November 29, 2005
Sizing Up the WinnersLast time out we looked at the HACKING MASS results, so this time we'll take a look at the other Baseball Prospectus fantasy contest, Predictatron. Our own Jim Baker was kind enough to share this contest idea with the tech crew, and the first annual contest has gotten rave reviews from staff and readers alike.
Previously, we looked at the distribution of picks and got into considerable statistical jargon. This time, we'll keep it lighter and look at how the contest actually transpired. In case you're not familiar with Predictatron, check out the rules. If you had your own entry, you can view the details of your ballot here.
Here is an example of the detailed ballots, this one for our winner, Greg Ferguson:
Greg's score: 1000 - 154 + 12 = 858
As you can tell, Greg did an excellent job predicting team wins this year. The only team he was more than 9 wins off for was the White Sox (as we'll see, that was fairly common this year). Greg notably had both of his mortal lock picks, Kansas City and Tampa Bay, pegged exactly right. He did make up a few points in the playoffs, getting 6 of the 8 playoff teams correct. Those 12 points were enough to propel him to victory, holding off the second place finisher, who earned 18 points for correct playoff predictions.
In talking to Greg, he revealed that he didn't have any secret formula. He entered originally as a way to procrastinate at work, and forgot about the competition for much of the year. Checking back into it, he found himself in first place at the start of the playoffs. As we noted before, his playoff bracket was good for 12 points, getting 6 of the first round teams correct. He's planning on using his winnings toward travel out to California's East Bay to visit his parents over the holidays. Happy travels, Greg.
As a point of comparison, here is a list of how far off the median win guesses were (MGD stands for Median Guess Difference):
Team MGD ANA 5 ARI 5 ATL 2 BAL 6 BOS 2 CHA 20 CHN 9 CIN 2 CLE 8 COL 2 DET 5 FLO 3 HOU 9 KCA 9 LAN 17 MIL 6 MIN 7 NYA 0 NYN 1 OAK 1 PHI 0 PIT 3 SDN 5 SEA 8 SFN 7 SLN 7 TBA 2 TEX 1 TOR 7 WAS 13
This median ballot would score 828 points before we subtract for whichever two teams were selected as Mortal Locks, and without getting any points back for playoff picks. Interestingly, a score of 828 would place about 135th, which is about 88th percentile for the 1095 ballots.
It is no surprise that the biggest misses for the ballots as a whole were the White Sox, Dodgers and Nationals. What is fascinating, though, is that the highest guess for the White Sox total wins was 95, which was 4 wins lower than their actual total, 99 wins. If we treat the team's win total like a binomial distribution, where p is the team winning percentage, then the highest win guess for the White Sox was actually half a standard deviation lower than the mean. One ballot had the White Sox at 64 wins, over 5.5 standard deviations below the mean.
Miraculously, one user did accurately predict this year's World Series winner. Congrats to Beau Underwood, who has bragging rights over friends for decades to come.
Regardless of what you think contributed to the Dodgers' poor season, Predictatron balloters didn't see it coming. Only 2 people guessed a win total lower than the Dodgers' actual total of 71 wins (99.8% of ballots missed this). For the Nationals, just 7 ballots were at 81 wins or higher (and 99.36% of ballots were lower).
To finish up, we'll look at the standings of Baseball Prospectus staff predictions:
Ironically, in the first annual competition, Predictatron's creator, Jim Baker, came in last among staffers. Thanks to all to for making the first iteration such a success, and we'll look forward to playing next year.