Paper Panics · Chapter 12

The Hot-Potato Algorithm

How a Volume-Pegged Sell Order Erased a Trillion Dollars in Thirty-Six Minutes

A single line of execution parameters recorded on the joint CFTC-SEC audit trail from Thursday, May 6, 2010, captures how modern electronic liquidity can evaporate in minutes. At 2:32:00 p.m. Eastern Time, European sovereign debt fears over Greece were already weighing on the market. In Overland Park, Kansas, a mutual fund complex named Waddell & Reed Financial initiated an automated program to hedge its equity portfolio. It entered an order to sell 75,000 June 2010 E-Mini S&P 500 futures contracts worth roughly $4.1 billion on the Chicago Mercantile Exchange's Globex electronic platform. Previously, executing a $4 billion hedge took more than five hours of careful slicing. That afternoon, the firm selected an off-the-shelf broker algorithm programmed with a single mechanical rule: as the official regulatory report later documented, the software was instructed to target nine percent of the trading volume calculated over the previous minute without regard to price or time.

Targeting a fixed percentage of rolling one-minute trading volume without a minimum price floor or a maximum execution speed created a closed positive feedback loop the moment it met high-frequency market makers. Between 2:32 p.m. and 3:08 p.m., the interaction of those two automated systems erased nearly one trillion dollars of market value across four distinct mechanical phases:

  1. 1. Initial liquidity absorption (2:32 p.m. to 2:41 p.m.): During the first nine minutes, the Waddell & Reed sell algorithm dumped 35,000 E-Mini contracts into the Globex order book. High-frequency trading (HFT) firms, traditional buyers, and cross-market arbitrageurs absorbed those initial contracts. The selling pushed S&P 500 futures down three percent while exhausting the top layers of resting buy orders.
  2. 2. Inventory saturation and hot-potato churn (2:41 p.m. to 2:45:27 p.m.): Once high-frequency market makers hit their strict intraday position limits of 2,000 to 3,000 contracts, their algorithms immediately flipped from net buyers to aggressive sellers to dump inventory before prices fell further. They sold contracts to other HFT algorithms, which instantly resold them back. The Waddell & Reed algorithm measured total traded volume over the prior sixty seconds without knowing that HFTs were merely passing the same contracts back and forth. It interpreted the churn as a massive surge in market liquidity and ramped up its own selling speed to match nine percent of the artificial volume spike.
  3. 3. Cross-market contagion and stub quotes (2:45 p.m. to 2:47 p.m.): Arbitrage computers buying cheap E-Mini futures in Chicago simultaneously fired rapid-fire sell orders into the S&P 500 ETF (SPY) and individual equities in New York. Sensing abnormal velocity, automated equity market makers switched off their quoting engines to avoid toxic order flow. With real bids withdrawn, sweeping market orders hit placeholder "stub quotes" posted at a single penny or $100,000 to satisfy nominal exchange rules. Blue-chip shares of Accenture and Procter & Gamble plunged from $40 to $0.01 in seconds while the Dow Jones Industrial Average dropped 998.5 points.
  4. 4. The five-second circuit breaker (2:45:28 p.m. to 3:08 p.m.): At 2:45:28 p.m., after the E-Mini had plunged five percent in four minutes and finished digesting all 75,000 Waddell & Reed contracts in just twenty minutes, CME Globex's automated Stop Logic Functionality triggered a five-second trading pause.

Those five seconds of forced silence broke the volume feedback loop and reset the rolling one-minute volume counter to zero. Human and algorithmic desks reposted real bids when trading resumed at 2:45:33 p.m. During a single fourteen-second window at 2:45 p.m., high-frequency trading firms passed twenty-seven thousand futures contracts back and forth while changing their net inventory by fewer than two hundred contracts.