Covid Prediction Markets at Polymarket

We now have some useful prediction markets up on Covid issues, so it’s worth looking at what they say and thinking about what other markets we could generate. I encourage you to suggest additional markets in the comments, with as much detail as possible.

As I wrote a while ago, if you want prediction markets to be successful, you need five elements:

  1. Well Defined. The resolution mechanisms must be clear and transparent.
  2. Quick Resolution. The longer it takes, the less people will be interested.
  3. Probable Resolution. There has to be a definitive outcome most of the time.
  4. Limited Hidden Information. If others know what I don’t, I’m the sucker.
  5. Sources of Disagreement and Interest. Suckers at the table.

To these, of course, we can add an implied sixth, which is

  1. Real Money. Money talks, bullshit walks. Are we doing this or not?

Thus, now that we have Polymarket posting real Covid-19 markets, we have the potential to learn things that matter, with a level of resolution we can use, and I’m glad I’ve been able to help them figure out what markets to offer.

We need real money prediction markets if we want to rely on their information. I’m glad Metaculus exists and puts up some markets, and it’s good for getting some idea at all, but it is not remotely the same thing. Their main advantage is that, exactly because they don’t use real money, they can violate the five pillars and those who are inclined to participate at all will still participate because there is no cost of capital, need to invest tons of attention or worry about being the sucker when it’s for internet points.

Whereas at Polymarket, a prime motivation of liquidity providers is to use the markets to generate information, providing motivation in turn for others to come participate.

There are three new markets up, let’s check them out.

Will Omicron be >1% of all USA cases by the end of the year?

Here’s what it would look like betting $1000 or $10,000 on this respectively.

It was my suggestion to use a relatively low threshold like 1% of cases, in order to allow the market to resolve faster. There is an interesting and important question of whether Omicron would then fully displace Delta, but if Omicron goes from 0% to 1% in a month, going from 1% to 50% is a lot less exponential growth than that and should happen that much faster. The game is very much over at that point.

If the chance of 1% happening in time gets very low or very high, or there’s sufficient volume, other markets with other percentages and/or longer time horizons (or other locations) can be created to continue asking the questions that matter and improving our estimates. For now, it seems better to focus on only one market of this type, to ensure better liquidity.

It’s possible that some people are trading thinking that 1% isn’t that much, as opposed to it being the majority of the way to 50%, which is one of many possible sources of good trading.

For those in other countries like the UK, this is still mostly the question you care about, since the growth rates in both places will be highly correlated.

Currently the market is putting this result at about 67% to happen within the month, and you can bet four figures without too much slippage in price, and five figures if you’re confident enough to buy up to 77% to do so, which could easily be a good buy especially if you’re reacting quickly to new information. If you think that the market is off by a lot, I’d encourage you to come in and participate.

You can also choose to provide liquidity to the market to encourage other participants. If there’s back and forth trading you make a profit, but you’ll be on the wrong end of any permanent market moves, including the final move to 100% or 0%, unless you withdraw your funds first. It’s not an easy way to make a profit and needs to be timed and monitored carefully, but it is helping to generate a public good. Here we already have $10k in liquidity, so we’re already doing pretty well, but more always helps and we see signs of at least some back-and-forth trading.

The current market price at least seems reasonable. If I had to either buy or sell right now, I would buy, but I definitely wouldn’t move the market much doing so. It’s a tight timeline. We definitely have cases already, but there’s a wide range of potential transmission advantages for Omicron and a wide range of how many existing cases there are, and 1% in a month could go either way. Discussion in the comments is encouraged, especially backed by toy models and actual math.

Will the FDA give emergency use authorization or otherwise approve Paxlovid before 2022?

As I’ve noted a few times, Paxlovid Remains Illegal. When will this change?

It’s not like this is urgent or anything. Looks like it’s a toss-up whether it will happen by the end of the year.

That’s a lot of liquidity, and it’s currently at 49%. If you can figure out the timeline first, there’s a lot of profit available. You take a hit to your price when playing big, but it’s a remarkably reasonable hit if you know something.

