Easy profit opportunities that retail traders once enjoyed on prediction markets Polymarket and Kalshi are rapidly shrinking.

On July 22, blockchain media outlet CryptoSlate reported that competition on pricing is intensifying as institutional investors, market makers, quant managers, funded trading firms and artificial intelligence (AI) agents enter prediction markets in earnest.

The immediate battleground is the U.S. Federal Reserve’s policy rate decision. The Fed holds a meeting on July 28 to 29. In a Reuters survey of 104 economists on July 21, all respondents expected rates to be held at 3.50 to 3.75 percent, and the price of Kalshi’s July contract also reflects an 87 percent chance of a hold. The remaining 13 percent is the area market participants have to price.

Market makers, quant managers, funded trading firms and AI agents are facing off over that range. They track prices in real time, compare differences among related contracts and continuously adjust probabilities.

Institutional investors are testing event-based contracts, and brokers are connecting institutional clients with liquidity providers. Funded trading firms have begun to identify algorithmic traders who reflect uncertainty in prices more accurately than the market, based on data from contracts after settlement.

Louis Regis (루이 레지스), founder of on-chain funded trading firm Propr, judged that event-based contracts are better suited than traditional financial markets to assess traders’ skill. Because contracts settle based on clear outcomes, it becomes relatively clear who priced probabilities more accurately than the market, he said.

Still, a sizeable sample is needed to statistically prove a trader’s skill. A benchmark by Foresight Arena estimated that about 350 binary predictions are needed to confirm a real edge that is 2 percentage points above the market.

AI’s predictive ability has also yet to translate into stable profits. In an experiment that ran autonomous trading from Jan. 12 to March 9 by allocating $10,000 each to 6 latest AI models, the models posted losses of 16 percent to 30.8 percent on Kalshi. On Polymarket, the average return was limited to minus 1.1 percent.

The next contest is more likely to be decided by deviations from expectations and immediately after data releases than by the Fed announcement itself. The U.S. Bureau of Economic Analysis (BEA) releases an advance estimate of gross domestic product (GDP) on July 30, and a July jobs report is due on Aug. 7. Contract prices will be adjusted again with each new indicator, and the side that reflects unexpected figures first or corrects lagging prices the fastest will take control of the flow of trades.

This trend shows prediction markets moving beyond an experimental phase centered on retail participants to a market where professional liquidity and automated trading compete. The presence of AI agents alone has not guaranteed profitability, and the key is likely to be how quickly and accurately uncertainty is reflected in prices.