Saturday, December 28, 2019
Bias in xwOBA: players or plays?
?As some of you may know, there's a constant tension in trying to decide if we care about players or plays.
This is easiest described with FIP: the reason FIP takes an agnostic position on batted balls in play (BIP, i.e., any contacted ball that is not an over the fence HR) is that those plays don't tell us much about the pitcher. If a pitcher happens to give up 10 or 12 singles in a start with 0 HR, and then he gives up 4 singles with 1 HR in the next start, it's that second start that tells us more about the pitcher (and in a negative sense). The first start may end up with worse results, but it's the second start that ends up with worse evaluation of the pitcher.
Now, most people think we do this because BIP involved the fielders. But this is not accurate. We do this because BIP involves a heckavu amount of Random Variation. We may call it FIP (fielding independent pitching), but it's more accurate to say LRVIP (limited random variation in pitching) or if you need something catchier and less accurate, LIP (luck independent pitching).
With Statcast, we can better assign values to PLAYS, but that's not really going to advance our understanding of PLAYERS. In this excellent article, the researcher notes:
The general takeaway from the research on whether players should hit with backspin is that backspin balls outperform expectations but the players that hit backspin balls more often actually underperform.
Last year, I noted this issue with Lorenzo Cain v Joey Gallo (Cain on the left, a line drive hitter, and Gallo on the right, who has more of an uppercut attack):
Given the same speed+angle, Cain gets far more distance in the line drive angles than Gallo. That's because at those angles, Cain is more likely to hit it just under, and Gallo is more likely to top the ball.
We can also see it in a player changing his approach, like Mookie Betts. In 2016-17, Betts had more of a Cain approach, and so, he'd get tons more distance on low launch angles. But in 2018, his MVP-level year, he was getting much higher launch angles. And as a result, the distance on the same speed+angle was much lower than in 16-17. Did he get worse? No, he changed his approach.
Do we care about the player or the play? Well, ultimately, we care about the player. And to get there, we need to know about the play. But the play is an intermediate step. If we make it the last step, then we don't learn enough about the player.
And so, if we want to evaluate the speed+angle of a ball, we ALSO need to know about the tendency of the batter. It gets tricky, which D.K. illustrates quite nicely. Sometimes the bias that exists is a good thing. We don't want to always remove the bias, because removing it at the play level may increase it at the player level.
This is why FIP endures, because it takes a careful focused view of the plays.
Recent comments
Older comments
Page 1 of 150 pages 1 2 3 > Last ›Complete Archive – By Category
Complete Archive – By Date
FORUM TOPICS
Jul 12 15:22 MarcelsApr 16 14:31 Pitch Count Estimators
Mar 12 16:30 Appendix to THE BOOK - THE GORY DETAILS
Jan 29 09:41 NFL Overtime Idea
Jan 22 14:48 Weighting Years for NFL Player Projections
Jan 21 09:18 positional runs in pythagenpat
Oct 20 15:57 DRS: FG vs. BB-Ref
Apr 12 09:43 What if baseball was like survivor? You are eliminated ...
Nov 24 09:57 Win Attribution to offense, pitching, and fielding at the game level (prototype method)
Jul 13 10:20 How to watch great past games without spoilers