Warren Buffett offers $1 billion for perfect March Madness bracket

http://www.forbes.com/sites/kellyphillipserb/2014/01/21/warren-buffett-offers-1-billion-for-perfect-march-madness-bracket/

by alphaoverlord • 13 years ago

70 70 13 years ago

forbes.com

confluence2• 13 years ago

EV = Probability X Payout = (1/(4.29 X 10^9)) X (5 X 10^8) = 0.11 cents per bet.

  Assuming equivalence of payments and probability FTA.

EV cost with 10^7 possible entries: (1/(4.29 X 10^9)) X (5 X 10^8) X (10^7) = $1.1m

  Ignoring secondary prizes and solvency costs.

I read some more details about the insurance policy here: http://www.latimes.com/business/la-fi-buffett-basketball-bet...

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...

txttran• 13 years ago

Probability is 1/2^63 (each match eliminates 1 team, 63 matches to leave one champ). Or 1/(9.22 x 10^18)

gibybo12• 13 years ago

This would only be true if every game was an exact 50-50 toss up and the games were not dependent on each other. In reality, many of the matches are skewed to one side or the other (some heavily so) and certain teams have strengths and weaknesses that work better and worse against various other teams.

eru• 13 years ago

They can probably get it insured for $2m, then.

IgorPartola2• 13 years ago

What's the point? If Buffet wins, no need to pay insurance premiums. If he loses, his insurance would cover 0.2% of his loss.

hkmurakami2• 13 years ago

Brackets will be made available on Selection Sunday, March 16, 2014, and will initially be limited to 10 million entrants, but only one per household.

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. :)

xenophanes2• 13 years ago

The payout is $500m though, plus false advertising.

prewett• 13 years ago

The payout is $25 million over 40 years = $1 billion. If you would prefer a lump sum, then you get less, because $25 million forty years from now will not be worth as much as it is now (assuming positive inflation). This is pretty standard.

soup10• 13 years ago

This is weird, besides the great odds straight up. It's extremely vulnerable to match fixing.

hkmurakami2• 13 years ago

How many teams out of the 64 does someone realistically have to bribe in order to have a good shot at getting the $1B?

velik_m• 13 years ago

Just a few refs...

confluence2• 13 years ago

I'm sure the contest contract has a clause making it void if match fixing is detected. I'm sure if there was a pay out, an investigation would occur to make sure it was legit, and if it wasn't, the feds would be brought in.

Ryel• 13 years ago

I wonder if Nate Silver is disqualified..

shawn-furyan• 13 years ago

I don't think that bracket submission interface supports the inclusion of error bars, so I'm sure Nate is welcome :)

xzel• 13 years ago

I'm disappointed by the 1 per household and million entry limit. I would really like to see someone hit it.

TheBiv• 13 years ago

When I first saw this, I thought "really cool" and then I thought about the time that my friends and myself would spend towards developing "the perfect" bracket.

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! :)

JAFTEM• 13 years ago

My brain had a hard time processing "billion" with a "B" in the title.

Anyway, 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.

gibybo12• 13 years ago

Or uncaught match fixing :)

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).

fuddle• 13 years ago

Buffet also insured a Billion dollar lottery created by Andrew Warner of Mixergy fame.. http://youtu.be/j6RSaqf9NsI

rahimnathwani3• 13 years ago

If you are the first person to write a bot to automatically submit entries, you could submit most/all of the allowed 10 million entries.

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.

mattlutze• 13 years ago

1 entry per household.

rahimnathwani3• 13 years ago

See the last paragraph. I am assuming you could use random, real addresses, and split the winnings 50/50 with the real resident.

midas007• 13 years ago

Given no information, the expectation of randomly guessing is around $ 0.16 USD.

But I suspect Voltaire and friends will find a way to win this. (The odds of rigged games just went up 300%.)

dreeves2• 13 years ago

From http://messymatters.com/the-perfect-bracket/ :

"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.

tlarkworthy• 13 years ago

so if you come up with the best strategy and figure a way to submit an 70 billion entries for 1c each around that optimal guess, you could net 2.8 million

AdamTReineke• 13 years ago

Except they're limiting entries to the first 10 million.

joosters3• 13 years ago

In other words, you could recreate Buffett's $1 billion offer by placing a 1.4 cent bet on the first game, and rolling your winnings on to each successive game. What's more, you're free to stop at any point and take some or all of your winnings home. (Which you would have to do anyway; you'll never be able to place the multi-million dollar bets near the end.)

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!

