One idea that is persistent in all sports, including hockey, is that there is a certain desirability to players that are consistent. We like the guys that show up to play every night, even if we need to sacrifice some upside for boredom. When it comes to the Leafs, this concept is even more pervasive, as the team largely appears inconsistent at times, at least according to our eyes. Certain players have been called out by fans at different times for disappearing for a few games, while others have a reputation for being slow and steady in their production. This post will work to quantify this ‘consistency’ to help us identify what we can think of as more reliable players.
The metric I’ll be using for our analysis is called GameScore, which was originally built for hockey purposes by Dom Luszczyszyn. Game Score aims to give an approximate measure of a player’s productivity in a single game by creating a composite of different metrics including goals, expected goals, primary assists, secondary assists, shots on goal, blocked shots, penalty differential, faceoffs, 5-on-5 corsi differential, 5-on-5 goal differential. GameScore has never claimed to be a perfect metrics, but I like that it includes both traditional counting statistics, along with the fancy stuff like xG and Corsi.
The piece of set up relates to our definition of consistency. A basic measure of the variation in our data would be the standard deviation, best thought of as how spread out data is relative to the mean. The last bit is important because if we were to use standard deviation to compare players, Auston Matthews could appear more inconsistent that Cody Ceci, because we’re comparing to a higher average. So our last transformation is to take our standard deviation and divide by their average game score – in fancy terminology, this is called the coefficient of variation. This gives us the risk per unit of return, and specifically in our case, the risk per unit of GameScore, meaning we ideally want players with lower ‘Consistency’, since this denotes a less risky performance.
Without further discussion, here are the Leafs ‘Consistency’ rankings:
We should be pleased to see the core four at the top of the list, confirming they aren’t being overpenalized for their higher average GameScore. At the top, Auston Matthews is the most consistent Leaf according to our metric, followed by Marner, Nylander, and Tavares. Among the depth pieces we see better(lower) consistency from Spezza, Hyman, and Mikheyev than forwards like Kerfoot and Kapanen. Defenseman appear slightly more unpredictable – Muzzin was the most consistent, follwed by Holl and Morgan Rielly.
Another way to look at consistency is graphically, by eyeballing players with wider variations in their distribution:
This supports some of our assertions above, and can help us pick out a few extreme results. Tyson Barrie had a particular rough October with 3 of his worst 4 results coming in that stretch (-4.26, -3.62, -2.59, -2.52). The best GameScore all year for the leafs was a 5.87 by Morgan Rielly against the Detroit Red Wings on November 27th – that was a doozy!