The Way We Think About Basketball "Upsets" Might Be Wrong
Before UMBC's historic win over Virginia, 16 seeds were 0-135 in the NCAA Tournament. Prior to the 2016 NCAA Tournament, FiveThirtyEight.com featured an article that made the claim that 16 seeds had actually been very unlucky not to get a win (0-124 at that time), and that they would have been expected to win about 4 times according to FiveThirtyEight's model. I believe this overestimates the chances of a 16 seed beating a 1, probably by a significant margin, and I'll tell you why...but first, bear with me as I make a detour into a story about poker.
In a hand of poker between 2 players, each hand has a specific, calculated chance of winning. The chance of winning is based strictly on the probability that the 2 starting cards will eventually combine with some of the (up to) 5 community cards to make the best 5 card hand. You can find online calculators that will give you these odds for any variations of starting hands; one is located here at CardPlayer.com. Using that tool, I can calculate my chance of winning if I start with the jack of spades and jack of diamonds, and my opponent has the six of hearts and six of clubs; it's almost exactly 80-20 in my favor. But if I sat down in a hypothetical poker game against a top professional player, and we were dealt these exact hands, it's very likely that I would win MUCH less often than 80%.
The reason why is because we're not just going to deal out the 5 community cards and see who wins; poker is a game of multiple betting rounds as more cards are revealed. The best players constantly make their opponents make decisions, and often intimidate their opponent into folding. Against a top player, I’ll probably often fold a hand that would have eventually won, because the opponent can cause me to lose confidence by betting aggressively. Even though the computer simulation of the hands I described earlier gave me an 80% chance to win, in reality I'm going to win less often by frequently making decisions to fold. When my pair of jacks doesn’t improve after community cards start being dealt, I may be scared off by the wide range of possible hands that can beat me. I’ll likely only have the confidence to keep calling bets if my pair of jacks improves to three of a kind, or a full house, or some other very strong hand. But that won’t happen anywhere near 80% of the time, so my winning percentage will be much lower than the strength of my starting hand would imply.
I believe an NCAA tournament game between a 1 seed and a 16 seed plays out much the same. Computer models can come up with rankings of the teams which give the 16 seed a small chance of winning; say, 3%. That 3% chance of winning reflects the relative strength of the two teams, just like the chance of winning a hand of poker reflects the strength of the cards. What it ignores, however, is that a 16 seed will face many points in the game where a lack of confidence will keep them from playing anywhere near their best. When trailing against a lesser opponent, a 16 seed might maintain some confidence and do the little things to stay in the game and pull out a win. Against a 1 seed, a 16 seed may believe they are fated to lose, and not expend what they might see as wasted effort to try to come back.
In the poker example, I needed to make a very strong hand to have the confidence to win. Similarly, I believe the scenarios where a 16 seed could beat a 1 seed involve the 16 seed being in a situation where they would be highly confident...such as leading by double digits with a few minutes left, or a 1 seed losing several key players to fouls or injuries. The UMBC game played out this way, as they took a double-digit lead early in the second half which they never relinquished. When they had a 14 point lead with 5 minutes left, the Retrievers had to have a huge boost of confidence which helped them keep playing near their best.
It's incredibly unlikely for a 16 seed to have any kind of lead against a 1 seed late in the game. Using some assumptions and math, I’d estimate that a 16 seed has less than a 0.5% chance of having a lead of 5 points or more against a 1 seed with 5 minutes left in the game. If a 16 seed doesn't have a lead like that, they are much less likely to have the confidence that they can pull out the victory. I don't think it's a coincidence that 1 vs 16 matchups have yet to produce a situation where a 1 seed hit a game-winning shot. In the rare occasions before last Friday where a 16 seed came close to beating a 1 seed, the 16 seed always needed to it a shot in the final seconds to win or tie. The 1 seeds had always been able to take a narrow lead late, even when they weren’t playing well.
With all that being said, I think UMBC's win over Virginia can represent a real tipping point. Every 16 seed now will tip off know that it is indeed possible for them to win; they've seen it happen. I really believe there had been somewhat of a mental block for 16 seeds that just didn't exist with 15 seeds or 14 seeds, because they had seen other similarly seeded teams win. 15 seeds are 8-128 against 2 seeds; while that's a very poor record, it's miles better than 1-135. Since Richmond became the first 15 seed to win a game in the NCAA tournament in 1990, the stretch between 2002-11 is the only period of more than 4 years without a 15 seed winning. Even during that stretch, a 15 seed lost by 3 or fewer points 5 times; Tennessee (2006) and Duke (2008) were 2 seeds who needed game-winning shots in the final seconds to escape the 15-over-2 upset. The point being, after the first 15 seed won, it’s been a pretty regular thing for 15 seeds to win (or come incredibly close).
The lasting impact of UMBC's win may very well be the confidence it inspires in other 16 seeds. In 15-20 years, we may look back at the Retrievers in the same way that poker players look back at Chris Moneymaker, whose victory in the 2003 World Series of Poker convinced a horde of amateur players to take a shot at beating the best poker pros. Every 16 seed from now until eternity should live by the words of Lloyd Christmas (played by Jim Carrye) from "Dumb and Dumber": "So you're telling me there's a chance???"
Thanks for reading my newsletter. If you have any questions, want to argue a point, or have some feedback, feel free to reach out via email at sean@hoopsinsight.com, or on Twitter @hoopsinsights. If you liked this, let me know as well, and tell your friends to subscribe at www.hoopsinsight.com.
You can forward this to others, but please ask them to subscribe as well so I can keep track of who's enjoying my insight and analysis