JanezStupar• 13 years ago

Don't forget the tax.

gibybo12• 13 years ago

He's giving you 1.4 cents if the games are random. If they aren't random, like with match fixing, then he's giving you $1 billion. That leaves plenty of money to bribe the relevant refs and players.

corin_2• 13 years ago

There is a slight difference, in that were you to get a couple of games away from winning the $1b there would be a huge incentive for fixing at least one match to kill your chance right at the end. Not that Buffett would do that, but if it did get to that point then he's certainly in a different position to "all I'm doing is offering the 1.4 cent first bet".

ErsatzVerkehr• 13 years ago

...assuming a completely random outcome.

sytelus• 13 years ago

This seems to be ripe for attack through ML models. It should be easy enough to create models and test it out on previous years to see its accuracy. May be some of us can do ensembles of models with bunch of best predictions (Netflix style!). Any idea where to get data sets for previous games?

dreeves2• 13 years ago

I'd bet that the strategy of always picking the favorites is about the best you can do.

ghshephard2• 13 years ago

If you had any success doing that on a full on march madness bracket, you could probably make a lot more than a Billion dollars (over time, don't be greedy) - betting on the sports betting scene, which is a (great) deal larger than $1B/year.

Houshalter2• 13 years ago

The odds of winning are 1 in 4,294,967,296 at random guessing. If your system can do 100 times better, that reduces it to 1 in 42,949,672. Or an expected utility of a measly 23 dollars, so don't waste too much time on it.

adestefan• 13 years ago

Says every naive sports gambler.

rfergie• 13 years ago

"it should be easy" is always the killer

jamespitts• 13 years ago

It is amazing how many people think that they can outwit Warren Buffett and his inner circle of actuarial geniuses :-) He probably has a part of his brain that runs the Kelley Criterion in his sleep, so long as there is enough Cherry Coke in his system.

http://en.wikipedia.org/wiki/Kelly_criterion

qznc• 13 years ago

If I were a US resident, I would just submit a random bet. A nearly-zero chance for a billion for a few minutes of my time and no cost. There is nothing to outwit here.

zen_boy• 13 years ago

The expected value for this bet is $1,000,000,000/4,294,967,296 ~= $0,23.

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.

gibybo12• 13 years ago

You are definitely not going to beat Warren Buffet in a game of actuarial wits where he's drawing the rules.

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.

adventured4• 13 years ago

How aren't you betting against Buffett?

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...

Houshalter2• 13 years ago

About a 0.2% chance of anyone out of the 10 million winning it (assuming they all guess randomly which probably isn't true, meaning it's even higher than that.) That's an expected payout of $2,328,306 for Buffet.

whizzkid• 13 years ago

Story itself is not really interesting but a paragraph in the article got my attention;

"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?

adventured4• 13 years ago

Take the lump sum every time.

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).

corin_2• 13 years ago

The only scenario I've seen...

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.

TheCoelacanth• 13 years ago

I have to disagree. In this case, the lump sum and the annuity are equivalent if you get a 4.2% return on investment. You aren't going to find a risk-free investment that gets a better return than that, so the only reason to take a lump sum is if you want to invest in riskier investments. The decision comes down to risk tolerance; there is no clear cut winner.

gibybo12• 13 years ago

This depends on the interest rate you think you can earn on the money.

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 sum

e12e• 13 years ago

"It’s not business income. Well, probably not. There are folks who enter games, contests and sweepstakes for a living (remember that Julianne Moore movie?) but that’s some serious dedication to the cause. Assuming that you did enough research and really dedicated yourself to winning, you could possibly treat it as your business (or more likely, a hobby). If you did, you would report your winnings (of course) but could also deduct any reasonable expenses associated with winning."

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...

santialbo• 13 years ago

Someone needs a 64 bit calculator.

losvedir• 13 years ago

How are the odds calculated as 1 in 4,294,967,296?

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.

gibybo12• 13 years ago

The article decided the odds were: 32 teams, each with a 50% chance of winning, so the odds must be 1 in 2^32!

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.

jrs235• 13 years ago

The journalist also ignored reality when calculating taxes owed too. They used the $1 billion up front but they reported it'd be $25 million per year for 40 years (or one time payout of $500 million).

fsk• 13 years ago

No, odds of a perfect bracket are closer to 1 in 2^47 than 1 in 2^63. The reason is that a coinflip makes too many stupid picks, such as a #16 seed beating a #1 seed. If you use the Sagarin ratings to generate a simulated bracket, the odds of a perfect bracket are approximately 1 in 2^47.

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.]

Warren Buffett offers $1 billion for perfect March Madness bracket

Related Stories