Hmm.

I’ll give you that the shootout is a crapshoot (albeit a weighed one), but dismissing overtime entirely still sticks in my craw. While overtime wins may be very dependent on luck, overtime goals themselves are still valid hockey goals and should be included in goal-differential calculations.

Are the Habs truly a terrible team, or a slightly-better-than-terrible team with a particularly flukey distribution of goals-over-gametime that results in them having scored an inordinate number of them past the 60 minute mark? They’ve clearly been a bad 5-on-5 team (and, scarily, getting worse) but the Leafs have a 0.82 GD/game if you do count all non-shootout goals, and the Habs have -0.52 GD/game - not anything good, but nowhere near as bad either. (Both teams have a single empty-netter, which cancels out.) Of course, that last game by the Leafs really didn’t help their GD, and the Habs are showing no signs that they’re playing any better, but the Habs’ overall body of work does seem slightly less awful. shrug

You missed my point. OT goal differential, taken by itself, isn’t predictive, sure. The sample size is way too small, so that’s not exactly a surprise.

The point is that OT goals (and not shootout goals) are still goals.

Overall goal differential is predictive. And while I wouldn’t use OT goal differential on its own to predict anything, not even OT success… I don’t see why there’d be much validity in removing overtime goals from overall goal differential either. I don’t see how eliminating OT goals from overall goal is going to make it a better predictor. It’s no surprise that teams that sneak into the playoffs on OT wins don’t do well, because OT wins are one-goal wins by definition so they don’t do much for your goal-differential. That, and teams that ‘sneak into the playoffs’ are going to be low seeds by definition and can’t be expected to do well in the playoffs regardless of how they scratch the points they need to barely make it, whether it be OTWs, OTLs, or otherwise.

Montreal isn’t a good team right now (they are merrily headed for a goal differential around -40, after all) but I don’t see any evidence that they are as bad or worse than the -65 GD/82 Leafs just because an unusually large proportion of their goal production happened to fall in overtime. That just strikes me as being a flukey distribution of goals, the same sort of thing that kept the Sabres out of the playoffs over the last couple of years despite positive GDs when negative-GD teams make it. If the Habs had flukily scored those four goals in the last 3 minutes of regulation, or even between the 12th and 15th minute of the second period, odds are the Habs’ record would be very similar but I doubt we’d be having this conversation.

I don’t see why one should be considered in overall GD and the other not. Overall GD is very predictive, so what inherent flaw does an OT goal have such that removing it from the overall GD predictor would improve it that, say, a 4-on-4 goal scored in regulation wouldn’t have? Of course, as a group, they don’t have much weight. They’re a very small sample of the data, after all. But do some types of goals individually have less weight in the total-GD predictor than others?

There’s not enough data to decide either way. For example, if you tried to correlate goal differential in the first 2 minutes of the second period to the rest of the game, you’d find a tenuous relationship at best, simply because of the small number of data points. It’s tough to tell whether 4-on-4 OT correlates with 5-on-5 or is an unrelated animal. My instinct is that it would be similar, but there’s just not enough data to know.

Just checked… last year the 4-on-4 to 5-on-5 goal differential correlation was 0.18 (R^2 = 0.034). Given the small sample, that sounds to me like there’s at least some relationship.