One month from now would be fast, slightly faster than the meeting on MolnupiravirThere doesn’t seem to be any sense of special urgency coming from the FDA. The flip side is that there is outside pressure on them on this one (that hopefully I am helping to increase somewhat), and the symbolic cost of this not stretching into 2022 seems non-trivial. With Omicron now on its way, there’s more reasons to apply pressure and to accelerate the schedule, and Metaculus had this as a coin flip before we knew about Omicron. So while this could definitely go either way, I’m guessing this is somewhat low.

Will the Omicron variant be marked as a Variant Of High Consequence by the CDC before 2022?

The definition is: A VOHC has clear evidence that prevention measures or medical countermeasures (MCMs) have significantly reduced effectiveness relative to previously circulating variants.

By the definition, it seems like there is a much higher than 10% chance that this will ultimately be a variant of high consequence. The second question then is how often that would cause the CDC to make that declaration by the end of the year? The CDC tends to always be behind the curve and hates doing useful things under uncertainty, so they’d essentially need proof.

A month seems like a long time to me. By the time Omicron is 1% of cases in the USA, it seems like we should have more than enough information to know if vaccine effectiveness against infection has declined substantially, purely through looking at the data from sequenced individuals, but will type of evidence, together with the mutations and lab analyses, be considered sufficient? That’s the question being asked here, as much as the question of the physical properties of Omicron. To win on Yes, you need to win both wagers, hence the currently low price.

I definitely wouldn’t sell here unless I knew a lot more about how the CDC operates and knew they would move sufficiently slowly or demand an impossibly high standard of evidence. I’d be worried before buying that someone else did sell down to here, and that they did have that knowledge, so I’d want to investigate further.

What we’d like to know, more than whether the CDC will officially acknowledge that Omicron is a VOHC quickly, is whether Omicron actually is such a variant, and to what extent. That would unfortunately require some combination of a longer-dated market and a means of resolution that would answer the question.

I’ll also mention the market that has the highest trading volume.

Will the United States report a record high 7-day COVID-19 case average before January 1, 2022?

This market is less interesting than it was earlier, since it seems clear that we’re not on a trajectory to set a new record before the end of the year, but it seemed worth highlighting that volumes can get meaningful over time. Here, the volume has almost reached $250k.


The problem of things like VOHC declarations not telling us what we want to know leads us to the issue of oracles.

The best way to resolve a wager is to have a reliable source and a well-defined outcome. Unfortunately, for many of the things we care most about, this is not possible. For example, suppose we wanted to know if Omicron cases were more severe, less severe or about as severe as Delta.

In theory, we could go the Metaculus route and say something like ‘of the first five published papers of the right type in the proper journals’ and come up with a tight definition. In practice, that’s not great. It potentially takes quite a long time, still involves a bunch of judgment calls and/or complex definitions, and could end up giving us the wrong answer.

This is one of the big issues with prediction markets in practice. It’s not that useful to go betting exactly how the resolution mechanism will work, or to wait around for months or years for resolution.

Another proposal is to use a social mechanism. The Augur protocol, for example, resolves markets by splitting its entire chain in two if there’s sufficiently robust disagreement, assuming the truth-tracking chain will be the valuable one. I do not much trust that assumption, and it turns truth into a social phenomenon or one about money, influence and power. You’re now predicting social reality not physical reality. That’s exactly what prediction markets are there to prevent.

As an exercise, what would happen if there was a disputed presidential election that wasn’t as clear who was right as it was in 2020? Who wins the bets on the 2020 election if January 6 goes differently? Would the actual winner win bets, or would it be a partisan dispute resolved along partisan lines?

Who do you think is going to make the College Football Playoff this year if Alabama loses a close game to Georgia? What do you think is driving that decision? The answer depends on whether you think the group that’s deciding is trustworthy, and what criteria you think they are using. Neither of these things are at all clear.

Thus, the proposal to use a trusted person, or trusted small group, as an oracle.

We can designate a trustworthy person or small group, and let that person decide the outcome. My suggestion, with his permission, is that this person be Scott Alexander.

(If asked, I would be willing as well, but Scott seems better for this.)

He’s probably not the best possible choice, but he seems like a very good choice.

