from Amazon
Thursday, August 28, 2008
Clay Davenport: replacement-level is…. .350 win%, not .150 win%?
Great job by Clay in going through a list of the best minor-leaguers over 30. He concludes that the best nonpitchers would score 3.7 runs per game, and the pitchers would allow 5.0 runs per game (and when you include the slightly below average fielding, all the way up to 5.2 runs per game). That works out to a .350 win%.
The best team of players who have no hope of playing in MLB would win 35% of the time. Most analysts have argued that the number is somewhere around the 27% to 35% level. Clay’s number here is a bit on the high-end, but certainly believable. I would not be surprised that the players have a selection bias and probably not enough regression, and so, we can see that perhaps they would really score 3.5 runs per game and allow 5.5 runs per game. That would imply a .300 win%.
Clay’s WARP however presumes a 25-win level for a replacement level team, or a .154 win%. I hope that Clay may have convinced himself that WARP needs to rethink its position about replacement level, and join the rest of us. A guy who puts in as much effort and thought in doing the work he does to distribute the work to as many people as he does deserves our respect. There is a sizable community that takes this WARP stuff seriously, and hopefully Clay can recognize that.
UPDATE: For those who want more of my takedown of WARP, you can read about it here.
UPDATE2: And here.
Clay’s numbers are probably a shade high because these are the “Replacement Player All-Stars,” if you will. But it’s really hard to swallow the idea that these guys are twice as good as your typical replacement level player, and I hope Clay figures that out.