Curiosities
Things that I think are fun and interesting, but not that important. Updated daily.
Third-period comebacks and blown leads, compared with teams of the same quality
A team trailing after two periods wins about one game in seven, and good teams win more than that, so counting comebacks mostly re-ranks the standings. Each team here is measured against what teams of its calibre manage instead, using a line fitted on the two seasons before this one. The numbers are games above or below that line.
1312 games in 2025-26. Expectation fitted on 2023-24 and 2024-25, so a team cannot move the bar it is judged against. Teams in colour differ from their own expectation by more than chance explains; everyone else is having a normal season, whatever the chart looks like.
Who’s lucky? Who’s gonna regress to the mean?
This one is PDO. Add the share of a team’s shots that go in to the share of shots against that its goalies stop, times 100. By definition, the league average every year is 100. People often call this metric a metric of “Luck” as it sort of seems to quantify something beyond skill. Teams pretty much always gravitate to land between 98 and 102, so anything above or below that is seen as good or bad luck. I think it just reflects the crazy amount of noise that exists in hockey, given it’s such a chaotic game. I think it’s more interesting to see who is likely to regress to the mean, who is on a heater right now, who is getting shafted, and call it a light quantification of noise.
1311 games in 2025-26. The crest is the figure for the window, built from totals rather than an average of per-game rates. The bar is the middle half of the team's single games: wider, on its own, than the gap between the best and worst teams over a season. Last 10 is each team's own last 10, so a team with a game in hand is still judged on 10; only a full season has to land on exactly 100. All situations, including the empty net, which analysts usually cut along with everything above five-on-five.