Bracket busting began here
April 03, 2018 3:00 AM
by Elliot Frome
I admit to pretty much stop checking my NCAA brackets after the opening weekend. When a 16-seed knocked off a 1-seed that I had going to the Final 4, I knew I wasn’t going to do well. This evening, I decided to check. It wasn’t pretty.
My overall potential score was not awful because my pick to win it all was still in the tournament, but one side of the bracket is completely done. When I checked the leaderboard on Yahoo, I was a bit surprised to see the leaders had only about 2/3 of the games correct. I know there have been a lot of upsets, but this number seemed very low.
First, I have to say I don’t really follow much college hoops. I root for UNLV, my adopted hometown school. I root for Albany State – my true alma mater, which has made the tournament a couple of times but never done much in it. And, I root for whoever is playing Duke – long story.
So, I’m not really surprised if my bracket sheet has a lot of wrong picks on them. For the most part, I’m guessing. I have about half the games right so far. So, for guessing, I suppose I’m not doing that bad, right?
If you flipped a coin 60 times (the number of games played so far), and guessed right 30 of them, you’d be doing “average.” I don’t think I need to go into a long math proof of why this is. But picking winners of sports events is not like flipping a coin for a variety of reasons. The most obvious reason is there is generally not a 50/50 chance either team will win.
While the exact probability that either particular team will win is not an exact science, we do know there are favorites and underdogs. A 1 seed will beat a 16 seed, 99% of the time. Until this year, the 1 seed losing had never happened! When you get down to the opening round 8 vs. 9, the 9 seed might be the favorite according to the pros.
So for simplicity sake, let’s say, on average, the favorite is likely to win 60% of the time. To be clear, this is just an over simplification. Two 60% favorites are not the same as a 70% favorite and a 50% favorite. Now, if you were picking 60 games where one team was 60% likely to win, what would “average” be? The math model shows us this will be at about 36. Perhaps, not surprising as this is also 60% of 60.
As stated earlier, to really calculate this we would need to feed in the exact probability of each team winning to get a more accurate model of what is average. I also suspect that in the opening rounds of the tournament, the average is above 60 as many games are considered more like 95%/5%.
But, there is another factor in the tournament that explains why many people probably do not achieve the average (more than one would expect?). First of all, we are not talking about 63 totally independent games (when the tourney is done). If you were one of the people who picked that number 1 seed that lost (pretty much everybody), there is a good chance you immediately lost two, three or four games. You picked that number 1 seed to go at least two rounds!
When you pick in a tournament bracket, you’re picking blind beyond the first round. You don’t know who the teams are but you have to pick a winner. You incorrectly pick an upset in round 1, so you pick what appears to be a favorite in round 2. But when the round 1 upset doesn’t happen, your round 2 pick doesn’t look so good anymore either! So, you are not picking 63 independent games, knowing who the two teams are. Your wrong decision in round 1, could lead to an even more wrong decision in round 3.
These last two factors combine to greatly lower the expected average number of picks one would expect. It is potentially impossible to create a math model that could accurately calculate this value. The number of variables is immense. Perhaps, the best number we could arrive at would be based on the actual results of past years matched up against a math model in terms of likelihood of upsets. There are probably some people out there who try to do this in order to maximize their chance of winning some pools.
Of course, the “big” prizes are the pools that promise huge sums of money if you get a perfect bracket. It is very easy to offer these because the odds are beyond astronomical. Even if you could predict each game with 80% accuracy – an absurd number for a sporting event of this nature, you would still be talking about 1.7 million to 1. If you bring it back to the 60% rate, you would be talking 94.7 trillion to 1.
All to win a million dollars. Some of the companies that are offering these types of prizes probably don’t even bother taking out an insurance policy “just in case.” All it takes is one Loyola Marymount per year to all but ensure they will never worry about awarding that money.