Prop LabGame script, momentum and scoring bias
NHL Arena Scorer Bias on Hits: How the Edge Died
For years, some NHL rinks counted far more hits than others, and the pattern was reliable enough to use. In 2024-25 the league moved event recording to a central crew. Within two seasons, the bias was gone.
By Edgehalla ResearchPublished 6 min read
On this page
How hits used to be counted
Hits, blocked shots, giveaways and takeaways are judgment calls. Someone has to decide that a bump was a hit. For years, that someone was an off-ice scorer in each building, and some crews were far more generous than others.
We measured it simply. For each arena and season, we took the average hits per game by both teams in the home team’s home games, divided by the same average in its road games. The home team’s own style is in both halves, so it cancels out. What is left is mostly the scorer.
The extremes of 2023-24
Hits per game in home games ÷ hits per game in road games (both teams)
Toronto’s rink logged 46 percent more hits than its road games in 2023-24; San Jose’s logged 29 percent fewer.
A big number is not enough on its own; it could be luck. So we split each season in half by date and checked whether the arenas that inflated hits in the first half also did in the second. From 2017-18 to 2023-24, they did: the split-half correlation was 0.61–0.78 every season.
That made it usable. Rostering heavy hitters who played in generous buildings was a real, repeatable fantasy signal. (We never tested it against hits-prop prices.) Before 2024-25, arena adjustment could move a home-heavy hitter’s total by 10–20%.
Then the league fixed the leak
Starting in 2024-25, the NHL recorded real-time events centrally instead of using a separate crew in each arena. We kept measuring.
Does an arena’s hit inflation repeat?
Split-half correlation of each arena’s home hit factor
Arena hit inflation was highly repeatable before centralized scoring and disappeared after it.
In the first centralized season, the bias shrank: the split-half r fell to 0.37, and the shrunk arena factors stayed within about ±6%. In 2025-26 it fell to −0.15. No arena effect was detectable, so every shrunk factor is now 1.00.
The other judgment-call stats followed the same path. In 2025-26, split-half r was 0.005 for blocked shots, −0.04 for giveaways and −0.21 for takeaways.
Arena bias in 2025-26, by stat
Split-half correlation of each arena’s home factor
In 2025-26, no judgment-call stat showed a repeatable arena effect.
One warning for anyone comparing seasons: the change also shifted league-wide counts. Giveaways and takeaways in particular changed in 2024-25. Compare players across that break with percentiles or z-scores, never raw totals.
Faceoffs, by contrast, never had an arena problem. Each building’s home faceoff residual was about the size of pure noise and did not repeat year to year. More in faceoff matchups.
The other hits myth: player-vs-team splits
A second hits story lives on many fantasy sites: “He always hits a ton against this team.” We tested that too, separately from the arena effect.
We flagged every player-opponent pairing from 2022-23 to 2024-25 with at least three games and a big split. On average they ran about 73% above the player’s normal hits (74% for blocks). Then we checked those same pairings in 2025-26.
Do big player-vs-team splits repeat?
Hits and blocks vs the player’s own average, flagged pairings
Pairings flagged at about 73 percent above normal ran only 5.6 percent above normal for hits the next season.
- When flagged (2022-25)
- Next season (2025-26)
Almost all of it vanished. The hits pairings ran only +5.6% the next season, and the blocks pairings +2.4%. A few games against one opponent is a tiny sample, and big splits are mostly luck.
What does carry over is the opponent at team level. Teams that give up lots of hits keep doing it, and adding that to a hits model raised the correlation with actual team hits from 0.33 to 0.40 in 2025-26. Back-to-backs, by contrast, changed hit rates by less than 1%.
What it means for fantasy and props
- For seasons before 2024-25, adjust. Raw hit totals from those years are partly the rink. Our arena-adjusted hits leaderboard divides every game by that arena’s shrunk factor.
- For 2025-26 onward, do not. Every arena factor is 1.00. A player’s hits are his hits.
- Ignore player-vs-team hit splits. Use the player’s rate, his ice time and the opponent’s team-level hits allowed. That is how Hits Tonight projects them.
- Keep watching. One season of zero is not proof the bias is gone for good. We re-check it every season on our arena pages.
More hockey myths, tested: momentum, game script and garbage time.
Frequently asked questions
Do NHL arenas inflate hit totals?
They used to. From 2017-18 to 2023-24, each arena’s hit inflation repeated within a season at split-half r 0.61–0.78. Since the NHL centralized event recording in 2024-25, the bias shrank to about ±6% and then became undetectable in 2025-26 (r −0.15).
Which NHL arena counted the most hits?
In 2023-24, the last season of in-arena scoring, Toronto’s home games logged 46% more hits than its road games (a factor of 1.46). San Jose’s logged 29% fewer (0.71).
What changed when the NHL centralized stat scoring?
From 2024-25, real-time events like hits, blocks, giveaways and takeaways were recorded centrally instead of by a crew in each arena. Arena bias in all four stats was undetectable by 2025-26.
Do NHL players hit more against certain teams?
Not in a way that repeats. Player-opponent pairings flagged at about +73% hits from 2022-23 to 2024-25 were only +5.6% against the same teams in 2025-26. The opponent’s overall hits-allowed rate does carry over.
Should I adjust hits for the arena in fantasy hockey?
Only for seasons before 2024-25, when adjustment could move a home-heavy hitter by 10–20%. From 2025-26 every arena factor is 1.00, so current hits need no adjustment.
More in Game script, momentum and scoring bias
Sources and method
- NHL stats (NHL.com)
- NHL API reference (community documentation)
- Split-half reliability and the Spearman–Brown formula
- Empirical Bayes shrinkage
- Responsible Gambling Council (Canada)
- ConnexOntario (1-866-531-2600)
- National Council on Problem Gambling (1-800-GAMBLER)
Research and analysis, not betting advice. Past results do not guarantee future returns; bet only what you can afford to lose.