How ChelHead's numbers are produced
This web site is not endorsed by or affiliated with Electronic Arts Inc.
This page explains where ChelHead's EASHL statistics come from, how often they update, which figures ChelHead calculates itself, and how well the Deserve to Win (DTW) win-probability model holds up when it is tested against real games.
Where ChelHead gets its data
Club, player and match statistics come from the public Pro Clubs statistics that EA publishes for the current NHL title. Raw counts such as goals, shots, hits and games played are shown as EA reports them. A small number of derived figures, such as club-level percentages and goalie save percentage, are recalculated from those counts (see below), and ChelHead's own metrics such as DTW are labelled as ChelHead's.
If EA does not answer, a page says the club or player is temporarily unavailable rather than showing guessed numbers. A few gaps come from EA itself:
- Short match history. EA keeps only a club's last 5 matches of each type (public, cup and private). ChelHead can show older games only for clubs whose matches it saved while they were still listed.
- Incomplete private and older matches. These arrive without some fields, such as the result or team art. Missing values are left blank rather than filled in.
- Players with no games. A player who has never played has a profile but no statistics.
How often statistics update
ChelHead refreshes statistics from EA regularly and briefly caches them so pages load quickly. Most changes appear within an hour, often sooner, but a new game can take longer to show during busy periods or when EA's own service is slow. A game EA has not yet published cannot appear on ChelHead.
How ChelHead calculates statistics
Some of EA's summary figures are misleading if shown as given, so ChelHead rebuilds them from the underlying counts:
- Club percentages. EA adds up each member's own percentage instead of computing a rate over the whole club, so a player with two shots counts as much as one with sixty. ChelHead totals the club's raw counts and divides once.
- Goalie save percentage. EA rounds it down to a whole percent (76% for 73 saves on 95 shots, which is really 76.8%). ChelHead recalculates it as saves divided by shots against.
- Shutouts. A shutout is any game with zero goals against.
- Possession. EA reports a season total, so per-game possession divides it by games played.
Understanding Deserve to Win
DTW estimates how often a team that played the way it did would be expected to win, based on everything in the game except the score. A 69% DTW means that out of 100 games with a similar performance, the team would be expected to win about 69. It looks back at a finished game; it is not a pregame prediction.
DTW draws on shot volume, puck possession, passing, faceoffs, takeaways and giveaways, defensive plays and discipline. It does not use goals, assists, saves, the final score or the result, and it does not consider which players or clubs were involved. The Deserve to Win guide explains how to read it, and the FAQ has the short version.
How accurate is DTW?
The model was built from about 50,000 past EASHL matches and then tested on 12,560 later games it had never seen. On those test games, the favored team's predicted win chance was within 0.7 percentage points of how often favored teams actually won, on average. The table groups the test games by the favored team's predicted win chance.
| Predicted range | Games | Average predicted | Actual win rate |
|---|---|---|---|
| 50% to 60% | 2,605 | 54.9% | 53.8% |
| 60% to 70% | 2,644 | 65.1% | 65.3% |
| 70% to 80% | 2,650 | 74.9% | 73.5% |
| 80% to 90% | 2,699 | 85.0% | 85.2% |
| 90% to 100% | 1,962 | 94.2% | 95.0% |
Limits of DTW
- EA does not record where or how a shot was taken, so DTW measures overall performance, not the quality of individual chances.
- Teams play differently when leading or trailing, which can shift the stats DTW reads.
- Accuracy was measured on complete, decisive games. Forfeits and games with missing stats are less reliable, and a game missing a whole team gets no DTW.
- The test games come from one recent stretch of play, and the current NHL title has a smaller sample than the last one. Accuracy is rechecked before each model update.
- The model is updated from time to time. DTW from an earlier version should not be compared directly with current DTW.
Corrections
If a number looks wrong, check the club or player in the game first: the usual causes are the update delays and EA gaps described above. If it still disagrees with EA's own data, tell us at support@chelhead.com or on the Support page, with the page address and what you expected. More about the project is on the About page.