We can certainly do a lot better with someone like Scott and a well-written question than an untrustworthy Oracle using murky criteria (again, see: College Football Playoff Committee) but that only goes so far. There’s still some amount of wagering about how the Oracle will think and make its decision.

I still think it is worth doing, because it allows us to ask the questions we care about and get reasonable answers. I am confident that if it is very clear by January 1 that Omicron kills more of the people it infects, that Scott will agree with this, and that if it’s similarly clear that Omicron kills less of those people, Scott would agree with that instead. There would be cases where it was less clear, and the decision could go either way and the trading gets to be about that, but that’s a price I think we need to be willing to pay. I hope we have the interest levels necessary to justify paying it.

Future Prediction Markets

What other prediction markets about Covid-19 would provide high information value, while being sufficiently well-defined and short-term? Let’s put our brainstorming caps on and see what we can come up with. There’s a lot of willingness to help get these markets going if we can figure out the right markets to offer.

In particular, what is needed are precise questions and good resolution mechanisms and information sources. The biggest barrier to asking the questions we care about, and getting the information now, is finding a reliable way to measure that information in the future.

This entry was posted in Uncategorized. Bookmark the permalink.

16 Responses to Covid Prediction Markets at Polymarket

  1. Stephen Hayes says:

    The issue with appointing oracles is that you need to trust them not just to come to the correct conclusion about the outcome, you need to trust them not to wager a million dollars and then declare themselves right. Everyone needs to trust them this much, otherwise smart money will think it’s the oracle’s sucker and stay away from the market. I’m not sure I can think of anyone that the general betting public would trust not to invisibly engage in a bit of market manipulation.

    • TheZvi says:

      I agree that this is a risk/cost, but if one’s reputation (and the role of oracle) are higher value than the amount of profit available, it’s mostly safe. It gets less safe as the deadline approaches in a situation where the outcome is unclear, but it’s pretty much fine for things to stop trading in those spots since there’s little information value left?

      • Stephen Hayes says:

        It’s not clear to me that reputation costs work here. A corrupt oracle can get great EV by playing the market, truthfully reporting the cases where it’s obviously yes or no (taking a loss if he bet against them), and always resolving the ambiguous “well maybe it’s yes” cases in his favour. The oracle loses no reputation, and you’re still his sucker if you play the market. You could use oracles only for cases with very little probability on the ambiguous middle outcome, but isn’t the whole point of oracles over objective criteria that they’re better at resolving ambiguity?

  2. Yellowface Anon says:

    Semi-OOT: I just re-read the Israeli decision to revoke the Green Pass of everyone who hasn’t received a booster (the new regulation says it is only valid for 6 months after the last COVID vaccination). They screwed over 2 million who’ve done their part and just planned to miss out what was until then optional. That rule implied no current vaccination protection is long-lasting & you’re stuck with 6-month-revaccination for the rest of your life. Clearly there will be no end in sight.

    Is this the future we’re facing if those who wants to keep us in “permanent twilight” wants to keep the whole vaccine milk cow up? (This sounds rhetorical but I’d like you to evaluate how likely it is by seeing whether boosters are long-lasting or not.) If many countries do this, even those who have been jabbed will drop out of the vaccine passport system & walk away from a relatively normal life to join the antivaxxers.

    • TheZvi says:

      There will be a weekly post later today for such discussions.

      I find it interesting that Israel is saying 6 month limit whereas USA is saying 6 month minimum. Prohibited, forbidden, etc.

      I am happy to strongly advocate for there being an end to the madness at some point, but I am not sympathetic to the framing that this is a cash grab by pharma companies.

      • Yellowface Anon says:

        “I am not sympathetic to the framing that this is a cash grab by pharma companies.”
        Everyone has their blind spots. Not that some vaccines shouldn’t be developed – but once they are there they should be produced as cheaply & as massively as possible.

  3. The first useful market question that comes to mind is: will the proportion of the fully vaxxed (alternatively: recently vaxxed, i.e. last dose in last 6 months) population hospitalized with COVID be at least $PERCENTAGE higher than it is now on $DATE?

