If the Oilers’ advanced stats are so good, why don’t they win hockey games?
9 years ago
The comments section on this website was a little testy yesterday, and analytics in hockey were a target for many. It’s understandable; there has been a lot of talk about the Oilers’ improved underlying numbers while at the same time the team isn’t winning hockey games. So what’s going on with that anyway?
An Early Hole
There are a lot of things going on that simply saying “the Oilers have improved underlying numbers” doesn’t cover. Among them is that brutal run at the start of the year.
*Need a quick Fancy stat tutorial before reading, click here.
Edmonton started the year with five losses. Outside of the debacle in Los Angeles the Oilers were not bad; competitive against Vancouver (both of those games were one goal losses) and the superior team in many ways against Calgary and Arizona. However, the goalies posted a glorious 0.845 save percentage in that span and no team wins games when that is happening. Some of that is on them (Scrivens and Fasth both legitimately struggled) and some of it is on the defence (Dallas Eakins said his team had to find a way to eliminate “the big mistake” and he was right on that) but the Oilers actually did a pretty decent job of spending more time in the opposition’s end than in their own end over that stretch.
That’s my view; nobody here is required to accept it. But assuming for a moment that I have it right on those contests, here’s the thing: when a team plays pretty well and has five losses, there isn’t a grand and impartial board of judges who will make sure that the team will make up those points later. The dice have no memory. So when a team starts five games below 0.500, if it wants to be above 0.500 by the 20-game mark it needs to go 15-10 the rest of the way, winning 60 percent of its games.
In other words, the big problem with an early hole is it can be awfully hard to recover from it. Edmonton has gone 6-6-1 in the 13 games since; in that span their even-strength Corsi has been 49.7 percent and their even-strength Fenwick number has been 49.4 percent. That to me looks a lot like a slightly below average team going just slightly under 0.500.
Analytics & Schedule Effects
None of these things happen in a vacuum, and while the numbers are absolute truth (X happened; we can disagree about the reasons or the importance of it, but it happened) it can be difficult sometimes to interpret them correctly.
For example: Schedule effects. I was harder on Ralph Krueger than I really should have been during 2012-13 because I failed to properly account for the fact that the Oilers were playing solely against Western teams. That matters to results.
Edmonton has played one game total against clubs that were in the playoffs in the West last year. If we continue to isolate those 13 games, they’re running a touch under 0.500 in a schedule that featured a friendly 8/5 split between home games and road games and saw the meet the following clubs:
- 2013-14 Eastern teams that made the playoffs: Tampa Bay, Montreal, Philadelphia, Boston, N.Y. Rangers
- 2013-14 Eastern teams that didn’t make the playoffs: Washington, Carolina, Buffalo, Ottawa
- 2013-14 Western teams that made the playoffs: None
- 2013-14 Western teams that didn’t make the playoffs: Nashville (x2), Vancouver, Arizona
We can talk about the hot starts in Vancouver and Nashville, but even so that isn’t exactly a murderer’s row of opponents. Eight of the 13 games came against teams trying to climb back into the postseason, and the Oilers had the additional advantage of catching New York in the second half of a back-to-back.
This has been a favourable chunk of the schedule, the part of the year when the Oilers needed to go well above 0.500 if they were in actuality a 0.500 team. Instead, they are sitting with a 6-6-1 record and even-strength possession numbers that show they’re basically full value for it.
There are some complications to putting it that plainly. The power play is on a cold streak and probably deserves better than it’s got so far; over the 13-game stretch we’re looking at the Oilers have all of five goals on 66 shots, a 7.6 shooting percentage which is well below the norm and which will probably come around.
Even so, that doesn’t excuse a pretty mediocre performance at evens over a soft part of the schedule.
The Numbers are Crap!
No, they aren’t.
It’s often said that unlike baseball, which is highly repetitive, hockey is super random and difficult to analyze. That’s self-serving nonsense from the Luddites and should be disregarded. Hockey is a series of constantly repeating events, like so:
- A team that doesn’t have the puck in the defensive zone has twin objectives: prevent a good scoring chance and regain possession. This happens at least dozens of times in every single game.
- A team that has the puck in the defensive zone has one objective: to advance it to the neutral zone. This happens at least dozens of times in every single game.
- A team that doesn’t have the puck in the neutral zone has twin objectives: prevent an opposition zone entry and regain possession. This happens at least dozens of times in every single game.
- A team that has the puck in the neutral zone has one objective: to advance it to the offensive zone. This happens at least dozens of times in every single game.
- A team that doesn’t have the puck in the offensive zone has twin objectives: prevent an opposition zone exit and regain possession. This happens at least dozens of times in every single game.
- A team that has the puck in the offensive zone has one objective: to create a scoring chance (well, actually to score, but this is done by taking shots from a dangerous area). This happens at least dozens of times in every single game.
This happens quickly, and the tactics teams use to achieve their objectives can vary, and special teams need to be accounted for but it’s not like we’re trying to analyze Calvinball here. The game can be broken down into discrete moments, and a team’s performance in those discrete moments can then be analyzed. It’s extremely time consuming (I’d hoped to break down the game in exactly this manner this year, but simply don’t have the time to do this much manual tracking) but it’s entirely possible and when the NHL brings in its new analytics package this is the way every numbers guy on the internet will be studying hockey, no matter how much Brian Burke huffs and puffs.
In the meantime, we use things like Corsi and Fenwick as shorthand. It’s not perfect, but it does capture a lot of the above simply by explaining how often given teams (or given players on given teams) are succeeding in the core objective of getting to the offensive zone and delivering the puck/preventing the opposition from doing the same.
And that matters a lot. Teams that constantly let the opposition setup shop in their own end lose games because no matter how good their defensive structure is they’re setting themselves up for failure by conceding the territorial battle. Hot goaltending or a good structure can cover that for a while, but not forever. The same is true for teams that can’t manage to get to the opposition’s end with high frequency; really good shooters (or a run of pucks going in) can cover that up for a while but it’s hard to win games consistently without an offence that can create chances in volume.
Looking at the numbers, the Oilers certainly seem to be better at those two things than they were a year ago. They’re not good enough at it, and it’s a good bet they’re going to get worse when they start playing against good teams, but as unpopular as it is to say they are making progress. But while ‘just under 0.500 in a soft part of the schedule after an 0-5 start’ technically represents progress it understandably doesn’t make the fans very happy. I’ve hand-waved away that 0-5 start to some degree in this piece, briefly explaining my view of what happened without going into a ton of detail, but regardless of the cause those five games form part of the team’s record. They also go a long way toward explaining why Edmonton is where it is despite improved (but still not great) possession numbers.
RECENTLY BY JONATHAN WILLIS
Recent articles from Jonathan Willis
