Friday, January 20, 2006

prOPS Introduction

Earlier last year, JC Bradbury at the Hardball Times, introduced a new metric called prOPS (predicted OPS) in this article:

Introducing PrOPS

What prOPS tries to do is cut through the "luck" factor of BIP outcomes (out, single, double, etc...). Over the course of a year, "loud outs", texas leaguers and bloop singles can add up in one direction or the other, skewing a player's stats. Bradbury negates this luck by ignoring the outcomes of a BIP (ball in play) all-together. Instead, he studies the underlying statistics that generate a ballplayer's OPS.

He found 6 statistics that seamed to drive OPS the most; three defensive independant stats (K, HR, BB) and three BIP types (FB,GB,LD). He then weighed the components, in terms of their relative effect on OPS. For example, line drives result in an out only 26% of the time and are most often extra base hits. As a result, LD% has the greatest weight in the prOPS model. Weighing all 6 stats and adjusting for park factors, his prOPS metric was complete.

How does he attempt to validate prOPS? In the Hardball Times Annual 2006, he does so in two ways. First, since we attribute much of the variance in a player's year-to-year stats to luck (statistical noise), prOPS should vary less than OPS over two seasons since it cuts much of that noise out. He showed that that's indeed the case. Second, he showed that prOPS better predicts a player's future OPS than his previous year's OPS does. To illustrate this further, he constructed a table containing the 25 highest overperformers and underperformers from 2002 to 2004. He found that 20 of the 25 top overperformers regressed the following yearand 21 of the top 25 underperformers increased the following year. After noting a mixture of good and bad players on each list, I was convinced that prOPS was at the very least onto something.

Enough introduction, how did Brewer players do, relative to their prOPS?

2005 prOPS, Milwaukee Brewers
            PA  PrAVE   AVE  PrOBP   OBP  PrSLG   SLG   PrOPS   OPS    Diff

C. Hart 63 0.291 0.193 0.357 0.270 0.457 0.368 0.814 0.638 -0.176
R. Weeks 406 0.259 0.239 0.350 0.333 0.432 0.394 0.782 0.727 -0.055
R. Branyan 238 0.268 0.257 0.380 0.378 0.527 0.490 0.907 0.868 -0.039
JJ Hardy 423 0.268 0.247 0.343 0.327 0.405 0.384 0.748 0.711 -0.037
C. Lee 680 0.279 0.265 0.339 0.324 0.502 0.487 0.841 0.811 -0.030
L. Overbay 614 0.290 0.276 0.379 0.367 0.460 0.449 0.839 0.816 -0.023
D. Miller 427 0.282 0.273 0.350 0.340 0.420 0.413 0.770 0.753 -0.017
P. Fielder 61 0.288 0.288 0.310 0.306 0.468 0.458 0.778 0.764 -0.014
G. Jenkins 610 0.293 0.292 0.380 0.375 0.508 0.513 0.888 0.888 0.000
B. Clark 666 0.292 0.306 0.359 0.372 0.430 0.426 0.789 0.798 0.009
J. Cirillo 219 0.262 0.281 0.350 0.373 0.386 0.427 0.736 0.800 0.064
B. Hall 538 0.266 0.291 0.317 0.342 0.435 0.495 0.752 0.837 0.085
W. Helms 187 0.250 0.298 0.317 0.356 0.384 0.458 0.701 0.814 0.113

I broke the table into three groups: unlucky, neutral and lucky. The "unlucky" list includes two notable rookies, Weeks and Hardy. Weeks under performed by 57 OPS points (25th largest underperformance of 2005), while J.J. Hardy underperformed by 37 points. These numbers suggest that if Weeks and Hardy had identical seasons in 2006, their raw stats would increase simply as a result of luck. If there's any merit to that, 2006 could prove to be a breakout year for both.

According to prOPS, Miller, Jenkins and Clark all earned that stats they received in 2005. It's comforting to see that Jenkins' excellant 2005 season wasn't just a result of balls bouncing his way. Leading the league in line drive percent certainly didn't hurt (26.7% of Jenkins' BIPs were line drives). Of course, whether he can keep that LD% up is another topic altogether.

Who were the most "lucky" Brewers in 2005? Seeing Cirillo and Helms in the overperformer category makes me happy to know that the former will be only be used off the bench and the latter isn't even on the team anymore. I skepticle of both's stats last year and these numbers don't do anything to change that. That leaves us with savior/whipping boy, Bill Hall. prOPS suggests that Hall's gains in 2005 were much a result good luck. Could that be right? Was it just a fluke season for Hall?

Hall's 85 point over performance was the 4th highest over performance in the majors last year. It also ranks as the 13th highest over performance between 2002 to 2005 (the length of this study). While I expect Hall to continue to improve his batting skills next year, I am very skepticle that he can carry a .800+ OPS next year. His 2005 numbers just don't add up. Perhaps Melvin knew what he was doing when he aquired Koskie?

Who was the top overperformer from 2002 to 2005? None other than Scott Podsendik, during his excellant rookie campaign of 2003. As you probably know, Pods hasn't even come close to matching that season, in terms of OPS. This highlights a glaring problem of prOPS; it doesn't account for the speed of the batter. For example, prOPS simply assumes that a league average number of ground balls will result in an out (75%). For a player like Pods, who converts more grounders into base hits than the average player, groundball should have a higher weight than, for instance, someone like Chad Moeller. Common sense tells us this is wrong.

While the author makes mention of including speed in one of his online articles on prOPS, it wasn't even mentioned in the HBT Annual. As far as I can see, it's still a glaring issue and one that needs to be addressed before prOPS can be taken too seriously. Still, I feel prOPS can give you a unique and useful insight into the role the "luck of the bounce" can have on OPS over the course of a season. In Hall's case, let's hope it was wrong.

Here's a link to the prOPS for all major leaguers. Next time I'll look at Bradbury's Brewer projections for 2006, which were based off his prOPS work.


Post a Comment

Links to this post:

Create a Link

<< Home