    This would quantify the extent to which vaccines will continue to keep people out of the hospital even if they get breakthrough Omicron infections. That in turn seems like the biggest variable in deciding how Big of a Deal it will be. The data should be available to decide the question, no?

    • TheZvi says:

      Would that be the total # of such folks, or as a % of all hospitalized?

      The problems either way are
      1) That’s a compound question of Omicron infection + its effect on the answer, which includes the vaccine rate at the time etc.
      2) We’d need to pinpoint a reliable source, I haven’t actually seen such a source yet, although presumably it exists.
      3) This doesn’t do much until after Omicron has been dominant for several weeks, so it would have to be long-dated.

      I think I’d be interested in largely
      A) The % of hospitalizations that are vaccinated.
      B) The total # of hospitalizations at some future time, which is definitely useful.

    • Nicholas Galauxy says:

      This runs into the big issue of “according to what reporting numbers?”

      When money is on the line, you start running into the issues we’ve been seeing over the past 2 years of people choosing their reporting conditions how it suits them, because we don’t have a supernatural dashboard reporting real-time accurate numbers.

  4. Pingback: Omicron Post #3 | Don't Worry About the Vase

  5. maline says:

    If Scott takes on the job as Oracle, then any time he comments on one of the relevant topics, there will be market fluctuations as people speculate on how his opinion will or won’t bias his final call. I predict that Scott will be very bothered by this and feel obligated to avoid all such comments. But of course “the relevant topics” will end up including pretty much all current events, so him being an Oracle will effectively be inconsistent with him being a blogger.

    OTOH, Scott’s scrupulousness is a major part of why he might be considered trustworthy in the first place.

  6. kronopath says:

    Worth pointing out that Kalshi also has a bunch of COVID-related real-money prediction markets as well.

    Whether the CDC will identify a VOHC (any VOHC) by March:

    Whether Omicron will make up >1% of cases in the US by new year’s:

    As well as weekly markets on vaccination rates over time:

    There’s some more in there including those that bridge COVID and politics, like whether vaccine mandates will be overturned, whether cities will close indoor dining, etc.

  7. myst_05 says:

    “What other prediction markets about Covid-19 would provide high information value, while being sufficiently well-defined and short-term?”

    1. Will Israeli hospitals have a higher 7-day average of COVID hospitalizations by Mar 1 2022 than in September 2021? => market on speed of spread, lethality, booster efficacy. Israel is the best choice due to highest spread of boosters and highest quality of hospital data, plus they already have Omicron. September 2021 is chosen as the peak of the Delta wave which happened when most elderly people there had two shots of the vaccine.
    2. Will New York hospitals have more patients with COVID on a 7-day average by March 1 2022 than on Nov 11th 2021? If Paxlovid is expected to work great, this should resolve to No. Nov 11 is chosen as the bottom of the autumn COVID wave in NY, a pretty hard number to beat.
    3. Will the state of New Mexico remove their mask mandate by Mar 1 2022? New Mexico is not as extreme as Florida or Oregon in terms of COVID precautions, so they’re a good barometer on public attitudes towards the virus. If they remove it by then, its a strong signal that the virus is no longer seen as important.
    4. Will the state of Arizona introduce a mask mandate by Mar 1 2022? Similar to New Mexico but in a different direction. Arizona is a Republican state but not as extreme as Texas/Florida. If they bring back the masks, the public is definitely worried.
    5. Will Apple mobility data show a 30% decline in movement within the UK by Mar 1 2022? This is effectively talking about a possible lockdown without having to argue over the definition of a lockdown. The UK is a solid choice as they’re not as lax as the US but also not as strict as Austria.
    6. Will a new COVID treatment clinical trial in the US be halted due to being too efficient by Mar 1 2022? Gives us info on any new drugs. It seems like “halted due to efficacy” is the fastest signal of future approval.

  8. myst_05 says:

    In regards to market makers, how would it work? If you bet $100k in both directions, it seems like the payout for new participants will still stay about the same for others? In a normal stock market you would offer buy and sell bids with a spread centered around the market average, but not sure what’s the mechanism on Polymarket.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s