3-2-1-0 would be 3 points for a regulation win, 2 for an OT win, 1 for an OT loss, and 0 for a regulation loss. You're correct about the other two.
I'd guess they'd play 4-on-3 and then 4-on-4 when the penalty expired until the next whistle
Actually there's some research suggesting that a 3-2-1-0 system might change how the game is played. Scoring rates drop off at the end of regulation in tied games between teams in difference conferences because they're playing for the extra point. If teams got 3 points for winning in OT, this would probably go away. …
Last year there were 129 games that ended in OT and 178 in a shootout. Slightly above average
In a typical season about 20-25% of games go to overtime. That's about 280 or 290 games per year.
No clue why they keep it around. It made even less sense in the early 2000s when a tie and an overtime loss would both earn you 1 point.
it's because the AHL's hybrid model is 7 minutes of OT. Three minutes of 4-on-4 and then four of 3-on-3
War on Ice just released an enhanced standings page where they show what the standings would look like if there wasn't the OT loss point, or if there wasn't a shootout, or if there was a 3-2-1-0 points system: http://war-on-ice.com/standings.html
I get it out of the play-by-play data since 2005/2006. I have some code the extracts the amount of time played at each strength level (5-on-5, 5-on-3, etc.) as well as the number of goals scored during each. You're right that 3-on-3's are rare, although there's still a lot of them over a 10 year period, particularly…
Thanks for catching that. I forgot the word "which" when I was editing.
That's a good point and just reinforces the fact that these estimates are ceilings. The true percentage of OTs that make it to a shootout would probably be even lower.
Yep that's exactly right. This article shows that goals in the NHL are Poisson distributed: http://hockeyanalytics.com/Research_files…
War on Ice just added an enhanced standings page to their site: http://war-on-ice.com/standings.html On the far right of the table they have the standings if there were no shootouts, if there was a 3-2-1 point system and if there were no points given for losses.
I used the base graphics in R.
I looked at every attempt. The first three shots (the first round and the first shot of the second round) always fit in the "neither" category. The fourth shot fits in the must-score if the 4th shooter is on a team that's down 2-0. The fifth shot is a must-score if his team is behind; it's a can-win if he's ahead by…
Maybe, but for how few games go to a shootout per season (something like 10 or 15 games per team) and the fact that the specialist could only shoot once anyway, it's probably better to use that roster spot on somebody who's going to help prevent games from getting to the shootout in the first place.
If it were just from one or two seasons, then sample size might matter. But this analyzes over a thousand shootout games, a thousand (or several thousand) shots in each of the three shootout circumstances. If there really is a statistically significant difference that's more than about 1%, that should be plenty of…
It's a decent point, except that teams tend to put their better shooters earlier. So if the early guys didn't score, then it's probably not very likely that the later guys (who the coach didn't believe were good enough to shoot in the first three) would score.
That's really great! It's such a good example of how the Central Limit Theorem can work, even with variables that are bounded. The standard errors for all the differences are in the text in parentheses.