1. In complex systems “more” is different
A single share’s transaction & quote volume is now measured in thousands of bid/ask quotes per second. To visualize this, imagine that a share now has the equivalent of 10,000 people offering bid/ask quotes every 6 seconds all day long vs 10-20 people a few years ago. Change the rate, number or nature of components in a system and th systems behavioral rules change. Nanex has pictures to help to visualize this micro-second world of market makers Nanex
2. Positive feedback loops grow exponentially
A microphone amplifying the loudspeaker leading to high pitch screech is an example of an audio positive feedback loop.
The flash crash occurred when one exchanges system clock got overloaded and was slightly mis-aligning bid/ask times stamps. The HFT traders interpreted the mis-stamped quotes as an arbitrage opportunity sending more messages and activity into the system further mis-aligning the time stamp and increasing the perceived opportunity. see Nanex report
3. Normal accidents are caused by common errors
in accident and systems theory terms the flash crash was a “normal” accident
. The timing mismatch wasn’t a large error in and of itself, but because of the HFT equity trading system structure (tightly coupled
& brittle) the normal error caused a large accident.
Summary & Flash Crash 2.0
The poorly understood system dynamics
of HFT mean that another Flash Crash is likely to be caused by a previously innocuous error. Greater risks may be associated with the sovereign debt markets and VaR related feedback loops
as High Frequency Trading expands its reach. The debate is one of speed (Liquidity) vs. safety (system integrity).
Explanatory pictures and video: