Quote:
Originally Posted by Das Boot
The primary argument is something like: "Steals, as a box score statistic, are undervalued."
Well if he just said then I think people like Fly and myself would have much less of a problem for it. Don't blame people for reading the stuff he actually wrote. He certainly went a lot further than arguing they are simply undervalued.
The supporting case is:
Quote:
1) Steals have outsized predictive weight relative to other box score statistics as measured by average change in team SRS. (His regression is the supporting evidence for this claim.)
From what I understand, he said:
Quote:
I created a regression using each player’s box score stats (points, rebounds, assists, blocks, steals and turnovers) to predict how much teams would suffer when someone couldn’t play
As measured by his difference in SRS (simple rating system, or average margin of victory/defeat adjusted for strength of schedule) with or without him. By comparing the regression coefficients for each variable, we can see the relative predictive value of each (all else being equal). Because we’re particularly interested in how each stat compares with points scored, I’ve set the predictive value of a single marginal point as our unit of measure (that is, the predictive value of one point equals one, and something five times more predictive than a point is five, etc.)
From what I understand from this, a successful steal is more valuable to the team's SRS than a successful point, rebound, assist, or block.
I don't see anything wrong with that, other than that it's pretty useless because box score stats can only tell us so much and it doesn't address the downside of a missed steal.
My objection is this: There are 180-200 possessions in a game. The best stealer is getting 2.5 steals per game and he's not close to 100% responsible for them due to the team aspect of basketball defense. With such small margins there, how can you be confident about the predictive value on steals to SRS, especially given that it disagrees with the market, haralabos, etc;?
Quote:
2) Each steal by itself is a major event (worth ~1.2 points on average). When a team loses expected steals due to player absence, those steals are almost completely not replaced. (The replaceability section is the supporting evidence for this claim.) Therefore we should expect steals per game to be an important metric.
This is where the missed steal attempts becomes so important. It'd be like giving players credit for scoring without factoring in how many possessions they used. Without that, #2 is completely useless.
I don't really get the replaceability thing either. If "Charges Taken" were part of the box score they would be even more irreplaceable than steals but that wouldn't them important because just like steals (and everything else) there is a downside for attempting it.
Thank you for making his argument 100 times better than he made it, but