Saturday, April 13, 2013

The Shootout and Sample Sizes: Addressing the Leafs' Shootout Woes

The Leafs lost on Wednesday night to the New York Rangers in the shootout, dropping their shootout record to 0 wins and 5 losses on the season. The players are collectively 3 for 24 on shootout attemps for a 12 percent success rate – 2nd worst in the NHL. Something must be done! We need to figure out a way to fix the team’s shootout woes lest we be forever doomed to leave with only one point in the vaunted 3 point games!

Would the Penguins trade us Jussi Jokinen? Is the trade deadline already passed? Damn. What about Jason Allison, is he still skating with those cement laced boots that prevent him from skating from the center ice line to the goal? Gosh darn. What about Sundin or Mogilny? Surely in their early 40s they still have some finesse and could be used as a “shootout specialist” if called upon.

Ok...deep breaths, everyone. Long, deep breaths --especially you Nick Kypreos-- the sky is not falling and the Leafs are not necessarily bad at shootouts.

Over the past week or so the mainstream media has been harping on the Leafs lack of shootout success this season. They’ve layered on statistic after statistic, showcasing the abysmal shootout percentages of all our forward corps. All 3 of the team’s shootout goals have been scored by Tyler Bozak, with the rest of the team sitting at a combined 0 for 21.

As a result of all this media attention and my inability to escape the mainstream media (try as I might) I have spent some considerable time thinking about the shootout. The way I see it there are two basic things we want to know:

1. Are the Leafs forwards good, bad, or average at scoring in the shootout?

2. If we are in fact incapable of scoring goals, what steps do we then take to become good at it?

The eureka moment came for me, ironically enough, while watching a Jays game earlier this week. The Jays are off to a 4 and 6 start this season and are being criticized by some as a result of the lofty offseason expectations. However, baseball experts have been continued to emphasize patience with the fan base, noting that baseball is a long season and not to make any snap judgements based on such small sample size of games. In fact, some pundits have suggested watching 50 to 60 games of America’s pastime before drawing any firm conclusions.

Herein lies the crux of my point and the obvious correlation between the NHL and MLB: A shootout between a goalie and a player is a statistically similar data point to that of a pitcher versus a batter. By statistically similar I mean to say that there is a specific outcome that can be attributed to the event. In baseball, a player can either strikeout, walk, or register a hit. To simplify this, we can say that the batter can either reach base safely, or be retired by the pitcher.

In hockey’s shootout there are a number of different outcomes; a player can score with a deke, a backhand shot, a slap shot or the goalie can make a save. In the simplest of terms a player can either score a goal or be stopped – like baseball, two possible results.

Since I’m not a baseball expert I had to look up how many at-bats a typical everyday major league ball player has each year. After reviewing a couple well known batters it appears a typical season entails around 600 plate appearances. As noted above, experts have cautioned fans not to look too closely at baseball numbers until the 50 or 60 game mark – approximately 34% of the way through a 162 game schedule.

Taking that 34% and multiplying it by 600 bats we can conclude that a baseball players batting ability should not be completely judged until he has approximately 204 at bats. Now that’s just for a single season, keep in mind that if you truly want to understand the proficiency of a player at the plate you would use multiple seasons of data – 1000s of plate appearances.

For the NHL I wanted to look at some of the most successful players in the shootout and try to discern if we did in fact have enough of a sample size to call them “shootout specialists”. Below are the career shootout numbers for some notable players:

Zach Parise: 32/68 (47%)

Pavel Datsyuk: 32/69 (46%)

Patrick Kane: 28/64 (44%)

Phil Kessel: 15/52 (29%)

Steven Stamkos: 5/27 (18.5%)

After reviewing the stats, available here from NHL.com, it appears that the most experienced NHL players have had in and around 60 shootout attempts. To put this into baseball terms, 60 shootout attempts is the equivalent of about 15 baseball games.

We’ve all watched TV broadcasts where at the start of a shootout they flash the career shootout record of Steven Stamkos and expose his career struggles. Conversely, fans marvel at Pavel Datyuk’s wizardry with the puck and he is routinely referenced as the top shootout producer in the league. The point is that neither conclusion may be correct. Stamkos may in fact be horrible at shootouts and Daystuk might be the best in the world, but we simply do not have nearly enough data or evidence to make a conclusion either way. Making firm assessments on a players shootout prowess based on 60 attempts is akin to pronouncing Vernon Wells a superior batter to Jose Batista two weeks into a baseball season.

If we take the thought process above and apply it to a single year the amount of statistical error becomes almost laughable. In one NHL season the most any one player is likely to participate in the shootout is 10 to 15 times, the baseball equivalent of 3 games.

To circle back to our original thought: How do we know if the Leafs are good at the shootout and if they are in fact bad how do they improve?

The only way to properly understand the team’s shootout ability or lack thereof is to substantially increase the sample size. One idea is for Toronto to conduct multiple shootouts over the course of training camp and record the results. For example, the Leafs could have all skaters shoot on the team’s goalies at the end of each practice.

Of course shooting on the same goalies over and over would no doubt pollute the data since the familiarity of the players to the goalies would be significant. To help mitigate this the team could look at including their AHL goalies as well, perhaps going as far as partnering with nearby college or junior teams to maximize the goalies present.

For the purpose of our exercise, let’s say the Leafs collect 10 goalies (2 NHL, 3 AHL, 3 ECHL, and 2 Junior). After 14 practices each player takes 2 shots on each goalie (yes, this would take a while). The end result would be 280 individual data points (2x10x14) on all Leaf skaters. If each result was recorded this would represent a statistically relevant amount of data.

By reviewing upwards of 300 shootout attempts across multiple goalies the Leafs could begin to accurately access if their ability in the shootout. Additionally, they could record how a player does when taking a shot versus attempting a deke. Over the course of hundreds of attempts they may see trends arise and subsequently look to coach players to become better.

I certainly have no inside knowledge as to the inner-workings of NHL teams, it’s quite possible they already conduct in depth reporting on shootout ability and tendencies. However, the articles that I’ve read in the past allude to the shootout being an afterthought at the end of practice where players simply experiment while shooting on their own netminder.

After the Rangers loss this week Randy Carlyle was asked about by the media about a stat indicating the Leafs went 24 minutes without a shot on goal. His reply below leaves one to wonder exactly how much statistical analysis is undertaken by the coaching staff:

“Stats are for you guys.”

There has been no indication that the shootout is going away anytime soon, as it will continue to serve as the game deciding mini-game during the NHL regular season. When it comes to the shootout it might be time for the Maple Leafs to heed the example of their MLB counterparts and embrace the new world of statistical analysis and sample sizes.

No comments: