Warren Buffett offers $1 billion for perfect March Madness bracket
forbes.com
forbes.com
"accounting for team strength, the chance that each of the 63 tournament games is won by the favored team is a mere 1 in 70 billion".
Which implies an expected payout (a billion dollars times the 1 in 70B chance) of 1.4 cents.
Hopefully this will put paid to all the crazy comments about match fixing... Buffett is 'giving' you 1.4 cents, and that's inspiring people to talk about organising a massive match fixing conspiracy? Get real!
All this talk of match fixing is just noise; sure, it could be an interesting topic of discussion, but it is completely unrelated to Buffett's offer.
Unless you really think people out there weren't going to try and fix matches, but suddenly they change their mind now they've been given a free 1.4 cent bet.
Say, it takes 10 minutes to make the bet. That would compute to the average gain of 6*$0,23 = $1,38/hour.
This disregarded the $100,000 paid to the top 20 as well as the possible psychological effects from making such bet, such as a positive outlook on life for the duration of the bet, etc.
There have been four $580m+ lottery wins in the US in the last 18 months (powerball & mega millions). $2.4 billion split among eight people. The odds were beyond absurd, and yet these people still won. Granted, I like my odds better in the lottery than Buffett's wager.
You're right, it probably over values it due to the marginal utility[1] of money and the non-zero value of the risk premium[2]
You have as many chances at winning a billion dollars as you want. All you have to do is make lots of smaller bets strung together. It may seem impossible to string 30 winning red/black roulette bets (starting with a dollar), but actually the odds of winning a billion dollars that way are roughly the same. Technically speaking Casinos don't have the capital to support the super large bets, but you could easily transition into 50/50 bets in the equities markets. There is enough capital there to give you a shot at a trillion dollars. All you have to do is take it.
[1] http://en.wikipedia.org/wiki/Marginal_utility [2] http://en.wikipedia.org/wiki/Risk_premium
No it doesn't. It's identical, except that rolling your bets gives you the additional option which can only be considered more valuable.
And interestingly in this case, Buffet will actually provide the same opportunity if it comes to it. If you get the first couple rounds of the bracket correct, Buffet will be right there with a check offering to buy you out of your position.
If I had (only) 500m USD the chance of me letting it ride with 2x return and 50% chance of winning is precisely zero.
So, with the roulette example, I would never win 1bn USD. From what you say about Buffet's early buy-out, the same may apply in this case.
They made the naive assumption that all games are independent events and that all games have exactly a 50% chance to go to either team. Then they computed 2^32, which is STILL massively incorrect even with those assumptions because there are 63 games, not 32.
However, in this case, it's Quicken taking the loss on the bet (in exchange for promotional value), not the people making the brackets.
That's not to say making the brackets is worth doing since there are other costs associated with it, but you aren't betting against Buffet by doing so.
Quicken isn't taking the loss. Berkshire Hathaway is insuring the bet, and will eat the radical majority of the loss should a win occur (minus the kicker Quicken is giving Berkshire to do it).
"Buffett said one of his insurance companies is writing the $1-billion policy in exchange for a premium, which neither he nor Quicken would disclose."
The best quote on all of this so far:
"This will be the most fun. Just imagine if there's one person left [with a perfect bracket] at the last game," Buffett said. "I will go to that final game with him or her and I’ll have a check in my pocket. ... I think we'll be rooting for different teams."
http://www.latimes.com/business/money/la-fi-mo-warren-buffet...
But what I meant is that it's not a bet against Buffet since you don't have to pay Buffet anything if you lose. The real bet that Buffet is making is with Quicken, where the stakes are probably $10 million to $1 billion, with 1:2000 odds in Buffet's favor.
OK, maybe your IP would get banned, but what if you have access to a botnet and, therefore, lots of IP addresses?
OK, maybe they have a captcha, but what if you trick your botnet victims into filling in the captchas during their normal web browsing sessions, by pretending to be a google bot-detection page?
OK, so you win, but the address is a random address that isn't yours. How much would you have to give the person whose address it really is, to pretend you live there? Half? Still not bad.
Completely random is 1 / 2^63 (1 in 9,223,372,036,854,775,808 or about ten million times worse). Granted, they're not completely random, but it seems like to get the odds they're saying you can take a LOT of the game outcomes for granted. I don't follow March Madness that closely, so maybe that's the case? Just seems odd for a tournament to have so many nearly guaranteed outcomes.
It's obviously wrong, of course, but journalists aren't typically known for their statistical math backgrounds. Interestingly, this article claims you can actually get the odds even lower, to about 1 in 1 billion: http://www.latimes.com/business/la-fi-buffett-basketball-bet... using some more sophisticated team/player data.
If you're trying to win the perfect bracket challenge, you can do much better than flipping a coin for each pick. Instead, flip a biased coin, with the higher seed favored based on Sagarin rating.
http://www.realfreemarket.org/blog/2013/03/16/what-are-the-o...
[Picking mostly favorites is also a stupid strategy if you're trying to win, because, if you do win, you're likely to be tied with others and split the prize.]
Well, it's a shame 1 billion probably wouldn't be enough to buy up all the 63 teams - and leave some winnings for the effort...
But I suspect Voltaire and friends will find a way to win this. (The odds of rigged games just went up 300%.)
I can't really decide if it's a good use of time bc time is not fungible (easily replaced) or whether or not I am contributing to something that values luck over hard work.
Or maybe I need to get some sleep and get off HN! :)
Assuming equivalence of payments and probability FTA.
Ignoring secondary prizes and solvency costs.
