As I poke around in the data from this past season, I’m coming across all sorts of interesting things. One of them is with the Oilers and what happens after they lose an OZ faceoff at 5v5. You can see that they actually started the season really well – they were touching 60% through the first couple months and then gradually petered out as the season went along.
This graph puts it a little more starkly.
That 60% number through December 7 is fabulous. If you’re up above 60%, you’re into the league’s elite in this aspect of the game. The 36% number is astonishingly bad – the worst number I’ve ever seen for a full season is Colorado’s 42.8% in 2009-10, so it’s pretty impressive to put up something seven points below that over 50 odd games.
As I am often “politely” reminded, Corsi% isn’t goals. We do know that, over time, Corsi% and GF% end up pretty close together. When you start slicing the game into segments that are this small, it takes an awfully long time to pile up enough Corsi events for the Corsi and the goals to converge. Just to give a sense of the scale of the problem here, my database, which covers 2007-14, has 38,122 Corsi events that I treat as OZ- faceoff events – shot attempts happening within 21 seconds of an offensive zone faceoff loss. There were 133,492 Corsi events at 5v5 in the NHL this year.
In other words, seven years worth of OZ- events for a team produces about as many Corsi events as we’d usually see in 23 of their games. We know that, in the short run, Corsi% and GF% are not necessarily tightly correlated – the percentages can overwhelm things. A bad Corsi% team can have a good 25 games in terms of the goals for and against or vice versa.
It seems to me like this would make it a somewhat difficult aspect of the game to coach because the connection between what you’re doing and the results that you achieve is pretty tenuous in any given season. In terms of the tactical decisions that have to be made here, people learn from watching these things play out and there’s simply not enough of them in a hockey game, in a season’s worth of hockey games, that someone who watched all the games would necessarily see teams that employ tactics and players that lead to the Corsi% being lousy post-OZ loss get punished. You can get it wrong and do fine, you can get it right and get killed. The trick is to identify enough areas where you’re making the right decisions that you build an edge that’s unlikely to be overcome by misfortune.
So it’s worth pointing out that this does seem to be the case over time. I’ve got seven years of data and I created groups of teams based on their Corsi% following OZ losses. The more Corsi% = more GF% relationship with which we’re all familiar holds.
Happily, I’m not required to resort to statistical flimflammery such as randomness and regression to the mean with respect to the Oilers lousy Corsi% after losing an OZ faceoff loss post-December 7. They got destroyed. Through December 7, they’d scored a goal and allowed a goal post OZ faceoff loss. From December 7 through the end of the year? They scored one goal and allowed six.
What we don’t know – and what we can’t realistically know unless someone watches all of the bloody faceoffs – is whether this was tied to a tactical change or luck (luck can affect Corsi% in small samples, obviously). There’s data that isn’t being gathered – zone time, with time stamps for when the puck entered/exited different areas of the ice – that would give us a much better sense of things but, unfortunately, Corsi% is where the data currently runs out.
There’s a view, held by some people who I respect, that the Oilers became less aggressive as the year went along. I’m always very skeptical of people who jabber about systems because I think it takes a lot of work to spot them and I’m skeptical that the people taking about systems have done the work that I’d feel is necessary to talk about it with any degree of intelligence.
That being said, we know that they abandoned the swarm about ten games into the season, that they were bleeding goals and that it seems reasonable to think that they might have wanted to cut risk out of their game given the goaltending that they were getting early. It doesn’t seem implausible to me that they would have tried to cut risk out of their game elsewhere as well. One way to do that is to try to limit the dangerous rushes that you give up.
I’ve gone through the goals that were scored off Oilers OZ faceoff losses and assembled a little video. I want to underline that I don’t think there’s anything definitive here, it’s just a way to add a little visual context to the issue.
The Anaheim video that I mention in that video can be found here.
Does any of that mean anything? I can’t really say for sure. Some of those, they look like they put some pressure on, although even that is going to result in goals sometimes – we’re trying to identify the marginal goals for/against here, which is a trickier thing to do. I have big doubts, given how badly they were outshot in these circumstances, that the Oilers were making the defensive zone exits particularly difficult for the opposition. Whether something changed around December 7 or whether they got lucky early, they got run over. The goals are just a symptom of an underlying issue.Email Tyler Dellow at firstname.lastname@example.org