I've been brushing up on my statistics and to give myself practice I've been crunching some of the game logs of each player on the roster. Anyways, first I averaged the GameScores of each of the guys on our team, with the following results: This shouldn't be too surprising. since it roughly corresponds with how prolific each player is. The one big caveat being that GameScore isn't rate based, so it favors players who have higher minutes which is why Derrick ranks higher than guys like Taj and Nikola, who have higher PER's. The secondary bars represent the standard deviation in in GameScore. Here's how their standard deviations compare to one another: Here things are evened out a little bit more, but I'd guess a raw comparison is more advantageous to bit players than larger ones. Ie, the more you produce the easier it is to have large game to game fluctuations. So here's each players standard deviation in per game output relative to what they produce: Here it's inverted.........the bottom of rotation players end up worst. I'd guess the reason why is because their output is more sensitive to outliers. If Tony Snell does next to nothing most games, an occasional decent game will really move the numbers since his output usually hovers around zero. So, finally, it might be useful to measure their median deviation, which would do a better job of capturing their typical day-to-day productivity and here the results are a bit different: These number, I think, make sense. Most of the time Snell really isn't doing that much, and it also confirms the observation that Niko is pretty all over the place on a day-to-day basis. A few closing thoughts: 1). The best comparisons might be between guys at the top of the roster and guys at the bottom, since they'll have different factors influencing how they have to play. In particular, role players can have large swings in their pecking order when guys get injured but top-of-the-roster guys don't. I think this would explain E'Twaun Moore's consistently high marks for inconsistency despite the fact that he's a seemingly steady player. 2). I think when everything's accounted for Taj would have to get the reward as being the most reliable player, followed by Jimmy Butler. 3). Collectively these graphs reinforce the observation that Tony Snell stinks. 4). A good route to follow down in light of this information is what factors affect each player's contribution. Injuries would be obvious, but it'd be interesting to look at opposing team record and maybe position played. I might go forward with more of this, but it'll depend on how much time I have to devote to it.
Thanks, Rosenthal. Interesting stuff. If you find that you want to exercise these muscles some more, might I suggest comparisons with players on other teams. How does Butler compare to Leonard? Wade? Rose v Conley? I've always believed that consistency is what separates the good from the average and the great from the good, but this important attribute is seldom quantified. Anyway, thanks again.
Will definitely do this. My hunch is that comparisons of consistency are more useful when done horizontally (same role, different team) than vertically (up and down the roster) for the reasons I mentioned above. I'd be interested to see as well.