question, and I keep the two apart. HoFW and Tier measure merit. Projected Year and Induction Probability are the real-world prediction. A player can have a strong case and still be projected several years out, because there's a line of equally deserving players ahead of them.
How Coach Score works
Coach Score compares regular-season head-coach wins minus losses. A score of 100 equals the average for the Hall's Coach-category members, using each inductee's record strictly before their election year. Coaches who returned after induction, including Paul Brown and Joe Gibbs, contribute only their earlier records to that benchmark.
Coach Score = max(0, 100 × (wins − losses) ÷ Hall coach mean W−L). The benchmark is available with the live coach data. There is no ceiling. Ties do not change W−L; nonpositive records receive zero. Missing head-coach records remain unscored.
The score covers regular-season head-coach records from the NFL, AFL, AAFC and APFA. A coach's other professional jobs are checked separately and don't enter it.
The score bands are 0 to under 50, 50 to under 100, 100 to under 150, and 150 or more. The bands only describe where a coach sits on this scale. The score gives no induction chance, projected class year or automatic ballot spot. Player HoFW uses its own formula and tiers.
I tested a fixed three-feature challenger using schedule-adjusted W−L, performance above the league's season average, and championships. In the frozen 2026 study, its leave-one-out AUC was 0.974 versus 0.963 for raw W−L, and average precision was 0.723 versus 0.664. The cohort contained 29 Coach-category inductees and 393 unelected coaches with reviewed histories. Paired 95 percent intervals for the improvements included zero, so I kept the W−L baseline. These retrospective comparisons do not establish accuracy for the new selection process.
A longer career and a longer schedule offer more chances to accumulate wins above .500. This baseline does not adjust for era, credit assistant work or innovations, or count postseason results and other professional leagues' records. Assistant-only candidates such as Alex Gibbs have no head-coach record to score. Contributors get no score.
The history audit includes professional assistant, coordinator and consultant jobs, plus CFL, USFL, XFL, UFL and earlier professional leagues. Sourced role dates provide career context; they are not official Hall eligibility rulings or proof that every job is documented. Sparse histories remain flagged. The Studio and forecast use a hand-maintained candidate roster that will be replaced when the Hall announces the new cycle's nominees.
Coach Score doesn't change how the forecast holds spots for coaches and contributors, or its labeled 2027 Belichick assumption. Comparing résumés tells us nothing new about how selectors vote.
Sources: manually captured Pro Football Reference head-coach seasons, the Hall's Coach-category list and media guide, and revision-specific Wikipedia biographies for professional histories. The study and source register document the comparison, joins and unresolved coverage.
How a score is built
Every player goes through the same five-step pipeline. Each step is position- and era-aware.
What drives the score
Coefficients are fit on standardized inputs, so bar width directly compares how much each feature matters to the model's induction predictions. Bigger bar = stronger signal.
Some offensive linemen use an honors-based proxy. Linemen covered by the main model use its fitted score. The separate proxy uses these inputs:
- All-Decade Teamheaviest single input in the OL formula
- First-team All-Proeach selection counts heavily
- Pro Bowl selectionsmoderate weight per selection
- Career value (wAV)lightest weight, often missing for players drafted before 1980
Reading the tiers
Tier cutoffs come from the scores of Hall members in the training data. They describe how a career compares with that group; they do not measure accuracy on future elections.
HoFW vs. Induction Forecast
A career score and an election forecast answer different questions.
A position-calibrated merit score. High HoFW means a real Hall of Fame–caliber career. It doesn't tell you when the committee will vote them in.
Projected Year is a deterministic queue snapshot, answering when the backlog reaches a player if nothing surprising happens. The queue ranks fitted career forecasts before HoFW’s position-specific merit rescaling, then accounts for eligibility, available seats and time spent waiting. Probability of Eventual Induction is the calibrated logistic number. The two can disagree.
Since the Class of 2027 there is one pool: every retired player competes together, and the Hall expects to elect five or six people a year in total. Part of that class is spoken for before any modeled player competes, so the queue budgets it in three parts.
| Share of the class | Who | Slots / year |
|---|---|---|
| Modeled players | Everyone scored on this site | — |
| Long-retired backlog | Long-retired candidates outside the modeled queue | — |
| Coach & contributor, combined | Reserved capacity for coach and contributor finalists | — |
The induction model ranks players. Coaches have a separate résumé score described above, without an induction probability. Contributors remain unscored.
The queue budgets coach and contributor places separately from the long-retired backlog. The coach and contributor still get protected spots, one finalist each on the 15-name slate, but once there they face the same room and the same 80% bar as the thirteen players. So each is budgeted like any other slate place: an expected class of five or six spread across fifteen finalists. A finalist place is not a guaranteed induction, which is why their row is a fraction of a slot rather than two. Their old election rate isn't a guide to this. Until 2023 they mostly got their own yes-or-no vote and nearly always passed, and in the three years they had to compete for votes, 2024 through 2026, none of them got in.
