Bryan Caplan  

What a Bet Shows

PRINT
What monopsony power in the la... Larry White on Monetary Econom...
Whenever I win a bet, critics rush to say, "This doesn't prove you know better than me."  They're right: Bets "prove" nothing.  But that's a silly standard.  For empirical questions, definitive proof is unavailable, so we have to settle for degrees of probability.

By this reasonable measure, how probative are bets?  Taken individually, most bets are only marginally so.  Even if one side offers 1000:1 odds and loses, this might only show that the loser was a fool to offer such great odds, not that his position on the underlying issue is flat wrong.

When we move from solitary bets to betting records, however, bets are telling indeed.  A guy who wins one bet could easily have gotten lucky.  But someone who wins 10 out of 10 bets - or, in my case, 14 out of 14 bets - almost certainly has superior knowledge and judgment.  This is especially true if someone lives the Bettors' Oath by credibly promising to bet on (or retract) any public statement.  A bet is a lot like a tennis match: one victory slightly raises the probability that the winner is the superior player, but it's entirely possible that he just got lucky.  A betting record, in contrast, is a lot like a tennis ranking; people who win consistently against any challenger do so by skill, not luck.

After losing our unemployment bet, Tyler Cowen objected that betting:
...produces a celebratory mindset in the victor.  That lowers the quality of dialogue and also introspection, just as political campaigns lower the quality of various ideas -- too much emphasis on the candidates and the competition.
Tyler's gets very close to the truth, but misses the most important lesson.  Namely: In a busy world, ranking people by accuracy is not only helpful in the search for truth, but vital.  No one has time to listen to more than a fraction of the armies of talking heads, or the memory to recall more than a fraction of their arguments.  If we want to know the future as well as we're able, we need to identify people with good judgment - and ignore people with bad judgment.  The first step, of course, is to unfollow pundits who refuse to put their money their mouths are.  We should take them as seriously as self-proclaimed tennis "champions" who've never publicly played a match.




COMMENTS (17 to date)
Nick writes:

Mainly agree, but I don't think this part is entirely right:

When we move from solitary bets to betting records, however, bets are telling indeed. A guy who wins one bet could easily have gotten lucky. But someone who wins 10 out of 10 bets - or, in my case, 14 out of 14 bets - almost certainly has superior knowledge and judgment.

A good bet is one that has positive EV in dollars. So I don't think W/L ratio is the right measure — a strong bettor could always find himself getting good odds betting for low-probability events. In that case, his W/L ratio could be quite poor, while his long-run earnings would tell a different story.

John Hall writes:

The fact that a person wins many bets may not be the most relevant. Think of Warren Buffett's baseball analogy for investing. You may only be swinging at the obvious bets.

Perhaps more relevant, add up the total amount of money made on these types of bets and compare who has won the most.

Eliezer Yudkowsky writes:

To the other commenters: beware of being too smart to see the bloody obvious. Bryan wins a lot of bets. It means something important. Nothing else to be said should contradict this.

Maciano writes:

You did lose to Steyn on Brexit, right?

Or do you still consider this a hanging bet?

MikeP writes:

It should be considered a hanging bet:

Before the June 23 Brexit referedum happens, I want to clarify that a majority vote for "Leave" is not sufficient for official withdrawal. I will happily concede defeat if and when (a) the European Union officially removes Britain from its list of members, or (b) the British government officially announces a unilateral withdrawal.

One of Bryan's betting strategies is to bet for inertia. One may not know the exact mechanism of the inertia ahead of time, but what is playing out in the UK today looks an awful lot like inertia in progress.

Dylan writes:

I'm not much persuaded by the types of bets that are won, because the things that I have interest in are not what will actually happen, but on what should happen. I'm interested in your arguments on why open borders is the morally correct position, but I have trouble seeing how you could construct a bet around that position. For any position that is remotely controversial, I don't think I have any special insight into what will happen, but that doesn't mean I can't have strong opinions on what ought to happen.

J Mann writes:

For me as a blogging consumer, the biggest value in bets is that they clarify positions.

Both parties have an incentive to spell out their prediction in detail, and to assign odds to the outcome. There's also a good value that both parties now have more skin in the game (mostly reputational) and will hopefully give some more thought.

