Original article by Felipe Montealegre

Original translation: Luffy, Foresight News

Once, a friend and I were discussing Robert Kennedy’s endorsement of Trump, and one participant confidently declared that Trump’s chances of winning had increased by 2% because Polymarket predicted so. This was a good observation because the event happened quickly and there wasn’t much other news to move the market. If Polymarket was an efficient market, this statement would seem to hold water.

But the problem is that Polymarket is still an inefficient emerging market that cannot predict small changes in the probability of an event (<5%).

The way efficient markets work is that large numbers of investors trade based on events. If you think Robert Kennedy's endorsement will increase Trump's odds by 10%, you will buy with leverage to get the expected return. If you think it will decrease Trump's odds, you will sell with leverage. The market is able to weight the average of well-thought-out forecasts to reflect the latest information.

The problem with Polymarket is that prices are affected too much for small deviations from consensus. In the market below, I give Trump a 50% chance of winning and Robert Kennedy’s endorsement bumps his odds up to 55%. The problem is that I can’t actually make $1 million on a trade to make $100,000 (assuming my 50% to 55% prediction is correct) because buying a $1 million bet would cause its average price to skyrocket by 62% and I would actually lose money if the market readjusts to 55%.

These issues can be solved with better liquidity and leveraging prediction tokens. Better liquidity would allow me to bet $1 million on Trump at 50% and exit at 55% with no slippage, thus making a 10% return. When a $1 million bet moves the market <5%, the market will be efficient within a <5% range.

On the other hand, using prediction tokens can greatly improve the efficiency of Polymarket. I think the 10% return on this kind of investment is low: you have to predict correctly and take about 10% risk on an emerging crypto platform. If you think you are right about 80% of the time, and if you are wrong, you lose 5%; then the final expected return is about 7%, which should be a good prediction relative to the market consensus. But the return is still too low.

Using Polymarket leverage, I was able to trade with 4x leverage and have enough left over to pay the lending platform’s 5% interest on the loan. This is a more appropriate return on capital for a researcher who spends energy and effort to find good trades and takes on risk.

In traditional markets we discuss 20bps interest rate moves because they have enough liquidity and leverage for analysts to earn high returns for correctly predicting small moves. Polymarket can discuss election odds of ~20%, or even 10%, but it cannot predict moves at the <5% level.

Counterarguments are that Polymarket is large enough to allow the wisdom of the crowd to work. The idea is that tens of thousands of small traders, including thousands in swing states, are constantly making small bets on the election using the latest information. These investors have biases and incomplete information, but the sources of bias are uncorrelated and cancel each other out, while the market aggregates many different sources of information.

The problem with this rebuttal is that Polymarket has about 4, 000 active traders betting on the US election every day, who are clustered in New York City and California and who disproportionately follow the same accounts on Twitter. Of these 4, 000 traders, I would expect half of them are not doing a deep investigation into each investment and are simply there to benefit from the airdrop “farming”, i.e. earning Polymarket tokens through trading, as Polymarket may reward active traders with airdrops. In addition, there is the issue of centralization, as 20 traders account for 95% of Polymarket’s trading volume.

Polymarket is an excellent business and source of information. I believe it is much better at predicting major political events than polls, expert models, and some analytical articles. Polymarket has demonstrated the ability to predict big moves before major media reports, but this does not mean that Polymarket is an accurate forecasting tool.

Original link