The Buffalo Sabres are a horrible hockey team. It’s not like anyone expected them to be good. Trading valuable players over the last year (see Ryan Miller, Thomas Vanek) for future assets and a lack of pursuit of top free agents shows the Sabres’ organizational embrace of a full rebuild. The team is developing a pool of talented young assets to become a dominant team in the future, whilst playing a mix of washed up veterans and AHLers on the ice this season.
Through 14 games, the Sabres’ tank is in full motion. With a record of 3-9-2 (8 points) the Sabres sit tied for last in the NHL in points, despite having no other team play more games. But the Sabres’ record doesn’t do enough to capture their season so far. The way the Sabres have been bad this first month of the NHL season has actually been one of the most impressively bad months in NHL history. With a Corsi of 36.3% (the lowest all time in a full season since 2007 has been 42.9%), the Sabres are playing at an incredibly poor level compared to the rest of the NHL.
Given this Corsi, it’s not really a surprise that the Sabres only have 8 points. The interesting thing is that with any bad team, there is also a certain degree of unluckiness that is associated with their lack of success as measured by wins/points. In any sports league, the worst teams every year are generally bad, but not as bad as their record indicates. A bad team with decent luck will look like a mediocre team in the standings, leaving the teams with the worst record as either bad teams with bad luck, or mediocre teams with bad luck. With a Corsi this bad, I was curious just how “unlucky” or “lucky” the Sabres have been in points relative to their actual skill they’ve presented, and how the luck of similarly bad teams in the NHL compared.
In order to measure this, I created a simulation of the Sabres season so far, according to what their Corsi would project (basically I created a “season” of 14 games, with Corsi events randomly scheduled based on their rate that they have occurred in the season so far for the Sabres). It’s a fairly simple simulation that doesn’t account for strength of opponent or score affects, but I feel that it will be able to give quality approximations. I ran this simulation 1,000 times and averaged the projected amount of points the Sabres should have so far this season. The result? Given a 36.3% Corsi, the Sabres would be projected to average 8.67 points so far this season (or .62 points per game).
With 8 actual points, or .57 points per game (and thus getting unlucky by .05 points per game), this seems to disprove my theory that the Sabres are an aberration on trends when it comes to bad teams being unlucky. This makes sense looking at the Sabres shot percentage (often a good proxy of luck), which is 5.38%, vs. the league average of around 8%.
It appears that the Sabres might be marginally unlucky, but I also wanted to look at the other “bad” teams currently in the NHL standings. I inserted the Corsi percentages, along with Corsi event frequencies, for each of these teams into the same simulation that I ran with the Sabres, and got the following results:
|Team||Games||Points Per Game||Corsi||Projected Points Per Game||“Luck” Amount|
Although the Sabres might appear to be getting slightly unlucky, we can see that their unluckiness doesn’t even approach the unluckiness of the other teams at the bottom of the standings. This is the interesting thing that makes the Sabres failure so unique. In fewer than 15 games, they were able to clearly establish themselves as the league’s worst team without the unluckiness that is typically associated with early failure.
So yes, the Sabres are an amazingly bad team. Having said this, however, I think it’s clear that the Sabres Corsi isn’t actually their “true” Corsi talent level. To call the Sabres so far a “true” 36.3% Corsi team would seem outlandish, given Corsi results for past NHL seasons. It’s very possible, with smaller sample sizes, for teams to get unlucky and present a different Corsi level than their actual skill (we have to remember, although Corsi is a fantastic proxy for how good a team is, and in the long run represents most aspects of a teams skill, it’s should not be looked at as a direct causation of success). I wanted to take the prior unlikeliness of having a Corsi this low in my calculation of the Sabres’ “true” Corsi talent level.
In order to come up with these priors, I first wanted to look at the chance of getting a Corsi of 36.3 through the number of Sabres’ Corsi events, given various values for a team’s “true” Corsi level. In order to look at this, I built another simulation that played out what the approximation chance of Corsi below certain levels would be for the Sabres through these 14 games (and the amount of Sabres Corsi events) given various “true” Corsi levels.
Here are my results (the colored values are the percent/100 of the time that the Corsi showed up under the threshold value):
Looking at just the .363 level as the “true” Corsi value increases, we can see that the probability of being under that threshold dramatically decreases.
Of course, this by itself doesn’t mean anything in projecting the Sabres actual value. We have to combine each of these true value projections for being under the 36.3 threshold with the value for the actual inherent likelihood of being a team with an actual “true” Corsi of these amounts. So far this season, Corsi league-wide has an average value of 50 (intuitively), a standard deviation of 4.34, and is distributed fairly normally. I was able to calculate the inherent probability of being at the Corsi given the distribution so far (based on the standard deviations away that the true value was) then multiply that likelihood by the percent chances that we found early in the simulation of being below the 36.3 threshold. The final prediction value I get through this multiplication is the relative likelihood for the Sabres’ Corsi level that they actually played at to be that value, given that the Sabres have a 36.3% Corsi through 14 games.
Graphing the End Prediction Values against the True Value we see that the curve resembles a normal distribution, which makes intuitive sense, given that we have been dealing with normally distributed values and simulation:
As we can see, the curve appears fairly un-skewed and seems to peak at 38% (I could calculate a more precise average, but it would be fairly trivial). This indicates that rather than looking like a 36.3% Corsi team so far this season, the Sabres have actually resembled a 38% Corsi team on average. It’s not as big a jump as I anticipated, but it’s a small step, and maybe it changes them from an incredibly historically bad team to just a very historically bad team.
Of course, this is even before adjusting for score effects. When teams are down in a game (as the Sabres are a lot), they tend to shoot more than they would at even strength, boosting their Corsi numbers past their “actual” Corsi talent level. This model predicts the Sabres to be at 34.67% Fenwick (through 14 games) after adjusting for score affects. So as bad as these numbers seem, it’s very possible that they are even worse.
Despite what I’ve shown here, however, I would not expect the Sabres to continue playing at this bad of a level (or even as bad as the 38% level I calculated) for the rest of the season. It’s not that the Sabres haven’t actually played at this bad a level so far this season (I think I’ve shown they have even accounting for chance), but with how low the value is compared to the rest of the NHL, and previous NHL seasons, it would take truly historical ineptitude in order for the Sabres to continue to play at these levels. Yes, I know the Sabres are probably tanking as hard as any team in the fancy stats era has, but it would take a level of horribleness never seen before in the NHL in order to sustain this for a season. It just doesn’t seem possible, but with these Sabres, maybe we can’t count them out.
Image Courtesy of NHL.com
Georgetown University Class of 2016