Make COVID Predictions, Take My Money: On Zvi, Bets and Taking Ideas Seriously

[Thanks to Charles Dillon (2.5x) and Daniel Frank (2.5x) for taking this bet. My capacity is now full, but if you email me with an offer on either side I’ll try to find you a match. So far, I’ve set up an additional two bets each at 2.5x.]

Zvi and Holden are betting on when Covid ends.

I am taking Zvi’s side and offering the same bet, except 10x larger. Stephen Malina has graciously accepted to take me up on half of this, so there’s another $300 on the table for anyone who wants it.

There are two issues to discuss:

  • Why am I doing this, and why for so much money?
  • Why did Holden and Zvi do this, and why for so little money?

Zvi has been writing weekly updates on Covid pretty much since last April, but didn’t catch my eye until his post exactly one year ago simply titled “We’re F***ed, It’s Over”.

Of course it’s melodramatic and alarmist in the way we should typically ignore, but Zvi is smart. Really smart. Or as Scott Alexander once described him, “[an] ​​illegibly smart [person] writing on the pandemic.” That sounds superfluous, but it’s just the warm up. In another post (about scoring Covid predictions), Scott continues:

[Zvi is] a former professional trader and sports gambler, and he does coronavirus modeling in his spare time with his family of expert biologists. Obviously he would win this one. I will never bet against Zvi on anything. Last year he bet against me on what restaurant I would have dinner at, without knowing anything about my situation or food preferences, and won anyway.

Or in another post:

When Zvi asserts an opinion, he has only one thing he’s optimizing for - being right - and he does it well.


If you’re planning the coronavirus response, maybe the best thing you can do is lock Zvi in a cave completely incommunicado and make him write one for you.


Here’s a story which is very flattering and which I hope is true: Zvi gives better advice than the Director of the CDC.

Best of all:

Maybe Joe Biden is an idiot for not appointing Zvi the Secretary of Health.

(That last one is taken out of context, but it’s too good not to include.)

I will be extremely pleased if I can ever do anything as well as Zvi does Covid modeling. Actually, I will just be pleased if anyone ever looks at me the way Scott Alexander looks at Zvi Mowshowitz.

Which is why I was disappointed and actually kind of miffed to see his recent update. With typical melodrama, he writes:

If these numbers are accurate, it’s game over, man, game over. That doesn’t mean there’s nothing to be done, but it does mean you lose, good day sir, let’s try to do it with dignity and save as many people as we can.

Or as he opens the post:

In a better world, I could focus on this full time and also maybe even hire a research assistant, and be better able to scour for information.

It’s kind of a throwaway line about civilizational inadequacy, but if you’re familiar with the whole Zvi ecosystem/melodrama/social scene it should stick out to you as extremely odd.

If you think about it for more than a second, you would ask: What on earth could possibly prevent Zvi from working on this full time? What could possibly prevent him from hiring a research assistant? He is a widely lauded public figure who regularly receives humorously high praise from Scott Alexander, who is, in turn, wealthy and well connected enough to run a grant program which includes in its description the line “I know a lot of nonprofits and rich people looking for interesting projects to fund.” So seriously, what could possibly be stopping Zvi?

As it turns out, the reasons are mostly his own opportunity cost and neuroses. As Zvi explains in the comments, he can’t work on Covid modeling full time because he’s busy making a trading card game on the blockchain. And he can’t apply for funding (which he would almost certainly receive) because “applying for large amounts of funding when you’re already well off doesn’t feel great. And I worry about the social dynamics and incentives”. And he can’t hire a research assistant not because he lacks funding or legitimacy, but because “I am confident that I am lacking key social technology to hire well and to direct such people well.”

That’s bad right?

To be clear, I’m not claiming that Zvi is acting poorly, and certainly not that he’s acting unethically. I don’t subscribe to the Copenhagen Interpretation of Ethics, under which Zvi would be prosecuted for doing far, far more than anyone else I know, just because he’s not exerting maximum possible effort on this particular issue. My view is that we should be grateful for the work he is doing.

