IMC Prosperity 4 · Spring 2026
Building a trading strategy.Algorithmic Trading Strategies
I developed and backtested Python strategies for the IMC Prosperity trading competition.
- Tools
- Python & backtesting
- Qualification
- Top 16% of 18,803 teams
- Progress
- All five rounds completed
Five rounds
On a four-person team, we qualified in the top 16% and completed all five rounds of IMC Prosperity 4. The top-16% figure refers to qualification, not our final standing.
An event-driven trading bot
I built Python strategies with per-product modules, bounded rolling histories, persistent JSON state, and inventory-aware limit-order generation. Signals included rolling OLS, EMA, local z-scores, and persistent regime states.
Different products called for different behavior: trend carry, mean reversion, and inventory-skewed two-sided quotes were not interchangeable.
Options, flow, and spreads
Later work added Black–Scholes option valuation, cross-strike volatility fitting, decaying counterparty-flow features, and pair/basket spreads. For basket signals, executable bids, asks, and unmatched inventory mattered more than an attractive midpoint relationship.
What I checked
I iterated 20+ backtest variants, using fill prices, turnover, inventory paths, and product-level PnL to diagnose execution costs and model bias. Simulator results were useful for iteration, but did not establish consistent profitability or out-of-sample performance.
