Baseball Graphs is dedicated to the better use and communication of baseball statistics. Below, you'll find excerpts from, and links to, some of the best baseball writers on the Internet. Follow the links above to read my own intermittent attempts at wisdom (the Baseball Graphs blog), and the heart of this site, historical graphs of every season dating back to 1900.
There are also two special sections you might want to check out. One is the graphical review of the 2003 season, which informed our work at The Hardball Times. The other is the Batted Balls Library, which includes a unique look at batters and pitchers from 2002 through 2005.
The Crawfish Boxes made an image of my WPA spreadsheet, which they call the “Win Expectancy Contraption.” Here’s what it looks like to them:
I think that’s pretty awesome, though WPA has nothing to do with OPS, RC, WARP, JAWS, EqA or any of the other esoteric sabermetric stats cited.
I found the WPA Contraption via a link in this post, which argues that a runner on third with two out vs. a runner on second with two out (in the ninth inning) is worth more than WPA allows. WPA says that there is a difference of 2% win probability between the two situations, and the author thinks that is low (though he doesn’t say what he thinks the difference should be). Actually, he thinks it’s “bullshit.”
I’m torn in these types of situations. On the one hand, it always kind of bugs me when people refuse to be open to statistical frameworks like WPA. The poster basically feels WPA is wrong in this case because it doesn’t jibe with his intuition. On the other hand, WPA isn’t perfect. It’s a model of reality. It could be off in specific situations.
So I looked at Chris Shea’s Win Expectancy Finder to see what the difference has actually been in real games. Of course, there’s a wide variance in the three time periods (1979-1990, 1991-1998 and 2000-2004). In fact, WPA was lower with a man on third than a man on second from 2000-2004. But, on average, the difference has been about 4%, or twice as high as my spreadsheet show.
On the other hand, there has only been a difference of 2% this year, according to Baseball Prospectus. And there are significant sample size problems with the “real life” data. For instance, the WE Finder found that, in the seventh inning, a man on second was actually worth 1% more than a man on third from 1979-2004. Just imagine how Crawfish Boxes would react with that data!
So the WPA contraption might be a bit off, or it might not be. I don’t know what Crawfish Boxes would consider the right number, but I think we can safely say it’s between 2% and 4%. That’s assuming Crawfish Boxes won’t object to actual outcomes…