|
In a recent "Hey Bill" Bill James wrote this about the 1986 season of Dave Parker: In 1986 Dave Parker drove in 116 runs for the Cincinnati Reds. The NL MVP voters named him the fifth most valuable player in the league. I credit him with 20 Win Shares, which would tie him for 21st in the league.
WAR credits him with 0.2 WAR.
This is not a small difference. Sure, there are many more cases on which we essentially agree, but we are the people who draw the maps on which other people rely. We cannot have discrepancies on that scale. This is where those kind of issues are worked out, worked through—here, and MLB.com, and a few other places. We are many years of hard work away from having it all figured out. If it irritates you, well. . . go away. We have work to do. I would have responded on Bill's site, but can't do that for two reasons:
So I'll post it here and hope Bill has enough interest to read it. You recently brought up Dave Parker's 1986 season. WAR has him at 0.3, while Win Shares gives him 20. Of course those numbers are not on the same scale, but for those not familiar, I would suggest that a 20 WS season for a full-time outfielder would usually be equivalent to around 3-4 WAR. Part of the difference is in defense, as BBref shows Parker as 17 runs below average. I don't think that's obviously wrong, as he was 35 years old, fat, slow, and 2 years later would become a full-time DH. But defensive Win Shares will not show such an extreme number, and probably has him at the equivalent of -5 runs. If BBref used a similar system for fielding Parker would probably be around 1.5 WAR. A bigger difference in Parker's value comes from how you look at his offense. Parker hit for power, but his OBP was only a little bit better than league average. BBref shows him as 8 runs better than average. This is based on a context-neutral interpretation of his stats. Win Shares gives extra credit for situational hitting, and Parker excelled at that. With runners in scoring position, he hit .349. 22 of his 31 homers came with men on base, even though he had more at bats with bases empty. One thing BBref could do to capture this is use the stat RE24 (run expectancy for each base-out situation) instead of context neutral linear weights. Parker's RE24 that year was +35, so if that was used he would be a 3 WAR player. BBref has made the decision not to use situation-specific numbers in WAR. From my experience when this debate is opened to a group of 25 sabermetricians you'll get 25 or more opinions on which way to go. You could go even further, and look at win probability added. As Win Shares works, Parker would get extra credit for hitting a 3 run homer (as opposed to a solo shot) but it wouldn't matter if it came in the bottom of the 9th, down by 2, or top of the 5th, up by 14. (Technically, Parker would get a little bit more credit for the bottom 9th, 2 out homer as team runs are reconciled to actual wins, but the credit would not go to him alone, it would be shared by the whole team.) It would be cool if someone with more programming skills than I have was able to create a WAR-interface that allowed the user to decide whether they want context neutral, run expectancy, or win probability considered and generate the batting wins based on that decision.
This page was last modified 01/07/2023 |