Home field is worth 25 rating points, not 65
Re-fitting home field on real results moved it from 55 rating points to 25 — barely a point of spread. What the 2025 data actually shows about playing at home.
Home field advantage is the most quoted number in football analytics and one of the least examined. The convention for years was around three points of spread, which in Elo terms is roughly 65 rating points. This model started at 55. After re-fitting on 2025 results, it settled at 25.
How the number was found
Not by argument. The constant was swept across a range of candidate values, and for each one the whole season was replayed and scored on log loss — a measure that punishes confident wrong predictions much harder than cautious ones. The value that minimised log loss won. Twenty-five points did, by a margin large enough to survive splitting the season into a fitting half and a held-out half.
| Constant | Original value | Fitted value | What it does |
|---|---|---|---|
| K factor | 20 | 28 | How fast a rating moves after a game |
| Home field | 55 | 25 | Rating points added to the home team |
| Rest | — | 5/day, cap 25 | Credit for extra days between games |
| Offseason regression | 33% | 25% | Pull toward 1500 between seasons |
What 25 points means at the betting window
The model converts rating points to spread at 25 points per point of spread. So home field is worth almost exactly one point — not three. On a matchup between two 1500-rated teams, the home side is a 1-point favourite and wins 53.6% of the time. That is a real edge and a small one.
It also reframes what a home upset is. A visiting team rated 100 points higher is still a clear favourite on the road; home field does not come close to closing that gap. Most of the time the crowd is worth less than the roster, and the fitted constant simply says so.
Why the old number was too big
- Travel and scheduling have flattened. Charter logistics, hydration and sleep science have all improved at the same time, and the road disadvantage they used to create has narrowed league-wide.
- Officiating and crowd noise effects are smaller than folklore suggests once you control for the quality of the teams involved.
- Older estimates were often fitted on decades of data, which averages in eras where the edge genuinely was larger.
The practical upshot on this site: every matchup page shows the home field line item explicitly, at 25 points, in the same ledger as the rating gap and the rest edge. If you disagree with the constant, the factor lab on each matchup page lets you drag it and watch the win probability and spread move in real time.