And if you really want to nitpick, then it depends on the tractability of the issue and how much value we’d really get, collectively, from another hour of Zvi’s time. I suspect that there’s a decent chance his blockchain game will do very well, and if so, a decent chance that he’ll donate a large percentage of the earnings to very important causes. So it’s debatable, but not immediately clear that spending more time on the Covid updates would even be desirable.

But it is bad, not in the sense of blameworthiness or even ethical harm, but bad in the sense that as I complain about incessantly, we are, as a civilization, squandering our potential, and in particular our human capital. Making a blockchain game might genuinely be the best use of Zvi’s time, and he might be acting both rationality and ethically in choosing to pursue it. And so this situation is Good, but only in a very limited and local sense. The tragedy isn’t Zvi’s decision, it’s that a scenario even exists where this is the decision he has to make.

Which brings us to the question of betting. Holden’s payout if he loses is just $40. As a reminder, he’s the co-CEO of a foundation that disburses around $300 million annually, which makes the bet just hilariously small, especially given that he recently called a related Omicron question his “Candidate for ‘highest-stakes question of the next several months’”.

So what’s going on? As Holden explains:

The bet is not for a life-changing amount of money. There’s simply no way that this sort of activity is an optimal way for either of us to work toward personal financial goals. That’s not why we’re betting; we’re doing it more as a sort of “holding ourselves accountable for our beliefs” sort of thing.

That’s admirable, and the fact the money is inconsequential is fine. The point is just to state your beliefs publicly, have clear resolution criteria, and then be forced to publicly and formally say “I was wrong”.

But this is only half the game. Using bets to hold yourself accountable is like using a hammer to crush graham crackers for your pie crust. It’s an acceptable use, but far from the tool’s potential.

If Holden and Zvi really want more people working seriously on high-stakes questions, they shouldn’t hold a trivial bet with each other, they should place really big bets publicly and allow anyone to gamble against them. As a specific example, if Holden thinks that there’s a 50% chance the pandemic will end soon, and if he thinks knowing the answer to this question is important, he should allow anyone to bet against him. And he should make this bet public for up to 100,000x the size of his current bet.

There are some logistical challenges here, like holding money in escrow and whether he should use his own funds or appropriate funds from Open Phil or convince a third party to fund the bet, but they’re entirely solvable. There may also be some legal issues.

At this point, you’re probably wondering: isn’t this just a prediction market?

And yes, taken to the extreme, it more or less is. There’s a variety of mundane reasons prediction markets don’t currently work that well (1, 2, 3), but they’re not that bad either. Kalshi has some good omicron questions, as does Metaculus.

What I’m really saying is that prediction markets are public goods, and so they should be subsidized. Zvi is actually the one who wrote the post on Subsidizing Prediction Markets, and doesn’t have to be convinced on this point.

But from Holden’s grantmaking perspective, what I’m really arguing is: If you think these questions are critical, and you think there’s a high moral value of information at this particular moment, then this is the time to enact those proposed subsidies.

Finally, prediction markets (or bets) are an easy way to resolve Zvi’s social inability to manage research assistants. Instead of managing people directly, Zvi could just say “here are the important questions we need answers to”, and use subsidized prediction markets to get answers.

So what’s my role in all this? Will my 10x bet change anything? I’m not sure. Stephen took the bet almost immediately, so I don’t think it incentivized any research on his end. But there’s still 5x left, and if you want it, presumably you’ll look into Zvi’s thesis at least enough to see if there are any glaring mistakes.

You might ask if that’s worth $300 to me, but in expected value terms I’m paying far less. If I just naively think that Zvi and Holden are both reasonably smart people, then the EV of this bet for me is close to $0. So just by providing the opportunity to bet against me, I’m potentially incentivizing some non-trivial amount of work on a very important question at very low cost.

If you’re interested, just get in touch:

Or if you want to make a similar offer with another 10x, 100x or far more, email me the details and I’ll add your proposal to the top of this post.

Thanks again to Stephen for taking me up on this, providing the opportunity to demonstrate that I’m at least serious enough to put my own money on the line.