March Madness: Beat Buffett!

[stanfr]

I was reading a story today (sorry no link handy but google is your friend) about Warren Buffett/Quicken Loans offer of 1 billion $ to anyone who submits a perfect NCAA bracket for this year's tourney. The story insisted the odds were one in a billion. That seems ridiculous to me, but I'm not sure how one would determine the true odds unless maybe one had a database of human picks.

But it gives me an idea--what if you could get a huge amount of data on all the matchups and do a hard analysis; looking at previous years' tourneys it is exceedingly rare for a lower than 12 seed to win in the first round, and usually those first round upsets are not as 'upsetting' as the rank would indicate. Then, it's even more rare for the rare upset to survive the second round. So, of course the key is to identify the potential upset games, and also to give oneself a better than 50/50 edge for the close matchups like the 8 facing the nine. Looking at previous years, I took a rough guesstimate that I could probably (if i knew anything about the teams, which unfortunately I don't...) increase my odds to 1 in 10,000.

Which make me wonder if the billion dollar offer is risky. Course I don't have stats on how any perfect brackets have materialized overall, so my 10K to 1 may be too generous. If it is reasonable, an easy way to beat Buffett would be to hoist a website and take to social media for recruits. Generate 10,000 unique brackets that cover all the possible (realistic, from a statistical point of view) outcomes. The assign each person who signs up a unique entry to the contest. The people who sign up contract with you to pay the site the 1 billion if their bracket wins. So, everyone who signs up as an excellent chance of walking away with $100,000-- It's a free entry, after all!

Am I on to something, or is my 1 in 10K estimate too liberal?

 

[boooeee]

I was reading a story today (sorry no link handy but google is your friend) about Warren Buffett/Quicken Loans offer of 1 billion $ to anyone who submits a perfect NCAA bracket for this year's tourney. The story insisted the odds were one in a billion. That seems ridiculous to me, but I'm not sure how one would determine the true odds unless maybe one had a database of human picks.

But it gives me an idea--what if you could get a huge amount of data on all the matchups and do a hard analysis; looking at previous years' tourneys it is exceedingly rare for a lower than 12 seed to win in the first round, and usually those first round upsets are not as 'upsetting' as the rank would indicate. Then, it's even more rare for the rare upset to survive the second round. So, of course the key is to identify the potential upset games, and also to give oneself a better than 50/50 edge for the close matchups like the 8 facing the nine. Looking at previous years, I took a rough guesstimate that I could probably (if i knew anything about the teams, which unfortunately I don't...) increase my odds to 1 in 10,000.

Which make me wonder if the billion dollar offer is risky. Course I don't have stats on how any perfect brackets have materialized overall, so my 10K to 1 may be too generous. If it is reasonable, an easy way to beat Buffett would be to hoist a website and take to social media for recruits. Generate 10,000 unique brackets that cover all the possible (realistic, from a statistical point of view) outcomes. The assign each person who signs up a unique entry to the contest. The people who sign up contract with you to pay the site the 1 billion if their bracket wins. So, everyone who signs up as an excellent chance of walking away with $100,000-- It's a free entry, after all!

Am I on to something, or is my 1 in 10K estimate too liberal?

Check out Take Buffett's Billion. It is pretty much what you proposed (full disclosure: I am currently part of the team behind the website; they'll be using my ranking system, among possible others, for generating brackets). So sign up, and tell your friends.

By my calculations, the odds of picking a perfect bracket are about 1 in 50 billion (with variation from year to year).

Even though the crowd-sourcing approach should tilt the odds further in our favor, we're still talking, at best, a one in a million shot.

 

[BowlOfRed]

Some speculation and numbers in this thread as well.
http://www.internationalskeptics.com/forums/showthread.php?p=9782930

 

[Bob001]

Check out Take Buffett's Billion. It is pretty much what you proposed (full disclosure: I am currently part of the team behind the website; they'll be using my ranking system, among possible others, for generating brackets). So sign up, and tell your friends.

By my calculations, the odds of picking a perfect bracket are about 1 in 50 billion (with variation from year to year).

Even though the crowd-sourcing approach should tilt the odds further in our favor, we're still talking, at best, a one in a million shot.

It's an interesting idea, but is it enforceable? Suppose, against all odds, one of your participants wins with one of your brackets. Who says he has to claim the prize in the name of "Winner LLC" or whatever mechanism you have set up? Will the contest rules even allow it? With a billion dollars at stake, he'll be able to afford pretty good lawyers and lengthy litigation.

 

[Bob001]

I was reading a story today (sorry no link handy but google is your friend) about Warren Buffett/Quicken Loans offer of 1 billion $ to anyone who submits a perfect NCAA bracket for this year's tourney. The story insisted the odds were one in a billion. That seems ridiculous to me, but I'm not sure how one would determine the true odds unless maybe one had a database of human picks.
....

