Stats Don't Tell Story... Yet
June 19, 2007
by Michael J. Marcinkus
As you may recall, back in April, I challenged the idea that statistical projections based on historical numbers are not much better (if at all) than a simple "guesstimate" based on knowledge of a player. In testing this, I used 25 random hitters and 25 random pitchers, posted the CBS Sportsline projections, then posted my projections, which were simply obtained by eyeballing some numbers and frequenting an educated guess at this year's totals. Imagine staring at the back of a baseball card and trying to predict what Geoff Blum will do this year and you'll understand my thought process.
I just updated the hitters statistics as of June 18th (attached is below) and projected them out to a full season. It's still too early to get a good read on the pitchers, since wins, saves, etc. can fluctuate greatly over the course of the second half of the season.
Focusing on the hitters, I can come to one basic conclusion: I am definitely not better at predicting future results than Sportsline and others which use statistical models. Despite my humble admittance of defeat, I still can't claim that Sportsline is better than me, however. Looking at the numbers unscientifically, I estimate that Sportsline has projected 10 of the 25 players better than me so far, I have projected 8 players better than Sportsline, and we are about even on the last 7 players. That's hardly a ringing endorsement for statistical models, but I certainly can't claim victory either. The numbers should be more clear cut by season's end, but this gives us a good starting point.
One factor that must be noted is the inclusion of bad players or has-beens. Steve Finley was released and Jody Gerut doesn't even play, so it's hard to give anyone an edge on those players. Also, I decided that strikeouts is a useless stat to project - if I thought a player would perform poorly, I tended to increase his strikeout totals, but if he plays above and beyond poor, he will be benched and his strikeout totals will actually decrease. I will remove the strikeout column in future updates.
Hopefully my predictions will fare better come September, but in the meantime, please peruse the uselessness of season long predictions using mid-June stats!
