For Entertainment Purposes Only¹

So back before the entire Oilersphere exploded in fury due to the never-ending debate over the nature, usefulness, and validity of new-age statistics, Matt Fenwick wrote a little post on BoA on the Oilers' shootout record. I felt it was kind of a tired pick at an old scab that really didn't tell us much about this year's Oilers and Flames, and said as much at the time. I wanted to write a post at some point about unrepeatability vs. unpredictability and the perils of ignoring those sorts of things, but that kind of flew out the window because of the ensuing brawl. That being said, I'm still interested in trying to figure out how a team's going to do in a shootout, even if small sample sizes and the psychological nature of the shootout put a serious cramp in any reasonably scientific method of prediction. Really, even if you were to run statistical analysis on the first three years of the shootout in the NHL, you're not going to pick out many trends, other than the Stars being fucking awesome at them², so take everything below with a hefty grain of salt.

Last year, Tyler Dellow developed a model for predicting shootout win percentage based on a team's shooting and save percentage, which actually worked not too badly based on the Oilers' first three years when I ran them³, so I thought I might at least try to take a stab at what the Oilers' SO record might look like in 2008-09, based on their known roster. It doesn't account for things like coach selection bias, shootout length, and game-over-game psychological factors, but it at least gives us a more likely idea than assuming the entire NHL goes 5-5. Below, we have the collective shootout stats for your 2008-09 Edmonton Oilers:

The first chart is the success rate of the Oilers' shooters, followed by the rates of the two goalies. The "adjusted total" for the keepers assumes (a) that the Oilers play an average number of shootouts for them (14), and (b) that Garon plays 10 to Roloson's 4, based on an approximate 55-27 game split (I have no way of accounting for Jeff Deslauriers, so at the risk of looking like a severe hypocrite, I ignored him). I initially thought Garon's mind-blowing .938 SP last year would skew the numbers, and while it did, it might not be as bad as you might think: his other two years were 0.812 ('06) and 0.611 ('07)⁴. Meanwhile, two of Roloson's three SO years are over .700 (0.550, 0.714, 0.730).

I also didn't attempt to adjust for shooter frequency, because frankly, I'm not sure how. Obviously, we'll probably see more of Nilsson and Horcoff than we will Pisani, which will probably affect the numbers, but by how much? Last year's numbers are skewed by Horc's injury, and it's also difficult to predict how often the game will go two shooters, three, four, etc. My best efforts usually wound up churning out numbers well over 40%, which while tempting, feels a little too optimistic, especially considering two of the Oilers' three shooting years are right in the range of the average above (35.8%, 27.3%, 36.9%). And while it's tempting to throw out all the one-and-dones, there's bound to be at least one odd choice or eight-round shootout, and really, they don't skew the data too much (they decrease the team S% by 2.0%, which doesn't change the number of final wins, based on the model used). The only player I've excluded who has taken a shot in the last three years is Gilbert Brule, because he's not, as of this point, on the team.

Anyway, given these numbers, I would say the Oilers' shootout record next year will fall between 0.625-0.650, based on Tyler's chart; given an average of 13.7-call-it-14 SO games, that comes out to 9-5. Not as good as last year, but not as bad as a coin-flip of a league-average number of SO games.

Now, I'm open to criticism here. Small sample size, incorrect shooter distribution, incorrect goalkeeper distribution, use of a player's career averages over their most recent production level, incorrect shootout frequency, even a disagreement over whether Tyler's model itself is valid. I don't think any of those things would change the prediction by more than one win -- the choice of model and various psychological factors, if anything, would be the ones to skew this ridiculously -- but if someone else can suggest something better, I'm willing to make adjustments, within reason. But just so we're clear here, this isn't meant to be anything more than a rough guess that takes into account actual team skill level, rather than assuming a saw-off of a league average number of shootouts.

For the record, using the same admittedly sketchy methodology, the Flames come out with a predicted record of 2-6 (figure below). Same caveats apply, including the exclusion of Anders Eriksson (waivers), the bizarre shooting frequency of Todd Bertuzzi with his previous employers (15 shots over three years, second amongst active Flames), and the little fact that Kiprusoff has gotten better at the SO with each passing year (0.478, 0.562, 0.647). Assuming he plays closer to last year's level, or about League average, bumps the record up to 3-5, which still isn't much, really, but then looking at the shooter list, Jarome Iginla is the only player whose career numbers exceed the League's three-year average of 32.9%, while Dion Phaneuf leads the ohfer brigade with six career misses. So while the goalie getting better certainly helps -- the Flames have improved their SO record dramatically every year since it started -- it only helps as long as his skaters can actually, you know, score.

¹ - Rejected titles: "Night of the Living Paretheses", "The Footnotes Came From Below"

² - After being a top-five SO SV% team the first two years, the Stars fell back to the middle of the pack in that category this year. Marty Turco is noted for trying to psyche out shooters by turning to one side until they reach the blue line, and was wildly successful at it the first two years (0.750, 0.761), but this year, he came crashing back to Earth (0.609). If I had to make a guess based on that theory, I'd say the shooters are now aware of this, thanks to video and experience, and have made the mental adjustment, though it also bears pointing out that he only took six shootouts last year, as opposed to nine and twelve the first two years, respectively. We'll see how the results shake out next year to see if that theory holds, or if it was just a blip; my guess is that Turco won't be below League average for a second year in a row.

³ - '06: PW ~= 0.43, AW = 0.438; '07: PW~= 0.40, AW = 0.500; '08: PW ~= 0.82; AW = 0.789

⁴ - Of further interest to Oilers fans may be this other post by Tyler on the subject of left-shooting skaters vs. right-catching goalies in the shootout, and the implications for Mathieu Garon.

- DoogieSeptember 15, 2008 - 11:27