Long-retired candidates are the other case. They lost their protected route entirely: their historical rate came almost entirely from the senior committee, which no longer exists. Across the last ten classes (excluding the one-off Centennial Class of 2020), sixteen long-retired players were elected and fifteen of them arrived through that committee. Exactly one won a place on the regular ballot. So their reservation is scaled from that merit path alone, which is why it is a fraction of a slot rather than more than one.
The long-retired estimate rests on a single player, so it is especially uncertain. The Class of 2027 will provide the first result under the new process. What Changed in the Hall of Fame Vote explains why the merge made this necessary.
Model accuracy
Validated two ways: cross-validated AUC-ROC on held-out induction outcomes, and a per-class rank check against every real induction class since 2005, testing whether real inductees ranked near the top when their player rows were held out from fitting.
The OL honors-proxy formula is validated separately. With no box-score stats to work from, it is its own model, calibrated against nflverse OL data from 1980 onward.
Careers that started before 1970
The model learns from every player who played a season since 1970. Until October 2026 it learned only from careers I could see in full from 1999 on, plus rough estimates for older Hall of Famers built from a single career total. Going back to 1970 gave it more than twice as many real Hall of Famers to learn from, each measured season by season, and in back-tests against past classes it ranked the eventual inductees higher than the old version did.
Some of the tracked real Hall of Famers started their careers before 1970, so part of their record is missing. They still count in training, flagged as incomplete, so the model can tell a partly seen career from a short one. The same goes for players drafted before 1980, where my draft records begin. In the feature chart above, neither flag is shown, because neither applies to anyone who could be on a ballot today. Already-enshrined players whose careers started before 1970 show as Enshrined with no number, because a score built on part of a career would understate it.
Where the data comes from
The model's career measures come from Pro-Football-Reference.com and Stathead: Approximate Value for every season since 1970, every Pro Bowl and All-Pro selection back to 1970, team playoff records, and the games-started figures the offensive-line formula uses. Season stats from 1999 on, the ones behind each player's signature stat, and the draft history come from nflverse, the open NFL data project, released under a Creative Commons licence.
I collect the Pro Football Reference and Stathead files by hand, using their export tools where available. The site publishes derived scores and forecasts, with source credit here and on the player database.
There is no free, openly licensed alternative for the pre-1999 seasons. I looked. The commercial sports-data feeds start in 2000 or 2010, and the open datasets that do cover the 1980s trace back to Pro-Football-Reference anyway. They did the work of compiling that era, so I'd rather credit them directly.
This site is free and has no ads, which is part of why I'm comfortable building on that data. If that changes, I'll have to revisit this first.
What it cannot see
The model uses career records and ballot history. It cannot observe selectors' arguments, personal judgments or changing priorities. Its historical tests cover the process used through 2026; results under the new rules are not yet available.
Kickers, punters and long snappers use a separate honors score, with no induction probability. The main training data starts in 1970; careers that began earlier are incomplete. Some historical inductees are listed without a score.
A score answers how a player compares to the players already enshrined at their position. That is a different question from whether the committee gets to them this year, which is why the projected class and the induction probability can disagree.
Data coverage and scoring details
- The main model learns from seasons since 1970. Earlier career seasons are missing, and incomplete records are flagged. Some historical inductees appear without scores.
- Offensive-line records provide less individual production detail. Some linemen use a separate honors-based score; those scores and their probabilities are not directly comparable with the fitted model's outputs.
- Older offensive-line records can lack career value. The proxy estimates missing value from games played. Players with no usable value or honors remain unscored.
- Specialists use a separate honors formula based on a small group of inductees. They have no induction probability or projected class year and are listed after other positions in combined rankings.
- Active-player scores describe their careers so far. Retirement dates are estimated from position averages, so projected class years can change substantially.
- Some positions have few inductees and share a scoring baseline with related positions. Their scores are less precise comparisons than those for positions with larger samples.
- Some defensive totals come from play-by-play attribution and may be incomplete. Name-matched award records can confuse namesakes; stable player IDs are used where available.
- The queue assumes a limited number of places each year. New retirements, voters' priorities and the treatment of long-retired candidates can change who gets those places.
- Coach Score covers regular-season head-coach records. It leaves out assistant work, innovations and postseason performance, and gives no induction probability. Contributors get no score.
The full database lives on the HoF Player Database, with every player I track, sortable and filterable by position, tier and eligibility, plus the Deep Dive breakdown behind every score.
Open the player database