A look at the “analytics” used for the Semin contract

As a hockey stat guru, it warms my heart to hear that front offices around the league are keeping an open mind into hockey analytics, which is  why I was excited to hear that the Hurricanes were using certain “advanced” analytics to back-up their decision to sign Alex Semin to a one-year contract worth $7 mil. This fact was revealed by Canes PR guy Mike Sundheim on Twitter and later confirmed by GM Jim Rutherford in a press conference. It’s always good to do you homework before you make a big signing and it’s good to see that the Canes did some fact-checking before signing Semin to this deal.

Blogs like mine and many others across the Internet will always do a thorough background check on a player before deciding if he is a good fit for the team they root for by using shot-based metrics such as Corsi and Scoring Chances along with looking at a player’s scoring rates. If you listen to the press conference from Rutherford, you will learn that he performed about as thorough of a background check with Semin as you could imagine but the type of analytics they used is a little different from the ones that you may be familiar with on this blog and other places across the Internet.

I looked back at Sundheim’s tweets to see if he mentioned anything about the type of stats the front office was using and he mentioned quite a few. First of all, it appears that the Canes were using percentiles to gauge a player’s value rather than just simple rankings. A percentile shows the percentage of a sample lower than the point (or in this case the player) of interest. They were also apparently looking at how often a player scored in “high pressure situations” and against “top forwards on the opposing team.”  Plus/minus for these situations were also looked at. These stats were used as a way to combat the opinions of Semin not being “clutch” in the post-season or when the game was on the line.

Per Sundheim, “high pressure situations” are when the game is tied in the second and third periods or when the game is close during those frames. Overtime is also included in this, obviously. The top 2-3 forwards on each of the opposing team are selected by whoever comes up with these statistics. It is also not clarified whether or not the team is looking at the amount of goals/points a player scored or the amount that he is on-ice for but either way, Semin ranked very high in all of these categories and that obviously factored into their decision to sign him.

There are some issues that arise with these stats and the biggest of which being that the “high pressure situations” could eliminate a good chunk of the game and cause them to look at some small sample sizes. A lot of hockey statisticians prefer to look at data when the score is close or tied, so that is understandable but I’m not sure that I agree with elminating the entire first period from the sample. We’re in sort of a “dead-puck” era with a lot of close games, which makes every goal huge. Throwing out a few goals because they were scored in the first period seems a bit strange, in my opinion.

The fact that they are only looking at data when a certain player was used against “top 2-3″ forwards on the opposing team also limits the scope because there are a lot of times when a star forward is matched up against a checking line, the latter of which probably doesn’t have one of the “top forwards” on it depending on the scorer’s definition. Therein lies another potential issue; How does the scorer decide who the “top 2-3 forwards” are? Seems a little too subjective if the team in question is playing a team with not much star talent. I understand the rationale behind this stat because it is probably another way of looking at “quality of competition” like I do often, but I also think that it could lead to an issue with sample size.

One final thing I would like to know about these stats is if they look at the amount of shots or scoring chances a player produces in these situations. I’m probably beating a dead horse with this by now, but it’s tough to use goals and points as a predictive stat because of how much luck is involved with them. Semin, in particular, had a very unlucky season with a poor shooting percentage and could have created more goals in these “high pressure situations” if he shot closer to his career average. Shots obviously don’t catch your attention as much as goals do, but they are also very important to look at because shots are what a player has the most control over. Seeing if a player can drive possession at a high rate usually means that they are likely to score more in the future than a player who isn’t.

I may have my issues with these stats and how valid they are but in the end, we can both agree on that Semin for one-year at $7 mil. is a terrific signing for Carolina. Semin not only can perform in these “high pressure situations,” he can drive possession at even strength, when the score is tied, and when the game is close and that is going to really help Carolina next season. Plus, he also has scored at a high rate at even strength for most of his career. It may not be the way that I, or many others, are used to doing things, but it is nice to see a team take an analytical approach to rate players and use it to factor into their front office decisions.

Quantcast