Inspiration
Prediction markets should reflect information instantly, but we saw that Polymarket lags behind faster exchanges like Binance on short timeframes. This delay creates temporary mispricing. We wanted to measure that lag and turn it into a consistent trading edge.
What it does
MALTA is an algorithmic trading system for Polymarket BTC, ETH, and SOL markets. It detects fast price moves on Binance, estimates the true probability of outcomes, and trades before Polymarket updates. ETH and SOL use a lead lag z score signal with Kelly sizing, while BTC uses a volatility based fair value model. The system only trades when signals are strong and pricing is favorable.
How we built it
We built on the provided backtester and iterated through several versions. Early rule based and ML models were inconsistent. The best results came from simplifying to a clean statistical signal with disciplined sizing. We tuned parameters per asset and built a hybrid system where each asset uses the model that fits it best. We also created analysis tools to study lead lag behavior, signal strength, and correlations.
Challenges we ran into
The edge is not stable and changes over time, which can lead to losses in weak periods. Position limits also reduced the effectiveness of optimal sizing. Another challenge was avoiding overfitting since short validation windows made some strategies look better than they actually were.
Accomplishments that we're proud of
We built a fully working hybrid trading system that consistently finds and trades inefficiencies. We identified and validated a real lead lag relationship between Binance and Polymarket. We also developed a strategy that balances performance with risk control instead of overfitting to short term gains.
What we learned
Simple, well understood signals can outperform complex models. Risk management is just as important as finding edge. Different assets behave differently, so one model does not fit all. Reacting to inefficiencies works better than trying to predict the market.
What's next for MALTA - Multi-asset Lead-lag Trading Algorithm
We plan to add regime detection to turn the strategy off during weak periods, improve execution to handle liquidity and slippage, and expand to more assets and markets. We also want to refine probability estimation and explore real time deployment.
Log in or sign up for Devpost to join the conversation.