Writing
Featured
Research · February 2026
Maker vs. Taker Returns in Prediction Markets: An Empirical Analysis
Analysis of 199,494 trades across 2,000 prediction markets reveals that makers earn positive excess returns at the vast majority of price levels, while takers face systematic adverse selection. We decompose returns by market category and find that structural advantages, not directional skill, drive maker profitability.
Alpha Beta
Research · February 2026
The Favorite-Longshot Bias in Cryptocurrency Prediction Markets
Cryptocurrency prediction markets exhibit a persistent favorite-longshot bias where extreme outcomes are systematically mispriced. Our Monte Carlo analysis of a SOL-focused FLB strategy shows a 7.66 Sharpe ratio and 194.4% ROI on capital with 100% probability of profitable year across 10,000 simulations.
Alpha Beta
Research · January 2026
Empirical Kelly Criterion for Prediction Market Position Sizing
Standard Kelly sizing overestimates optimal position sizes when applied to prediction markets due to edge estimation uncertainty. We derive an empirical Kelly adjustment using the coefficient of variation of edge estimates and show that half-Kelly sizing produces superior risk-adjusted returns.
Alpha Beta
8 articles
February 2026
Maker vs. Taker Returns in Prediction Markets: An Empirical Analysis
Analysis of 199,494 trades across 2,000 prediction markets reveals that makers earn positive excess returns at the vast majority of price levels, while takers face systematic adverse selection. We decompose returns by market category and find that structural advantages, not directional skill, drive maker profitability.
Alpha Beta
February 2026
The Favorite-Longshot Bias in Cryptocurrency Prediction Markets
Cryptocurrency prediction markets exhibit a persistent favorite-longshot bias where extreme outcomes are systematically mispriced. Our Monte Carlo analysis of a SOL-focused FLB strategy shows a 7.66 Sharpe ratio and 194.4% ROI on capital with 100% probability of profitable year across 10,000 simulations.
Alpha Beta
January 2026
Empirical Kelly Criterion for Prediction Market Position Sizing
Standard Kelly sizing overestimates optimal position sizes when applied to prediction markets due to edge estimation uncertainty. We derive an empirical Kelly adjustment using the coefficient of variation of edge estimates and show that half-Kelly sizing produces superior risk-adjusted returns.
Alpha Beta
February 2026
The State of Prediction Markets in 2026
Prediction markets have evolved from niche curiosities to legitimate financial instruments. We examine the current landscape across Polymarket, Kalshi, and emerging platforms, analyzing liquidity depth, market efficiency, and the growing role of systematic traders.
Alpha Beta
January 2026
When AI Meets Market Making: Systematic Liquidity in Decentralized Markets
The convergence of AI-driven strategy optimization and on-chain market making creates a new paradigm for liquidity provision. We discuss how category-specific calibration and VPIN-based toxicity filtering enable adaptive market making across diverse prediction market verticals.
Alpha Beta
February 2026
The ZeroPoint Thesis: Bitcoin as Foundation for the AI Economy
Two singularities are converging: artificial intelligence driving cognitive costs to zero, and Bitcoin's discovery of absolute digital scarcity. We argue these forces are complementary and together form the basis for a new economic equilibrium we call ZeroPoint Capitalism.
Alpha Beta
February 2026
Announcing Polybot: Multi-Strategy Trading System for Prediction Markets
Today we are releasing Polybot, our systematic trading engine built for prediction markets. Polybot implements four distinct strategies across Polymarket and Kalshi, with empirically calibrated parameters, Kelly-based position sizing, and comprehensive risk management.
Alpha Beta
January 2026
Releasing Our Kalshi Trade Analysis Dataset
We are open-sourcing select outputs from our analysis of 199,494 Kalshi trades spanning 1,292 events and 2,000 markets. The dataset includes maker-taker return decompositions, category-level calibration metrics, and hourly volume distributions.
Alpha Beta