What an unusual bowl season it was. Huge upsets were the story of the season, and they played out in historic fashion. How about this for a stat, courtesy of @StatsByLopez: Three of the five biggest bowl upsets in the last 20+ years all happened this year. Check out his graph to see it visualized.
With college football now over, it’s time to evaluate how our bowl season pick’em and betting picks did this year. As usual, we’ll start with our bowl pick’em advice and then move on to bowl betting picks.
Bowl Pick’em Performance
It’s much tougher for us to give a blanket statement about how our pool picks did than it used to be, since we now algorithmically optimize our recommended picks for every user’s specific pool type and for criteria such as pool size, payout structure, and anticipated opponent picking trends.
As a result, more so than in past years, it’s quite feasible for some of our customers to do very well in their bowl pick’em pools, while other users do not so well, all in the same year. Maybe our picks for small game winner confidence point pools kick butt while our picks for non-confidence spread pools are just middle of the pack, or vice versa, etc.
So from now on, a “good year” for our office pool picks is best defined as a year where most of our customers do really well, even if not everybody did. This year, though, that most likely was not the case.
The Good: Conservative Picks
The good news is that for the fourth straight season, our most conservative pick sets for game winner based bowl office pools performed very well compared to national benchmarks. (For the unfamiliar, we generated three types of pick sets this year for every user pool: Respectable Finish, Win Something, and Win Big, based on the amount of risk you were comfortable taking.)
Non-Confidence Picks Finished In The Top 6% of ESPN. Our non-confidence picks that used our Respectable Finish strategy, entered on ESPN the day before bowl season began, outperformed almost 95% of the nation. That’s good enough to contend for a top three finish in the average 60-person pool.
Confidence Picks Finished In The Top 12% of ESPN. Our confidence point based picks that used our most conservative approach beat over 88% of the nation, good enough to contend for a top three finish in the average 25-person pool.
Some portion of our customers likely finished in the money in smaller pools using these pick sets, especially in non-confidence pools, which is good.
Coming in top 12% on ESPN on the confidence point side is not as good as we’ve typically done over the past three years, but still, beating 8 out of every 10 opponents is nothing to sneeze at. These are both clearly “respectable finishes,” so the mission of the conservative pick sets was accomplished.
The Not So Good: Aggressive Picks
The bad news is that our more aggressive pick sets for game winner pools did not fare very well this year. As we explained in the notes accompanying those pick sets, playing pick sets like our “Win Big” strategy is risky. Some years, using them will produce huge returns; most years it won’t. In the long run, using the Win Big strategy is a smart gamble, but season to season, results can fluctuate wildly.
This year, most of the Win Big pick sets for small to midsize pools we tracked hovered around the 60th-70th percentile of the nation on ESPN, a finish that is highly unlikely to win anything for our customers. So why didn’t they perform better?
The answer is primarily because the fate of very aggressive pick sets for office pools largely rides on the results of a select number of key “bets” made on value picks that exploit irrational public opinion. As luck would have it, this bowl season happened to be one where the best value picks of the season, including teams who were favored to win their games, lost more often than expected. That’s going to happen sometimes. As they say, that’s why they play the games.
Using the Data Grid (subscription required) section of our Bowl Pick’em tool, here’s a breakdown of the results for our top seven “favorites with value” in bowl pick’em pools this year, according to our win odds and national public picking trends data we collect. As of just before game time, our models — which are strongly influenced by Vegas lines — favored all of these teams to win their bowl games, AND the public was also under-picking them, therefore creating value:
- WIN — Kansas State (69.9% win odds – 38% picked = +31.9% value differential)
- LOSS — Wisconsin (57.8% win, +31.8% value)
- LOSS — Houston (52.6% win, +24.6% value)*
- LOSS — Oklahoma State (50.5% win, +20.5% value)**
- LOSS — Buffalo (51.9% win, +17.9% value)
- LOSS — Stanford (71.0% win, +16.0% value)
- WIN — Florida State (80% win, +14.0% value)
* Houston was technically a small Vegas underdog, but based on other data points, our models liked Houston’s chances slightly better than Vanderbilt
** Missouri was initially the favorite to beat Oklahoma State, but by game time the betting odds had moved to favor Oklahoma State to win
As you can tell from the win odds listed above, four of the top value favorites (Wisconsin, Houston, Oklahoma State, and Buffalo) were close to being toss-up picks. So we certainly wouldn’t expect all of them to win. We also wouldn’t expect all four of them to lose. Lo and behold, they all lost. There was only about a 1 in 20 chance of that happening.
Overall, based on win odds, if you had picked all seven of the “value favorites” above to win, you would have expected about 4.3 wins, but you only got two wins this year.
An outcome like that really hurts in an aggressive confidence point strategy, since our pick optimization algorithms tend to give teams like Stanford (high win odds for a bowl, plus value) and Wisconsin (only a modest favorite, but with great value) much higher confidence points than the majority of your opponents will give them. Those were “high-leverage” games, in other words, and both of these favorites were upset.
