Edmonton Oilers Are Losing Seasons In The First Period


The Edmonton Oilers are a better possession team this year. They are however just as bad at scoring and keeping the puck out of their own net as they have been in the past nine years. Perhaps even worse. There was a piece on the Edmonton Journal today about how bad the Oilers were in first periods this season. It caught my attention because right now secondary stats and primary results are as far away from one another as they can possibly be. The quote I was more interested in was this one:

"They are a team that appears to not be ready to play games, has given up the first goal 19 times, and been outscored 31-15 in the opening 20 minutes."

A recurrent thought through out this season has been that the Oilers do seem to be chasing the game almost every time. I wrote a while back about a twitter exchange with a TSN analytics expert where I argued about scoring effects and quality of shots. To make a long story short, even adjusting with score effects, the Oilers still come close to even on possession numbers like Fenwick and Corsi. As for quality of shots, there is quite a bit of writing out there by the analytics community making a strong case that, when comparing numbers for ALL NHL players, shot quality is not a reproducible skill in terms of statistical significance, at least not with current methodologies ( there is just no practical way to measure it).

Thus, instead of focusing on how the chances are created and how good players are at putting them in the net, the advanced stats community has opted to pay attention to volume of shots, a very reproducible skill where game situations become a bit of a moot point. Possession stats like Fenwick and Corsi measure exactly this, how many shots a team is directing shots at the opposing net vs how many they are allowing (be it blocked, missed, or on net). Again, a strong case has been made that the more a team shoots on the other net, regardless of quality of the shots, the more they have the puck, the more they control the game, the more they score, the more they win.

Players like Steven Stamkos, Alex Ovechkin and Sidney Crosby are gifted talents for whom we can make a case that their finishing skills are above average and thus important.However, an important factor that makes “shot quality” difficult to measure is game situations: a bad bounce creating a 2 on one, a goalie having a bad day, a ricochet off a skate, etc. There are just too many factors creating scoring chances in a hockey game to accurately and reliably measure and be able to create a significant statistic. A good statistic will expose a trend from within a large sample size, discarding as much noise as possible.

More from Oil On Whyte

Going back to the disconnect between possession stats and results, a decent predictor of scoring success,  PDO, a measure of “puck luck” that is nothing more than a type of plus/minus that adds shooting percentage and save percentage when a player is on the ice, is a good place to look at when trying to explain the Oilers woes. Currently the Oilers have the worst PDO in the league at 96.8%.

The concept is rather simple: every shot produced either ends up as a save (SV%) or as a shot (goal, missed, blocked, making up shooting %). Because of this we say that PDO regresses to 1(or 100, or 1000, whatever notation a writer wants to use, but it means the same). This is how I understand it. In other words, the league averages for SV% and sh% account for 100% of the shots taken. So if a team has a PDO higher that 100(or 1, or 1000 and so on), either their goaltending is above average, or they are on a hot scoring streak, or both. Conversely, if a team, like the Oilers right now, has a PDO below 100, they are either getting poor goaltending or their players are stuck on a rut, or both.

In the Oilers case, it is both. Right now the team has a 5 on 5 SH% of 6.68, ranked 27th in the league, and a SV% of 90.14, good for 29th overall. Those two numbers combined gives us the PDO of 96.8% we talked about earlier. Now it is important to keep in mind that, like any other stat, the larger its sample size, the its predictive ability. As I mentioned before, PDO is a nothing more than a more complex plus/minus and as such it is subject to limitations and context of use, such as wild variations between individual games and individual players. It is still a useful stat when looking at why a team is not scoring and losing.

For comparison, Calgary is riding a high PDO of 101.0 (8th overall) with a high sh% of 8.96, fourth in the league and a middle of the pack SV% of 92.03 (18th). They have allowed the 13th most goals against in the first period while netting 16 of their own. Also their top 2 d men are having career seasons. Their horrible possession stats and over performing players do suggest a team going back to their average and crashing down, but as explained before, the larger the sample size, the better the predictions, and such a crash could not come up until the end of the season, or even not until the beginning of the next, all things staying equal. All those stats tell us is that, all things being equal, Edmonton should, at some point in time, improve their record, and Calgary should regress. But the stats don’t say WHEN that will be or how dramatic the rise or fall would look like.

A Tree Born Crooked Will Never Straighten Its Trunk

Having Jim Matheson’s quote at the beginning of this post in mind, I went and looked at the goals against allowed by the Oilers during the first periods of every season. Since the 2009-2010 season that gave us Taylor Hall, the Edmonton Oilers have finished First in Total Goals Against during the first period in 2 seasons (including last season), and have been in the top 5 on 3 of the last 5 seasons. Of all the 25 teams (including repeat offenders) who were the top 5 in that category each season, only 4 made the playoffs and 12 ended up being lottery teams. The one things we can conclude for certain is that allowing so many goals in the first period tends to end up badly more often than it does good.

Eakins’ charges seem to have trouble getting their legs underneath them until the game is well on the way. Why is that? who knows. If anyone has the answer I recommend you knock on the door at the kingsway offices and ask for a truck load of money in exchange for your knowledge.

Another piece from the Journal from yesterday talked about how the Oilers, despite being close to trading chances with the opposition, their goal differential hasn’t translated. One reason discussed in the article, besides poor goal tending, had to do with the distance the Oilers are shooting from as a possible explanation for the disconnect between scoring chances and goal differential. Here’s the quote that stuck with me the most:

"Under Dallas Eakins, Edmonton’s defence has done a pretty decent job in suppressing shots against from point-blank range; the Oilers’ opposition lands shots at that distance eight percent less frequently than the NHL average. Shots from the slot are and point are bang-on with the league average. This is a massive shift from last year, where the Oilers allowed far more than an NHL-average number of shots from the crease."

The same data is often prone to different interpretations. This is true for any type of analysis. And it becomes more frequent the less technical our interpretation is. In the case of the above quote, the argument made is a reasonable one. And just as reasonably we can make a different inference, based on everything else written on this post: the reason Edmonton deals with less shots from point-blank this season can be because teams score early and often and are rather content with limiting their shots to the outside and dumping the puck in wearing the Oilers down. Consequently, the Oilers may have allowed more shots from the crease because Ben Scrivens and Viktor Fasth arrived mid-season and had a stellar run, where instead of goals going in on the first chance, more bounces were created at later parts of the game, and thus more shots taken by the opposition on the slot.

And that’s the fun of analytic in sports, unless you are getting paid to do it or are trying to prove a point really hard, almost beyond doubt, you don’t need to have formal training in statistics. If you understand the numbers and what they mean, you can make a solid argument for any opinion you hold dear. Sports analytics are not based on experimentation and hypothesis testing, they are just descriptive statistics based on historical data, some of which turn out to be decent to good predictors of events. A lot of work goes into them yes, but they are not as difficult to produce nor understand as, say, how to make a robot land on a comet (I really don’t, yet, maybe never).