Take political betting markets literally, not seriously

US election betting markets have been having a moment. Relative newcomers Polymarket and Kalshi are attracting billions of dollars in trades, while the more established PredictIt continues to operate a few markets in the midst of a legal battle with the CFTC.

These markets all function in a generally similar way: bettors can trade shares tied to some specific outcome, such as the winner of the US presidential election. The shares trade at prices between $0 and $1. When the underlying event resolves, shares tied to the correct outcome pay out at $1 and the rest become worthless.

Betting markets have exhibited some strange behaviour in the past few weeks, though, with the implied odds of a Trump presidency climbing in a way that didn’t track with polls or election models.

The markets, of course, may end up being right. Trump could win. Whether the current odds are fair is a more philosophical, unanswerable question.

Polymarket is the biggest election prediction market by volume, and is supposedly only accessible to traders outside of the United States. As I wrote over on MainFT, the price movement has been driven largely by a small group of anonymous accounts on Polymarket that use disproportionately large limit orders.

The largest account in this group, which uses the handle @Fredi9999, owns more than 20mn shares in the main presidential market alone. Together, the four largest holders of Trump shares have nearly $40mn invested across markets on the site, almost all in holdings that imply a large Trump victory.

By analysing trading and deposit patterns, traders and internet sleuths have speculated that the four accounts (referred to hereon as FrediGroup) may be a single person or group. It’s hard to verify any connection with certainty, but there’s plausible evidence that the accounts are linked. Moreover, their public comments exhibit a similar (swaggering) tone and the same (chaotic) writing style.

Comments and messages to other users have revealed some sparse biographical details. One account claims to be French and to have lived in New York from 2000 until 2006, working as a trader. Another calls themselves an “investor and statistician”, repeatedly claiming to have “no political preference”, but rather a sincere belief that Trump is likely to win the election and that the true odds should be ”75 per cent Trump”. Très intriguant !

Yet there are some reasons to believe that FrediGroup FrèdiGroup could be playing a little dumb.

One comment appeared somewhat confused about the definition of a basis point, an odd mistake for an “investor and statistician”. The accounts have used at least four different spellings or abbreviations to refer to the state of Pennsylvania, all incorrect or uncommon. And they seemed to offer up a surprising amount of personal details, unprompted, when contacted by another trader.

Overall, their tone and reasoning come off as strangely blasé for somebody with millions of dollars at stake in a prediction market.

Is FrediGroup one or more deep-pocketed true believer(s) taking advantage of what they see as a bargain, or is there something more complicated going on here?

One possibility, much-discussed in some corners of the internet, is that FrediGroup is intentionally manipulating the market to create the perception of swelling momentum for Trump.

This theory is made more compelling by the fact that numerous conservative commentators have spent the weeks since FrediGroup’s spending spree began plastering Polymarket screenshots all over X, touting it as incontrovertible proof that a Trump landslide is coming.

In other words, the market’s movement can contribute to the perception of a large Trump lead — at least in some parts of the internet and media. This could be beneficial to a pro-Trump partisan, or possibly to someone invested in adjacent markets that would be influenced by the perception of a likely Trump presidency.

Perhaps some motivated partisan is behind FrediGroup, either directly, or by providing funds behind the scenes. But proving this is near-impossible with the public information available.

Another possibility raised by Barnard economics professor Rajiv Sethi, whom I spoke with last week, is that FrediGroup could be trading on some kind of inside information. They may know something about Kamala Harris or her campaign that is particularly damning.

If this were the case, it might be in their best interest to behave exactly as they are: playing the part of a cocksure, potentially irrational trader would keep the prices on Trump shares lower and improve their margins.

Insider trading is not impossible, but is it likely? Voting is already well underway in many states. After one of the most chaotic six months in American electoral history, what are the chances that one explosive secret has been held back that will swing the election in Trump’s favour?

Is it possible that FrediGroup is trying to make a tidy profit off of an elaborate “pump-and-dump” scheme? After all, as of Friday, the 41mn shares owned by the four accounts were up about $2mn from when they bought them.

Looking at the size of their holdings and the market’s order book, such a plan is highly unlikely to work.

All trading on Polymarket is public (it is ~on the blockchain~, after all). This would presumably make it difficult for the FrediGroup accounts to try to start slowly selling off shares without causing a stir.

For weeks, they’ve done nothing but buy Trump shares, with occasional breaks. As soon as they very publicly reverse course, the price could start to drop quickly and they don’t have much wiggle room. At an average cost per share of about $0.55, it wouldn’t take much to put them under water.

If selling shares off slowly is a problem, could they do it quickly? Nope, not even close!

As of Friday, there were bids open for a total of 3.5mn shares above their breakeven point of $0.55. If FrediGroup could fill all of those bids at once, in the moment before the price tanked, they’d make about $145,000.

This would leave them stuck with about 38mn shares with a breakeven value of about $21mn. In fact, each of the suspected group’s four accounts holds a number of shares greater than total shares bid at over 1¢.

So why are traders not gobbling up all of the potentially under-priced Kamala Harris shares? It’s possible that there is a genuine fear that FrediGroup is trading on inside information.

It’s also possible that with overall market liquidity thin, and with most pundits view the election result as a coin toss, the value-at-risk of backing Harris remains unattractive. Certainly, there is not much to gain from rushing in. So long as the large limit orders from FrediGroup keep coming, the Harris shares are likely to be cheaper tomorrow than today.

And there are other structural features of some of these sites that may interfere with their predictive power. PredictIt, for example, takes fees equivalent to a little more than 5 per cent on withdrawal, distorting the incentive to trade on small perceived margins. This may not matter as much to the large or habitual traders who roll winnings into subsequent bets, but discourages a new bettor from making a deposit for a one-off bet with limited upside, particularly as the site caps traders at $850 in any individual market.

Maybe the Trump trades are authentique. And the traders may be proved right in the end! Viewed purely as a momentum trade, they have been right so far. But a trader could be acting, to varying degrees, on other motivations.

All of this complicates the idea that prediction markets are “more accurate than polls” as Musk tweeted in early October, when he was either unaware or choosing to ignore all the research that tests the hypothesis.

It’s easy to imagine Musk and fellow influential Trump supporters creating a feedback loop that directs followers to the prediction markets then cites them as a measure of the campaign’s momentum. Of course, as an actual gauge of the election, it would be like holding a thermometer over a radiator and then holding it up as proof of how hot it is outside.

So when can we trust these markets, if we have reason to believe that the prices could be distorted or overly susceptible to meddling?

They may be at their most useful in the midst of big uncertain events, when there’s a possibility of the underlying event resolving at any time. This dramatically increases the risk of playing games with the order book, as there’s always a chance that one side of the bet could suddenly be wiped out.

One recent example of this was ahead of Joe Biden stepping aside in the presidential race. Prediction markets were particularly useful for quantifying swings in the general consensus in a way that didn’t have an obvious substitute.

Importantly, they also had the potential to resolve at any moment. Any attempt at market manipulation carried a greater risk of being left with nothing.

This will, of course, also be the case on election night. As results are coming in from key swing states, there will be a moment when the candidates’ odds reset either to 0 per cent or 100 per cent. (It’s probably best not to expect a repeat of the 2020 election, when bookies including Predictit left “next president” betting open after Joe Biden had been declared the winner.) The returns for a bettor with a strong hunch will be the same (and in a much shorter period), but there will be no reason left to wager against their true beliefs.

At that point, we can probably start taking the odds from these markets a little more seriously.