> Miller, the Duke professor, came up with his 1-in-1-billion probability through an equation that placed games into categories ranging from close games that could go either way to near locks. Based on his finding, Buffett would need to charge a premium of about $10 million to break even against his expected results, Miller said.
> "If I were Warren Buffett, anything over $10 million, I would probably do it," Miller said. "If $1 billion were going to ruin me, I wouldn't. But it's not going to ruin Warren Buffett."
> Buffett said his company is big enough to survive such a hit. "We've lost more money in a given event before," Buffett said. "Hurricane Katrina probably cost us $3 billion. "We will put more at risk in a given insurance transaction than anyone in the world. But we have more capital than anyone in the world."
> Berkshire Hathaway investors can take comfort in some news Buffett disclosed Tuesday. He said he would probably strike a deal — at significantly less than $1 billion — with anyone who gets deep into the tournament without missing a game.
> "If you get to the Final Four with a perfect bracket, I may buy you out of your position," Buffett said. "I'll make you an offer you can't refuse."
Buffett appears to be channelling Vito Corleone in that last statement right there: http://www.youtube.com/watch?v=SeldwfOwuL8
So assuming the new 1 in a 10^9 chance figure, we get:
EV cost: (1/(10^9)) X (5 X 10^8) X (10^7) = $5m
Now the previous time Buffett made this kind of a bet was here: http://en.wikipedia.org/wiki/Pepsi_Billion_Dollar_Sweepstake...
That had an EV cost of ~$1m as well (http://www.bloomberg.com/news/2014-01-21/buffett-makes-milli...), and he got paid a ~$10m premium from Pepsi for that risk.
So I'm guessing the same is true in this case. That is a 10x EV cost premium, with a range of ~$10m-$50m. I highly doubt that Quicken is willing to pay so much, so it'll be biased towards the lower end of the range. At the end of the day this is essentially a ~$10m-20m advertising project which uses free distribution through news/blogs/PR releases/forums/TV/radio/word of mouth/mind share, in addition to getting access to private consumer data when people sign up for the competition (email/address/name/age/etc).
Hopefully Quicken will have more luck and consumer buy in than Pepsi had in 2003 (http://www.psychologytoday.com/blog/the-decision-tree/201306...). I doubt it though. The more likely outcome here is that Buffett made himself a cool ~$10m-20m in one day for doing very little work, since he doesn't pay for any of the operational costs of the competition.
> Jay Farner, Quicken's president and marketing chief, said his company would benefit from the contest in two ways — news coverage and access to the email addresses of millions of potential customers. Anyone who enters the contest will have the option of receiving email offers from Quicken, he said.
Source: http://www.latimes.com/business/la-fi-buffett-basketball-bet...
Expected loss by Buffet, assuming all 10 million brackets are different:
5x10^8 x 10,000,000/4,294,967,296 = 5x10^8 x 0.0023 = $1,164,153.22
He can make this up by offering a lunch [1]
[1] http://gawker.com/5917080/the-cost-of-private-lunch-with-war...
edit: whoops looks like Confluence updated his post with this number too. :)
"The $1 billion will be paid in 40 annual installments of $25 million. Or if you don’t want to wait around that long, you can claim a lump sum payment of just half: $500 million."
Interesting!
Which one would be more effective if any of us would face this decision?
The dollar has lost an immense amount of real value in the prior 25 years (tracked against almost anything of consequential value), and the Feds / Fed weren't being anywhere near as irresponsible as they have been lately. Taking annual payments puts you up against having to match that devaluation just to stay even. I don't like what might happen to the dollar in just 25 years, particularly in the era of massively heightened currency competition likely to put even more downward pressure on it (eg bitcoin and whatever comes next).
Also, while it's possible tax rates will be lower in the future, I'd bet against that strongly given the bills we have coming due. I'd lock in today's tolerable tax rates, versus potentially ending up with Carter era 70% rates or 79% to 94% (1930s-1950s era).
The only scenario I've seen that makes any sense, in which you shouldn't take the lump sum, is if you have some personal circumstances that go beyond the sheer math of the situation (eg you have an intense lack of personal control over spending, and think you would manage smaller annual sums better, although you can still borrow against annual payments and bury yourself; or perhaps if you have an estate that you want annual payments to go through to your kids, to prevent fighting over a larger lump sum; or if you actually think you can significantly beat inflation).
The biggest one is the fact that the lump sum is less than the 40 year payments. If it was $500m now vs. $100m each year for 40 years you'd definitely chose the latter. Or if it was $500m now vs. $300m x2 years.
The break even point is about 4.2% annually. If you can earn 4.2% annually and you invest the entirety of the earnings as you get them, you'll end up with about $2.6 B with either option.
If you earn less than 4.2%, you will end up with more from the 40 year annuity. If you earn more than 4.2%, you will end up with more from the lump sum.
Remember that interest rates are historically very low right now and are likely to vary significantly over 40 years.
0%: $1B with annuity, $500m lump sum
5%: $3.2B with annuity, $3.5B with lump sum
10%: $12.2B with annuity, $22.6 B with lump sumAnyway, it says your odds are a 1 in 4.2 billion for correctly picking the winning 63 games. What that basically means is that we're going to find out if time traveling exists sometime in March.
The article is very wrong about 1 in 4.3 billion being the odds though. That's just 1 in 2^32 which isn't even close to the correct math for figuring the odds.
Accounting for skilled handicapping, the consensus seems to be 1:1 billion odds. Maximum 10 million entries, so there's roughly a 1 in 1000 chance we'll have a legit winner (well, that would be assuming everyone played with near optimal strategies).