Even for picks that weren’t value picks, the big upsets that happened this year also had a negative impact, although it was slighter. For instance, Baylor, Alabama, and Arizona State were clearly the three biggest favorites of bowl season. Despite the obvious risk that those teams wouldn’t be highly motivated for their bowl matchups this year, each team was still favored to win by more than two touchdowns in Vegas, and plenty of real money was being bet on them at those lines. Without the motivation risk, these three teams may have been even bigger favorites than they ended up being.
Even so, the public didn’t unanimously have teams like Baylor, Alabama, and ASU ranked at or near the very top of the confidence point scale. That typically creates a good opportunity to pick up some high-probability points on your opponents, but not so this year — not when three of the biggest upsets in the past 20 years of bowl history all happen during the same bowl season. That’s called a fluke, and it could easily be another couple decades before something similar happens again.
Obviously we all prefer to win, but hopefully nothing here is shocking. As far as the aggressive strategies go, results like this year’s are the very real downside of using a high risk, high reward strategy. You’re going for broke (relatively) to not only just finish in the money, but to maximize your expected profit. If key games don’t break the right way, you’re going to suffer. In our pick set notes, we stressed the differences between the Respectable Finish and Win Big approaches to help all our customers understand that dynamic.
Picks For Spread Based Bowl Pools
On the point spread pools side, it’s difficult for us to benchmark how we did because bowl pool spreads tend to differ from site to site. We entered some of our portfolio strategies in Fox Sports’ spread based pick’em, and using Fox’s spreads the results this year ranged from the 50th to the 70th percentile. Not horrible, but not spectacular. Our Model Driven pick set, which is entirely based on our algorithmic point spread picks and ignores public picking trends, beat about 60% of the nation on Fox Sports.
Winning a pool that consists of picking 35 games against the spread is always going to require a good bit of luck. Even doing very well by sports betting standards (e.g. hitting 55% of your spread picks) may not even put you near the very top of a spread based bowl pool. With just 35 games to pick, all it takes is one random streak of good luck for an opponent to do much better than 55% for one given bowl season, even if he or she is only a 50% long term spread picker.
Our model-driven pick sets didn’t catch fire this bowl season, but we look forward to hearing how customers using our custom portfolio picks for various spread based pools ended up.
To conclude, this clearly wasn’t our best bowl season from a pick’em standpoint. Given the poor performance of the most aggressive pick sets, even for smaller pools, we’d say it was actually the worst we’ve done overall in four years making bowl pool picks. At the same time, it’s hard to be too frustrated about it. Even in a fluky year, the Respectable Finish pick sets finished, well, respectably. The riskier pick sets had an off year, but that came on the heels of several good performances in recent years, and we expect highs and lows over time.
Bowl Betting Picks Performance
By and large our bowl betting picks performed quite similarly to their long term performance record, which can be summarized as follows:
Higher confidence over/under picks have done well. Playable-rated (2 and 3 star) bowl over/under picks have won more than 55% of the time since we started making bowl betting picks in 2006-7.
Higher confidence spread picks have not been profitable, but interestingly, lower confidence picks have done very well. Playable-rated bowl point spread picks have been losers overall (assuming typical -110 juice), but ALL point spread picks (including 1 star rated picks) have been profitable. 1-star bowl spread picks have hit at almost 56% since 2006-7.
This year, our playable over/under picks for the bowl games, measured by our closing picks for each game, went 9-6-0 (60%), improving on our already strong long term track record for bowl totals. Bowl season also capped our best year ever for college football totals picks, with playable picks going 152-130-2 (53.9%) for +8.1 units of profit on the season.
Our playable bowl spread picks did poorly this year, going 6-11-0 (35%). Our 1-star picks went 12-6-0 (67%), for an overall bowl ATS record of 18-17-0 (51%). That’s still not profitable at typical -110 payout odds, though. We’re going to do more investigation in the offseason to see if we can find an actual root cause for the 1-star bowl picks consistently doing better than the 2- and 3-star picks.
Finally, our money line value picks for the bowl games delivered profits if you were flat-betting each pick, with playable picks going 3-4 for +0.8 units of profit. All money line picks went 10-16 for +5.9 units of profit, thanks to some nice underdog picks hitting.
With the college football season over, we have our sights set on several future projects to do before next season kicks off. We made great strides with our newly-launched office pool tools this season, both for season-long college football pools and for college bowl pools, but we still have a laundry list of ideas for improvement. Adding support for combination NFL-plus-college football office pools and different types of pool formats are high up the list.
At the most basic level, we also have several investigations lined up to try to improve our algorithmic predictions for college football, and bowl games as well. Historically, college football hasn’t been one of our best sports in terms of betting picks, and we’ve struggled on the point spread front in recent years.
However, our playable-rated over/under picks have now strung together two straight seasons of profits, and we’ve had some interesting conversations with TR customers who have done very well applying our raw predictions for college football in their own homegrown betting systems. We have plenty of new “angles” to look into as a result. Stay tuned for next year!