Accuracy
Why trust our Goal %?
Our top pick scores 1.7× as often as a random pick.
Tested on all 543 Hockey Challenge lists from last season and this season.
How often the pick actually scored
One player per list. Share of lists where that player scored. Whiskers show the 95% range.
Differences among the sensible systems are within noise; every one of them beats a random pick, and last-10 alone or ice time alone are weaker.
When we say 20%, it happens about 20%
Every player on every list, grouped into six equal buckets by our Goal %.
| Bucket | 1 | 2 | 3 | 4 | 5 | 6 |
|---|---|---|---|---|---|---|
| We said | 4% | 9% | 14% | 20% | 26% | 33% |
| Actual | 5% | 7% | 14% | 18% | 24% | 33% |
Six equal-sized groups of player-games, sorted by Goal %. Predicted should track actual; at every level, what we said is close to what happened. At the very top we said 33% and it happened 33%.
How we tested
Real lists
Every Hockey Challenge list from the 2025-26 regular season (501 lists), plus 42 from this season before we launched: 543 in all.
No peeking
Tested on every 2025-26 Hockey Challenge list, using only stats available before each game.
Held-back check
On the 207 lists we kept aside while building it, the top pick still scored 30% against 18% for a random pick.
By list: List 1 36% vs 27% random · List 2 29% vs 18% random · List 3 25% vs 10% random.
7 in 10 top picks still miss. Hockey is random; this is an edge, not a guarantee.