Prediction

In the high-stakes arena of prediction markets, a counterintuitive pattern emerges: retail traders who correctly pick winners more than half the time still lose money, while automated traders with coin-flip accuracy pocket nine-figure profits. Using 222 million prediction market tradeswith directly observable terminal payoffs, the paper “Who Profits from Prediction? Execution, Not Information” presents a clean answer to why it is so. The authors decompose trader returns into a directional component and an execution component, revealing that the execution component, not the directional component, determines which trader types earn positive returns. 

This study examines the profitability of mean-reversion trading strategies applied to binary outcome contracts on Polymarket, the world’s largest decentralized prediction market platform. We analyze three distinct contracts representing varying risk profiles: a quasi-risk-free instrument (No to “Will Jesus Christ return in 2025?”) and two high-yield speculative contracts (No to “Will China invade Taiwan in 2025?” and “Will the US confirm that aliens exist in 2025?”). Using high-frequency price data sampled at 10-minute intervals over approximately one year, we implement a parameterized mean-reversion framework across twelve strategy variants, testing robustness under varying liquidity constraints and transaction cost assumptions. Our findings reveal that while mean-reversion signals generate substantial alpha under passive limit-order execution (zero-spread scenario), strategy performance degrades significantly when more aggressive market orders are accounted for.

For many investors, short-term interest rates are often treated as something the market “discovers.” In reality, the Federal Reserve has enormous control over how the front end of the yield curve evolves. While textbooks often portray the Fed’s policy rate as a flexible tool that reacts quickly to economic data, the actual behavior of the Federal Open Market Committee (FOMC) looks very different. In practice, monetary policy tends to move in long, persistent cycles. The Fed spends years hiking rates, or years cutting them, and only rarely reverses direction quickly. For anyone trading rates, bonds, or rate-sensitive assets, this persistence matters. It means that the path of short-term interest rates over the next one to two years is often largely shaped by the Fed’s policy trajectory rather than by constantly shifting market expectations.

This observation has an important implication: the short end of the Treasury curve often behaves less like a forecasting market and more like a gradual reflection of the Fed’s policy cycle. When the Fed enters a tightening or easing phase, that trend tends to propagate through Treasury yields from one month out to roughly two years. In this article, we show that these policy-driven trends can be measured and used. By identifying whether the Fed is in a tightening, easing, or neutral phase, investors can improve their expectations about the near-term evolution of the yield curve. For fixed-income portfolio managers and macro traders, recognizing these policy regimes can help sharpen rate forecasts, improve duration positioning, and better manage risks tied to interest-rate movements.

Everyone is excited about the potential of large language models (LLMs) to assist with forecasting, research, and countless day-to-day tasks. However, as their use expands into sensitive areas like financial prediction, serious concerns are emerging—particularly around memory leaks. In the recent paper “The Memorization Problem: Can We Trust LLMs’ Economic Forecasts?”, the authors highlight a key issue: when LLMs are tested on historical data within their training window, their high accuracy may not reflect real forecasting ability, but rather memorization of past outcomes. This undermines the reliability of backtests and creates a false sense of predictive power.