When Algorithms Feed on Each Other: The 2010 Flash Crash as Market Feedback Loop

The WH Trading case illustrates feedback loops at the firm level—one algorithm responding to its own signals until regulatory intervention breaks the cycle. But there’s a far more dangerous variant that we understand primarily through historical study, because we’ve managed to prevent it from recurring at scale despite massive growth in algorithmic trading. This variant emerges when multiple algorithms begin responding to each other’s responses, creating market-level feedback loops where the collective behavior becomes systematically unstable. The 2010 Flash Crash on May 6 remains the canonical example, and studying it is non-negotiable if you want to understand what can go wrong.

May 6, 2010: One Trillion Dollars in 36 Minutes

At 2:32 PM Eastern Time on May 6, 2010, an asset management firm called Waddell & Reed initiated a massive liquidation order. They sold 75,000 contracts of the E-Mini S&P 500 futures—approximately $4.1 billion notional value. This was a routine institutional order, the kind of thing that happens regularly in derivatives markets. Nothing about it was unusual except for scale, and scale alone shouldn’t crash markets. The Dow Jones trades trillions of dollars daily without disruption.