Now that the 2011-2012 college bowl season has come to a close, we wanted to recap how we did with our picks and advice for college football bowl pick’em contests.
Overall, we did very well relative to the public this year, although not quite as good as last year’s super performance. Still, we’ve received several emails and comments from customers who either won or scored in the money in their bowl pick’em contests, which is always good anecdotal evidence.
Based on final pre-kickoff predictions, our algorithmic game winner picks went 24-11 (68.6%) overall, their third best performance over the past seven bowl seasons.
Straight Up Winners With Confidence Points
Most bowl pick’em pools are based on picking game winners while also ordering your picks with a “confidence rating” from 35 (highest) to 1 (lowest). Get a pick right, and you get the associated confidence number as points, and the person with the most total points once all the games are done is the winner.
Note that I put “confidence rating” in quotes. In fact, determining optimal strategy for most confidence point based pools is a lot more complex than just actually ordering your picks based on the level of certainty you have in them winning. That’s where we come in, with data-driven strategies to help our users game the system, in effect, to increase their overall shot at winning a pool.
This year, for pools with confidence points, our conservative picks finished in the top 5% of ESPN’s contestand the top 8% of Yahoo!’s contest. Given that these picks were designed for 10-20 person pools, that should have been good enough to score a first, second, or third place finish in the average pool of that size, and hopefully take home some hardware.
Our aggressive picks finished in the top 8.5% of ESPN’s contest and the top 15% of Yahoo!’s contest. In a 50-person pool with multiple prize winners, most people would probably have been on the borderline in terms of finishing in the money. In a 100-person pool, using our picks likely gave you a very respectable showing, but didn’t bring home any actual bacon. (Apologies to the vegetarians in our audience.)
Interestingly, we also tracked the performance of an alternative aggressive strategy that we explained in response to a question in our comments by customer Eric Scheel. (Comments at the very bottom.) That strategy ended up doing best of all, landing in the top 2% of ESPN and top 3% of Yahoo! So any users who had decided to draft off of our secondary advice to Eric most likely did extremely well, and had a good shot to place in the money in a 100-person pool, or flat-out win a 50-person pool. Thanks for prompting us, Eric!
Finally, our very aggressive picks finished in the top 7% of ESPN’s contest and the top 12% of Yahoo!’s contest. Typically, you need a combination of getting nearly all of your key bets correct AND having lots of favorites lose in order to win a 500-person or bigger pool, and that didn’t happen this year. But you can still hold your head high if you used these picks. Plus, we nailed three of our four significant upset picks (Utah, Ohio, and West Virginia).
We’re happy with this performance overall. The picks did well, and just one more break going our way with our big bets would have resulted in another truly outstanding year. Vanderbilt, tied 14-14 with Cincinnati going into the fourth quarter, came close, but it’s hard not to grimace at the “woulda, coulda, shoulda” San Diego State game. For those who didn’t follow the game, with just seconds left on the clock, SDSU was winning the game, with Louisiana-Lafayette facing the prospect of a 55-yard field goal attempt to win. If you had asked us at that point how many confidence points we would risk on San Diego State to win, the answer would have been…a lot.
However, that was before the ragin’ cajun voodoo football spirits decided to intervene. SDSU was called for a 5-yard penalty (nooooooo!), thus shortening the field goal attempt to 50 yards. Then, ULL’s kicker stepped up and made the boot of his life to win by two points. Sigh. That’s why they play the games, I suppose. (To add insult to injury, a minute earlier SDSU had failed a 2 point conversion attempt.) Those confidence points could have made the difference between an in-the-money and an out-of-the-money finish for a lot of our customers. It was that close. We absolutely can’t feel bad about it, though; given the chance, we’d take those odds again.
Straight Up Winners Without Confidence Points
Our non-confidence point picks also fared well, although the performance varied much more greatly across pick sets. Strategy for a pool without confidence points is much different, and requires you to take different approaches to maximize your odds to win as your pool size increases.
Here, our conservative picks dominated, finishing in the top 1.5% of ESPN and the top 2% of Yahoo! You were in the driver’s seat to win a 10 to 20 person pool if you used them. Our aggressive picks finished around the top 25% of ESPN and Yahoo!, and our very aggressive picks finished in the top 35% of ESPN’s contest and the top 50% of Yahoo!’s contest. That’s not as good on the aggressive side as our confidence point picks did, but in straight-up winner pools, you end up needing to pick a lot more upsets in order to differentiate your picks. We may have been a little too aggressive, something we’ll re-evaluate next year, but the fact remains that winning a 500-person pool is a boom or bust strategy, and this year was not a boom. Even coming in top 1.5% only puts you in seventh or eigth place.
The final tally? Overall, we made seven sets of bowl pick’em picks this year, and tracked both on ESPN and Yahoo! All fourteen entries finished in the top 50%, 8 of them finished in the top 10%, and three of them finished in the top 2%. In addition, all three of our confidence point based pick sets got 16 of their top 20 picks right, a win rate of 80%, which is much higher than expected given the public’s assessment of team win odds.
We can always look to improve our analysis and tactics, which we will, but given how we performed last year as well (even better than this), it appears we’re doing something right.