* Berkeley Lab seeks early warning of future flash crashes
* Lab focuses on VPIN, HHI variant as key monitoring tools
* Lab research also precursor to consolidated audit trail
By Herbert Lash
NEW YORK, Nov 9 (Reuters) - A system that could be used to monitor the ultra-fast algorithmic trading that dominates Wall Street is taking shape at the U.S. government’s supercomputing complex in Berkeley, California.
Scientists at Lawrence Berkeley National Laboratory (LBL) believe a “yellow flag” alert akin to warnings used in auto races would be better in preventing catastrophic market turmoil like the flash crash of May 2010 than the circuit breakers now in place.
The scientists examined two metrics they believe hold promise for an early-warning system of unusual market conditions.
Their research also addresses desires for a “consolidated audit trail” that regulators have urged since the flash crash revealed the difficulty in monitoring high-frequency trading.
It took regulators almost five months to reconstruct what happened in a span of minutes on May 6, 2010 because of multiple marketplaces and diverse auditing systems. A single, integrated record of all U.S. equity trading activity should help regulators better understand and police markets.
Regulators should prevent disruptive incidents instead of reacting to them like aviation investigators, said David Leinweber, an expert in algorithmic trading who co-founded the Center for Innovative Financial Technology at LBL.
“After the crash you get the flight data recorder and try to figure out what happened,” said Leinweber. “Let’s manage the congestion, manage the flow and understand when things may be going wrong and try to avoid them.”
The LBL study was prompted after discussions with Gregg Berman, a senior Securities and Exchange Commission advisor who was a lead investigator for the flash crash, Leinweber said.
The center was started in 2010 to bridge supercomputing and financial markets, an area that has drawn little research.
“After the flash crash it became pretty clear that understanding markets was more of a data-intensive science than the federal regulators charged with understanding markets had the resources to deal with,” said Leinweber, a former portfolio manager at quantitative hedge fund First Quadrant, where he oversaw $6 billion in equity assets.
Regulators face issues tracking trade data, including incompatible systems and inadequate computing power.
The requirements of a consolidated audit trail and who will run it have not been decided. The Financial Industry Regulatory Authority, a self regulator of U.S. brokers, has said its OATS system for tracking trade information could be expanded to meet the SEC’s goals.
An LBL study found two indicators that exhibited early warning properties preceding the flash crash: A variant of the Herfindahl-Hirschman Index, a measure anti-trust authorities use to determine competition in an industry, and a metric known as VPIN.
VPIN was developed by Marcos Lopez de Prado, head of high-frequency futures trading at Tudor Investment Corp, and Cornell University professors Maureen O’Hara and David Easley.
The metric anticipates market imbalances hours before the flash crash and can identify short-term “toxicity-induced” volatility, a research paper by the three said.