From the official rules (emphasis added):

Odds of Winning. The odds of winning the Grand Prize depend on the outcome of each of the Tournament games and the skill of each Entrant. Mathematical odds of winning the Grand Prize are 1:9,223,372,036,854,775,808, which may vary depending upon the knowledge and skill of the Entrant, and the odds of winning a First Prize depend upon the outcome of each of the Tournament games, number of eligible Entries received during the Entry Window and the skill of each Entrant.

http://www.quickenloansbracket.com/rules/rules.html

 

[s4zando]

I was reading a story today (sorry no link handy but google is your friend) about Warren Buffett/Quicken Loans offer of 1 billion $ to anyone who submits a perfect NCAA bracket for this year's tourney. The story insisted the odds were one in a billion. That seems ridiculous to me, but I'm not sure how one would determine the true odds unless maybe one had a database of human picks.

It is. It's closer to the quintillions.
ETA: See Bob 001's link.

But it gives me an idea--what if you could get a huge amount of data on all the matchups and do a hard analysis; looking at previous years' tourneys it is exceedingly rare for a lower than 12 seed to win in the first round, and usually those first round upsets are not as 'upsetting' as the rank would indicate. Then, it's even more rare for the rare upset to survive the second round. So, of course the key is to identify the potential upset games, and also to give oneself a better than 50/50 edge for the close matchups like the 8 facing the nine. Looking at previous years, I took a rough guesstimate that I could probably (if i knew anything about the teams, which unfortunately I don't...) increase my odds to 1 in 10,000.

There are folks that have been doing this for decades. None have resulted in 100% accurate bracket predictions. See Bob001's links. I'm actually more interested in seeing you prove them wrong. The human element in this game, manifested by over 1000 19-23yr olds (the most predictable of all age groups) is the x-factor no-one has ever overcome.

Which make me wonder if the billion dollar offer is risky. Course I don't have stats on how any perfect brackets have materialized overall, so my 10K to 1 may be too generous. If it is reasonable, an easy way to beat Buffett would be to hoist a website and take to social media for recruits. Generate 10,000 unique brackets that cover all the possible (realistic, from a statistical point of view) outcomes. The assign each person who signs up a unique entry to the contest. The people who sign up contract with you to pay the site the 1 billion if their bracket wins. So, everyone who signs up as an excellent chance of walking away with $100,000-- It's a free entry, after all!

Am I on to something, or is my 1 in 10K estimate too liberal?

The answer to the highlighted is 0, thus the odds. If this is still a discussion topic in 2020, I may recommend to the PTB on this site that it be included in the MDC. Resoning: with every passing year, there is more and more technology, biology, kinesiology, and 'bracketology' going into these predictions. Although people can put together brackets that are outperform 'chance,' no-one seems to be able to put together a bracket that can predict the outcome of 67 games, even when the field is 'whittled down' to the most qualified teams*

*This is up for a dubious debate that would be better suited for another sub-forum.

By all means, nothing is stopping anyone. I encourage you to beat the system. The point of this reply is to point out that no-one ever has, and there are a lot of folks out here that are pretty good at this, professionally.

 

[stanfr]

Thanks for the responses, guys!

It looks like i did underestimate how difficult it would be to pick a perfect bracket. Of course, I was not talking about the 'pure' stat of 64 or 68 coin flips, which i knew was astronomical, I was talking about an informed calculation. Upsets are the exception.

boooeee--very cool! Thanks for the response; ill check it out!

Bob001: it would definitely be enforceable. A contract is a contract. Lottery pools happen all the time.

I still am 'skpetical' that a perfect bracket is as difficult as you claim s4--I know for a fact that when I participated in pools in years past, I know of at least one occasion where a person got all but one game right--I saw the brackets beforehand! It wasn't a fluke, just a perceptive individual. Really, the 'perfectly matched' team games are the exception--while upsets are possible it is usually (from a stat point of view) the team that has an inherent advantage that ends up winning. The key is to properly determine who has the edge, which of course is not trivial. But if you can 'whittle it down' your odds obviously increase significantly.

 

[Bob001]

.....
Bob001: it would definitely be enforceable. A contract is a contract. Lottery pools happen all the time.
....

Sure, people agree to chip in on buying a block of lottery tickets and agree that if any single ticket wins they share the money (although there have been some nasty lawsuits even about those). Then they claim the prize as a trust or partnership. But in this particular case, the participant registers for the drawing as an individual and submits his entry as an individual. If he wins, the drawing will announce that he's the winner. He might even be personally liable for all the taxes, even if he does try to comply with the agreement. (Lottery winners have been hit with gift taxes for trying to name family members as co-winners.) It's all totally theoretical -- Warren Buffett is pretty sure nobody's going to win -- but I'm not sure the distribution is that cut-and-dried.

 

[stanfr]

well...at least i learned my lesson. 1 in 10K was wayyyy too optimistic. Probably closer to 1 in 1,000,000. Intuitively it still doesnt make sense to me, but i think that is bias from playing monday-morning quarterback (Dayton beating Ohio St for example makes perfect sense NOW when you look at how they played their final few games of the season). Buffett knew what he was doing, and I wasted valuble time signing up...ah well.
Thanks for givning me the shot at it, Boooeee!