I have been thinking a lot lately about what I'm coming to think of as index stats. Index stats are, in my ideal Sabermetrics meets volleyball world, what I call a comparison to a standard or the team as a whole. In its current iteration, index stats compare an average, like an average passing score, to the team average for that stat. Alternately, if could compare to an ideal, or par value for that stat.
I started looking at stats this way while looking at NBA box scores and game recaps. There was a player who was scoring a lot of points, but he was doing so by taking a lot of shots, a lot of shots. Some might use a points per attempt stat in addition to field goal percentage to assess scoring efficiency. I wondered how he stacked up against the team field goal percentage, and by extension whether or not someone else taking more shots might be better for the team. I came up with a shooting index start (individual's field goal percentage/team field goal percentage). This stat is going to be a number around 1 if the player's stat is roughly equivalent to the team stat. If it is higher than 1, the player performed better than the team in that stat.
Leaving if there is fine. With that kind of index stat, I can make all kinds of armchair quarterback, and watercooler coach statements and assessments. I can say player A on team B is better than player X on team Y because he had a better shooting index. Because apples aren't oranges, interteam comparisons are of less value than intrateam comparisons. I could instead say player A should shoot more than player B on team C because he has a better shooting index, and it could have a little more basis in reality. It is still not the whole picture, and is reducing things down to the point of absurdity, but it could have some utility. Even more useful might be comparing the individual's performance instead to a goal value. The coaches and team might have a goal of having a perfect pass to the setter 65% of the time. That value replaces the team's average, and the individual's index stat in a sense is now a measure of the player's performance quality. It can even be used with the team average divided by the goal value.
The practical value for coaches is in evaluating who might be the best choice for the last shot, and also in determining what skills and situations need to be worked on and simulated in practice. Today I started contemplating further utility in this kind of stat. Comparing the player to himself. With enough data, you could create situation indexes. You could have a fifth set player hitting efficiency index that illustrates whether a player does better or worse in critical situations. Giving them more reps in a drill that replicates a tense end of match setting should help the player do better in the future, and hopefully translate to more success in competition. Keeping that simple, common feature (greater than 1 on an index stat is good) makes gleaning meaning easier.
Something I've wrestled with in trying to make this work with volleyball is the quirky nature of the hitting efficiency stat. Superficially it is similar to the hitting percentage in baseball. In reality it is a goofy stat that could range from -1.000 to 1.000. If either the player or team efficiency is negative, it gives a negative index value, and could kill any utility it might otherwise have. The best I have so far is to add 1 to the efficiency stat giving it a range of 0 to 2.000, and then divide by 2. This puts the hitting efficiency range at 0-1.000 with positive efficiency stats being anything above .500. Another weird one is going to be any state where lower is better. Take a hitters error percentage. This would be a great place to subtract the average from 1, like In Play Efficiency. This will maintain the higher is better nature of index stats.
It's an interesting concept. It could be a pretty big deal. Index stats could be a simple way to get a whole lot more meat from the stats skeleton. They could be put into play fairly easily with some basic spreadsheet skill.
No comments:
Post a Comment