There's a million things about this game and this team that I'd love to unpack, but earlier today I posted a question in the comments of Scoring Chances: Game 30, and then Elvis added his own set of inquiries. The whole thing got me thinking--and I recall someone asking about why the heck should we care about scoring chances--so I said screw it, let's get into some tables.
[N.B.: If anybody else did this, post your results, too. Replication is too often overlooked.]
First, let's start with Neutral's own perspective of looking at scoring chances:
"While chances generally line up with shot plus-minus, or Corsi, in the long run, shot quality can play a significant role in determining single-game outcomes. Scoring chances essentially operate as a check on Corsi to verify whether the team generating more shots in a given game, or over a series of games, is getting those shots from dangerous locations. At the single-game level, I think it's the more valuable metric and it's certainly far more valuable than traditional plus-minus."
I would push this a little bit further with some specifics. When I see a regulation loss AND I see the Sharks dominate possession (via Corsi or Fenwick), I tend to expect to be out-chanced. That is, scoring chances (i.e. high-danger shots / "shot quality") fills some kind of explanatory gap when LOL FENWICKS still leads to a regulation loss. On the other hand, I expect that if the Sharks are dominated possession-wise, but get a regulation win, it's because they out-chanced the opposition. Caveat, yes, I think special teams plays a big role, but for the moment, all of this data will be even-strength (EV). Possession metrics only really matter at evens, since winning possession 4v5 is like totally wtf.
So, here's the question. Do we see any correlative pattern between regulation wins (Wins) and winning possession (>50% FF%) and winning chances (>50% Ch%)? This is a mishmash of some of Elvis' questions. (And, btw, I'm not a professional statistician, though I can read numbers.)
[Method, quickly: This is totally preliminary--sample sizes. I just went with the first 30 games of this season because holycrap extraskater.com makes it easy. Also, I used Fenwick because well, just because. As it turns out, if you use Corsi, one pair of wins switches places (Cell 1 Cell 3; leading to the exact same 2x2 results.]
I coded each game so far this season with three digits: X, Y, and Z; where X is indicates a Sharks regulation win (0-1), Y indicates a FF% win (0-1), and Z(ed) indicates a Ch% win (0-1). I created a 2x2 (yes!) where each cell contains the regulation win-loss record given the FF% win-loss and Ch% win-loss parameters.
In 30 games, the Sharks have 14 regulation Wins (and 8 losses). Let's setup some expectations/guesses/half-baked hypotheses:
HBH1: They should win games when they win both Fenwick and Chances.
HBH2: They should lose games when they lose both Fenwick and Chances.
HBH3: They should win and lose an equal number of games when they win only Fenwick or Chances. (Just for giggles.)
So, in HBH1 conditions, they went 9-6. Pretty good, though something might be weird. We'll come back to this. In HBH2 conditions, 0-7. Yup, expected. In HBH3, 4-1 and 1-2 (5-3 overall). Sort of 50/50, but really small sample.
Also, in these 30 games, winning FF% and Ch% did not correspond 8 times. Sample size leads us to no strong conclusion, but it's suggestive evidence of correspondence between FF% and Ch%. Notably, the Sharks were 4-1 where they lose the Fenwick count, and win the Chance count.
For what it's worth (and Elvis asked-ish), in all games where the Sharks win FF%, they were 10-8. In all games where the Sharks win Ch%, they were 13-7. Not enough data. I suspect that in the big picture it will be real tough to distinguish which is a significantly better predictor.
What I do think this very small bit of data DOES illustrate is two weird sets of cases. The first set are those games where the Sharks lose in regulation, but capture FF% and Ch%. I looked at those 6 losses, and thought, well, they are shooting the puck a lot and are doing so from dangerous areas. What can account for the loss? My suspicion was Sh%. Indeed, the Sh% for those 6 games? That's right, 5.8%. In addition, the Sv% was 87% (i.e. PDO sat at a crappy 928).
Now, the other set of weirdness are games where the Sharks get a regulation win, win Chances, but lose the Fenwick count. These are odd because the Sharks are a good possession team, and we'd expect territorial wins to come with lots of shots. In those (only) 4 games, they shot 15% and had a Sv% of 98% (i.e. PDO at 1130!). Now, of course, losing FF% in these cases could not only be an effect the high Sh%, which, I think, typically results in lower FF counts, but also scoring effects.
Bottom line: I think the more precise value of "scoring chances as a check" is that we can more strongly make the case that a regulation loss can be blamed on some crap bounces or a goaltender having an abnormally good game, rather than poor Sharks play.
I don't about you, but I'm cool with that. Spreadsheet here.
[Last thing for Elvis: I haven't done the counting on this, but one could add up missed "tip-ins" as a measure of bombing away from the outside--which, of course, takes work to setup--AND still working down low to get in position for a tip, even if it misses. My two cents.]