The Mechanics of a Real-World Arbitrage Algorithm

Implementing a real-life multi-asset, statistical arbitrage strategy.

Even after accounting for fees and slippage, statistical arbitrage and pairs trading still remains a profitable niche.

To get started, let’s take a quick look at what the big picture of statistical arbitrage is. Essentially, for securities that are fundamentally correlated (e.g., corn and wheat futures prices due to similar risks, yields, seasonality cycles, etc.), a profit can be made when one is overbought/oversold relative to the other.

You may have already heard of this strategy, but most that describe it implement it in the wrong way. What isn’t said in discussions of this strategy however, is that even if the securities have a fundamental correlation, the edge is only scalable if automated on a timeframe of seconds/minutes and is multi-asset to reduce the overall volatility of the strategy.

I learned this when I actually tried deploying the system in the real-world, but as you will see below, with a few tweaks, I was able to rig the strategy into uncorrelated and scalable profits.