Several years ago, Krugman and Mankiw got into a little dust-up over Mankiw's criticism of Obama's recovery predictions. Mankiw offered to bet, but Krugman didn't respond.

Later on, when history proved Mankiw right, Krugman clarified that he didn't mean that Mankiw was wrong in his conclusion, only that Mankiw was a tiresome fool to cite to the unit root hypothesis in reaching that conclusion. A bet would have helped those of us following the dispute to know what the specific disagreement was. :)

http://www.themoneyillusion.com/?p=26068

KevinDC writes:

Maciano,

I believe the bet was about whether members of the EU would exit the EU by a given date. Thus, to win the bet, it's not enough for a nonbinding referendum to have been passed, or even for formal negotiations to have begun. The conditions of the bet require an actual withdrawal, not a statement of intention to withdraw at a later date.

marke writes:

tyler can be such a stick-in-the-mud.

:p

- mark

Andrew_FL writes:

"Bryan wins a lot of bets. It means something important."

Well it certainly means he's good at betting.

"If we want to know the future as well as we're able, we need to identify people with good judgment - and ignore people with bad judgment."

We should probably discount people on subjects where they consistently make incorrect predictions. The problem with Bryan's track record is that many of his bets cover different subjects, each of which he may or may not know what he is talking about on. 14 out of 14 bets on a single subject is much more impressive than 1 out of 1 each on 14 subjects.

There is a division of labor in expertise.

LD Bottorff writes:

1. Bets are a way of indicating how strongly a person believes in a prediction.
2. A consistent winning streak is a sign of good judgement. Either the winning bettor knows the subject well, or he has the judgement to limit the odds. Professor Caplan seems to have done this well.
3. Winning a bet does not require the winner to have a celebratory mindset. Professor Caplan bet that Hillary Clinton would not win the 2016 election. Does anyone think that he's celebrating the election of Donald Trump?

Even if you end up losing on the UK pullout from the EU, you still have an impressive record!

Jack Davis writes:

Nick is right on the money (sorry for pun). Evaluating a bet isn't as simple as getting it right or wrong. Have to take odds into account.e.g. I lost a bet recently on the Cavaliers not because I thought they would probably win but I figured they had a better chance than the odds.

Project idea: A website like PredictIt or BetFair where pundits can make bets with one another and develop a public track record. Users can view the personal profiles to get a sense of the track record of a given pundit, but should also respect the fact that a given pundit participated at all, even if the pundit is a perennial loser. Pundits can propose the bets (publicly) and/or accept and negotiate the terms of the bets with one another (i.e. the outcomes, odds, and payoffs are fully controlled by pairs of pundits), and can also copy or pile onto one side or another of the bets of other pundits (this latter feature is tricky though). I'm not sure how open or closed something like this ought to be, or what constitutes a "pundit," or who should make that determination, but the purpose would presumably to focus more on the louder of the talking heads...

Ordinary users of the site could also bet against one another on who will win what bets, similar to PredictIt, which would give a sense of how contrarian the pundits are in taking their side of the bet.

Chip Smith writes:

I'm a little surprised that Bryan took the election bet since it seems out of step with what I gather to be his strategy of hedging against contrarian fashion and in favor of resilient trends.

One thing that occurs to me is that if Bryan starts losing, it could be an early indication of serious unraveling.

Andrew_FL writes:

"I'm a little surprised that Bryan took the election bet since it seems out of step with what I gather to be his strategy of hedging against contrarian fashion and in favor of resilient trends."

Bryan won his bets against a Clinton presidency and a Cruz nomination for different reasons than he I think he expected.

1. Betting Clinton would not win made a lot of sense from the PoV of resilient trends, before Trump was nominated. Parties rarely successfully win three Presidential elections in a row anymore

2. Virtually everyone assumed the Republicans would nominate some moderate establishment type like they almost always do.

Bryan did not explicitly bet that Trump would get nominated and beat Clinton so there's no reason to infer from the fact that he won those bets that he saw that coming.

EDJ writes:

Another possibility is the skill of a debater is high. Debating is an art with skills. Was the person adequately prepared. I agree this record is a good thing but the 0-16 Detroit lions could have likely beaten the College Champions of that year.

Jobin writes:

Is there somewhere all of Bryan's bets and the outcomes are listed?

Comments for this entry have been closed
Return to top