The closing line beats us. Here is by how much.
We grade every forecast against the closing line, the sharpest public prediction of a game. In every sport we cover, our average miss is larger than the market's. Use these ratings to understand teams and games. Do not expect them to beat a sportsbook.
0.42
College basketball: the closing line's average miss was smaller by this much, on 295 games this season (10.13 against 9.71).
0.55
NBA: our average miss was smaller by this much, on 33 games this season (10.04 against 10.59).
0.27
College football: the closing line's average miss was smaller by this much, on 110 games this season (13.42 against 13.15).
0.15
NFL: the closing line's average miss was smaller by this much, on 100 games this season (9.37 against 9.22).
This season, game by game
Each finished game with a forecast made before the start. "Public" means that the forecast was on this site before the game started. "Private log" means that we logged it with a time stamp before the game, before this site existed.
| Sport | Forecasts | Graded games | With a close | Our average miss | Closing line's |
|---|---|---|---|---|---|
| College basketball | Public | 21 | 21 | 13.31 | 12.26 |
| College basketball | Private log | 394 | 274 | 9.88 | 9.51 |
| NBA | Private log | 191 | 33 | 10.04 | 10.59 |
| College football | Private log | 484 | 110 | 13.42 | 13.15 |
| NFL | Private log | 100 | 100 | 9.37 | 9.22 |
In points. A miss is the distance between a forecast and the final margin. When a closing line is recorded, both averages use the same games. Without one, the "ours" value covers all graded games.
Past seasons
| Sport | Rating | Ours | Closing line | Gap | Sample | Test type |
|---|---|---|---|---|---|---|
| College basketball | Power, game margins | 8.96 | 8.85 | 0.11 | 5,597 games with a closing line, 2023-24 | In-sample |
| College basketball | Power, game totals | 13.16 | 12.92 | 0.24 | 5,597 games with a closing line, 2023-24 | In-sample |
| College basketball | Power, game margins | 8.95 | 8.80 | 0.15 | 5,540 games with a closing line, 2024-25 | In-sample |
| College basketball | Power, game totals | 13.02 | 12.63 | 0.39 | 5,540 games with a closing line, 2024-25 | In-sample |
| College basketball | Power, game margins | 9.06 | 8.97 | 0.09 | 5,432 games with a closing line, 2025-26 | In-sample |
| College basketball | Power, game totals | 13.91 | 13.44 | 0.47 | 5,432 games with a closing line, 2025-26 | In-sample |
| NBA | Strength | 10.02 | 9.83 | 0.19 | one season of the 5,281 games scored in 2022-23 to 2025-26, 2022-23 | Walk-forward |
| NBA | Strength | 10.81 | 10.67 | 0.14 | one season of the 5,281 games scored in 2022-23 to 2025-26, 2023-24 | Walk-forward |
| NBA | Strength | 10.87 | 10.64 | 0.23 | one season of the 5,281 games scored in 2022-23 to 2025-26, 2024-25 | Walk-forward |
| NBA | Strength | 11.11 | 10.96 | 0.14 | one season of the 5,281 games scored in 2022-23 to 2025-26, 2025-26 | Walk-forward |
| College football | Rating | 12.43 | 12.00 | 0.43 | 2,883 FBS-vs-FBS games, 2022 to 2025 seasons | Walk-forward |
| NFL | Rating | 9.94 | 9.78 | 0.16 | 1,610 games, 2020 to 2025 regular seasons | Walk-forward |
| NFL | Rating | 10.25 | 10.04 | 0.22 | 56 games, 2026 weeks 7-10 (live, final games so far) | Forward |
Average miss on the final margin, in points, from backtests. In-sample tests use seasons that also helped build the model, so they flatter it. Walk-forward and holdout tests do not.
Did the line move toward our number?
| Sport | Test | Movement toward us | Games | Period | Type |
|---|---|---|---|---|---|
| College basketball | Margins: our margin 3 points or more from the morning spread (3 points) | +0.83 points | 81 | 2026-11-03 to 2026-11-16 | Forward |
| College basketball | Totals: our total 4 points or more from the morning total (4 points) | +0.22 points | 61 | 2026-11-03 to 2026-11-16 | Forward |
| NBA | Margins, gaps of 2 or more points against the line at tip - 3 h (median of 3 or more books) (2 points) | +0.19 points | 2,479 | 2022-23 to 2025-26 | Walk-forward |
| NBA | Margins, gaps of 2 or more points against the line at tip - 3 h (Pinnacle) (2 points) | +0.17 points | 2,490 | 2022-23 to 2025-26 | Walk-forward |
| College football | Spreads, entries with The Odds API lines at entry and close (any nonzero gap) | −0.08 points | 18 | 2026-10-29 to 2026-11-15 | Forward |
| NFL | Early-week line value against Pinnacle (not an edge: the pre-registered test failed) | +0.17 points | 543 | 2024 and 2025 regular seasons | Holdout |
A positive number means the market moved toward our forecast between the time we logged it and the close. It is a sign that the forecast carries information. It is not a profit figure, and it does not include the sportsbook's margin.
How much of the season is graded
A forecast can only be graded against the close if we recorded a closing line for that game. This is the share we caught, so you can see what the averages above leave out.
- College basketball
- 415 finished games: 415 with a forecast before the start, 295 with a closing line (71%).
- NBA
- 192 finished games: 191 with a forecast before the start, 33 with a closing line (17%).
- College football
- 484 finished games: 484 with a forecast before the start, 110 with a closing line (23%).
- NFL
- 100 finished games: 100 with a forecast before the start, 100 with a closing line (100%).
Some days have no forecast, for example during a computer outage from November 4 to November 10, 2026. Games without a forecast made before the start are not graded.
Every game, week by week
Each week we list every finished game: our forecast, the closing line, the result and who was closer.
Take the whole record
Every graded forecast, with its time stamp, the result, our miss and the closing line's miss.