Back in the day, Richard Hanania had some decent prediction market takes. But recently, he seems to have turned on the very asset class he was once a proponent of, calling them a “giant scam.”
The full manifestation of his updated belief comes through in his essay from this previous weekend: “Prediction Markets Need Prestige Signals to Fulfill Their Mission.”
Richard’s argument boils down to the following:
Bots are taking advantage of retail traders and capturing most of the profit available on PM exchanges.
Actual forecasters are becoming less important in the market compared to high-frequency traders. This differs from the original vision for prediction markets back in the Manifold/PredictIt/Good Judgement days.
Also:
The largest prediction market exchanges, Polymarket and Kalshi, are providing a venue for insider trading and publishing false advertising.
Since the questions that actually matter (e.g., economics) won’t attract dumb money, there’s no opportunity for sharps to profit.
Other financial markets can replicate the utility of prediction markets.
This is a hodgepodge of various anti-PM arguments. While many of these arguments may be able to stand on their own, smashing them all together does not create a coherent thesis.
I want to focus specifically on #1 and #2 above — essentially the claim that the best forecasters are becoming obsolete in PMs because most of the profit is captured by bots exploiting the bad decisions of retail traders. He says:
A May analysis from the Wall Street Journal shows that the top 0.1% of traders on Polymarket make 67% of the profits. And if you sort traders by trade frequency, only the top 0.1% make money on average. Those at the top are not smart pundits using their forecasting skills to improve their reputations, but largely bots that have been trained to take advantage of everyone else.
Factually, he’s probably not wrong. A February paper outlines this (quote from Marginal Revolution):1
Retail traders correctly forecast asset price direction yet lose money. Using 222 million prediction market trades with observable terminal payoffs, we decompose returns into a directional component (did the trader pick the right side?) and an execution component (did the trader get a favorable price?). Traders with above-random accuracy earn negative returns because they arrive late and pay unfavorable prices; traders with near-random accuracy profit through superior execution. …automated traders pay 2.52 cents less per contract than casual traders, and this gap alone accounts for the sign of returns across trader types. Being right and making money are not the same thing.
So, yes, in PnL terms, making sure your trades are well-executed is more important than making sure they’re factually correct. That’s not to say some trades aren’t accurate and well-executed.
Hanania is implying that the main issue with recent PM exchanges is that the financial incentives don’t necessarily align to the prestige incentives.
What Hanania is critiquing is, essentially, the incentive structure of the very markets he was once a proponent of. These aftereffects were an entirely predictable and natural byproduct of their existence. Unlike many legitimate criticisms, this isn’t malpractice from the companies themselves.
Critique #1: Bot PnL share is relative, not absolute
Let’s assume Hanania is right, and the profit incentives are not aligning with the prestige/reputation incentives. If that were true, here’s what would happen:
Bots are making most of the money on prediction markets.
The best forecasters are no longer awarded with profit.
Forecasters exit the market.
Calibration collapses because nobody is actually trying to answer the question being asked by the market.
This has not happened yet. During the article, Hanania never disputes PM’s superb calibration.
Additional evidence to back up this claim comes from this April paper.2 It classifies a subset of specific wallets as “skilled winners.” They derive this by taking an account’s trades and randomly flip the direction of each one. If the wallet’s trades beat the randomness benchmark by p < 0.05, the trader is skilled.3
Their study shows the skilled traders capture 27% of the profits on PMs. That’s all flowing to genuine forecasters.
The fact that retail-exploiting bots would capture a majority of the profit was also entirely predictable. Think about this from a pre-Polymarket/Kalshi perspective. It’s not hard to figure out that this would happen naturally:
Most retail traders won’t be running HFT desks.
There’s money to be made by correcting stale prices (regardless of whether they were right or not).
That “slow money” that comes from most retail traders is essentially free money for a bot.
There is a lot of slow money, so it dwarfs the forecasting profits when you compare them. That doesn’t mean the forecasting profits are nonexistent or small.
All of this evidence points to:
Yes, the bots are making most of the profit.
But, Kalshi and Polymarket are both very liquid! There’s still enough money floating around for actual forecasters and pundits.
The mistake Hanania makes is thinking about forecasters’ profit in relative terms, not absolute terms.
Critique #2: This is irrelevant
The title of Hanania’s piece is “Prediction Markets Need Prestige Signals to Fulfill Their Mission,” implying that the ability for a forecaster to build up a track record on a prediction market is the primary goal of prediction markets. While that is a goal, I believe it’s a distant second compared to (1) the forecasts and data provided by the market itself and (2) the hedging venue the market provides. Hanania himself admits the utility:
I still believe that prediction markets are pretty well-calibrated. When, for example, I hear about an up-and-coming politician and want to see whether he has a chance in an election, I will check the prediction markets first. That is the single best source of information for many questions one might ask.
Recap
My two major critiques of Hanania’s claim are:
Forecasters still must be making money, otherwise the markets wouldn’t be calibrated (a claim Hanania never disputes).
Even if the reputation perspective hasn’t been as successful as early PM proponents had hoped, it doesn’t matter as much as the other benefits provided by PMs.
I invite Richard to respond!
This quote is from a March version of the paper. That exact text no longer exists in the updated version.
Don’t worry, this doesn’t include market makers. I’m not a cheater like Kalshi.