High-frequency trading, flash crashes & regulation Prof. Philip Treleaven Director, UCL Centre for Financial Computing UCL Professor of Computing www.financialcomputing.org p.treleaven@ucl.ac.uk
Normal Accidents quote from Charles Perrow With complex and little understood system characteristics: Multiple and unexpected interactions of failures are inevitable in complex systems. Unexpected and complex interactions between faults that are tolerable individually can lead to catastrophes. Tight coupling allowing little opportunity for mitigation or defence once a fault occurs. picture of Three Mile Island when it had a partial core meltdown in 1979 Chernobyl
Flash Crash May 6, 2010 1180 SPX 1160 1140 1120 1100 1080 1060 9:30 AM 9:49 AM 10:08 AM 10:28 AM 10:47 AM 11:07 AM 11:26 AM 11:45 AM 12:05 PM 12:24 PM 12:44 PM 1:03 PM 1:22 PM 1:42 PM The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again. 2:01 PM 2:20 PM 2:40 PM 2:59 PM 3:18 PM 3:38 PM 3:57 PM $600 billion in market value of US corporate stocks disappeared
High-frequency Algorithmic Trading Systems Algorithmic Trading Systems becoming increasing complex, and the market interactions NOT understood. Algorithm behaviour behaviour, interaction & risk of individual algorithms NOT understood. Algorithm ecology reduction in diversity potentially detrimental. Algorithms deployment systems deployed with minimal testing. Financial Regulation deployed without understanding of potential impact. Law of Unintended Consequences MiFID II may actually increase instability. Recommendations Need to Certify Algorithms before deployment Need to simulate Financial Regulation before deployment
Definitions Algorithmic Trading (AT) any form of trading using sophisticated algorithms to automate all or some part of the trade cycle. Systematic Trading similarly, any form of trading using sophisticated algorithms ; users refer to algorithmic trading just to trade execution; High-frequency Trading (HFT) execution of computerized trading strategies is characterized by extremely short position-holding periods involving trading speeds in excess of a few milliseconds. Ultra High-frequency Trading or low-latency trading refers to HFT execution in sub-millisecond times through co-location of servers at exchanges, direct market access, or individual data feeds offered by exchanges and others to minimize network and other types of latencies.
Algorithmic Trade Process what, when, how Pre-trade analysis analysis properties of asset using of market data or financial news. Trading signal identifies trading opportunities based on the pre-trade analysis (what to trade). Trade execution executing orders for the selected asset (when and how).
Algorithmic/Systematic trading Research Data (Real- (me/historical; market/non- market) Pre-trade Analysis Alpha Model Risk Model Transac5on Cost Model Trading Signal Por8olio Construc5on Model Trade Execution Execu5on Model Post-trade analysis
Alpha Trading Models (predicting the future of instruments) Quant Style Input Approach Trend Following Strategy Price- data Algorithms Theory- driven (hypothesizing the way markets behave) Mean Reversion Stat Arb Fundamental Yield Behaviour/ Sen(ment Growth Quality Empirical Data- driven (data mining to iden(fy behaviour) Real- (me Market Data Historical Non- market Data Implementa5on Issues Forecast target Time Horizon Bet Structure Investment Universe Model Specificatio n Run Frequenc y Data Availabilit y Regulation Compliance Linear Models Non-linear models Machine Learning
Risk Models eliminating or reducing exposure Risk Model Theory-driven (modelling systematic risk) Empirical (uses historical data to model exposure) Portfolio Exposure Risk Economic Risk Size Limits (constraints, Penalty) Volatility Single Position, Group Position Regime Change Risk Exogenous Shock Risk VaR CVaR
Flash Crash charts of quote-stuffing a high-frequency strategy that involves firing a massive number of quotes at a market and then quickly cancelling them
Flash Crash Causes Fat Finger in single-stock / index future Stop-loss Triggering If the market price falls through the stop loss trigger price, then the order will be activated and the long position will be automatically closed out. Inconsistent Trade Halting Rules Stub Quotes ultra-low bids that are placed when reserve size is depleted NYSE Delay Quote Stuffing The image cannot be displayed. Your computer may not have enough attempt to overwhelm a market with excessive memory to open the image, numbers or the image may have been corrupted. of quotes by Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again. traders. This involves placing and then almost immediately cancelling large numbers of rapid-fire orders to buy or sell stocks SEC Report http://www.sec.gov/news/studies/2010/marketeventsreport.pdf
Regulatory Changes Circuit breakers - based on the Dow Jones Industrial Average instituted for the market following 87 Crash Best price - Must quote within 30% of best price Trading Pauses - for single stocks that drop 10% in 5 minute period Applies to all exchanges and derivatives
Dark Pools Dark pools: are crossing networks that provide liquidity that is not displayed on order books. This is useful for traders who wish to move large numbers of shares without revealing themselves to the open market. Dark liquidity pools offer institutional investors many of the efficiencies associated with trading on the exchanges' public limit order books but without showing their hands to others. Dark liquidity pools avoid this risk because neither the price nor the identity of the trading company is displayed. Gaming: Manipulating the price of a stock to increase profits at the expense of the investor on the other side of the order in a dark pool. Common anti-gaming features offered by dark pools include setting minimum order sizes and prequalifying participants that meet a certain profile. Ping: The most common way to glean information about an order in a dark pool. Often considered a type of gaming, an investor submits a small order to a dark pool to gauge liquidity. After determining there is liquidity, the pinger will drive the price of the stock up or down on the public exchange by buying or selling a few shares in the market. The pinger will then return to the pool to execute at the manipulated price. The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.
Resilient Design Classic Principles Diversity greater diversity of algorithms increases resilience. Redundancy reduces efficiency but increases resilience. Modularity & Independency independence of components reduces system failure. Feedback Sensitivity increases a system s ability to detect & respond. Capacity for Adaptation to changing circumstances. Environmental Responsiveness & Integration reduces the probability of structural failure
Recommendations Financial Regulation lessons from engineering! Banking System increasing complex & inefficient; arguably not fit for purpose Regulation - is increasing system complexity and the chances of catastrophic failure Algorithmic Trading Engineering use engineering principles for resilience; greater diversity of algorithms increases resilience Science use computational science to evaluate algorithms Recommendations Regulator changes Dark Pools?, and discourage gaming, pinging etc. Authority to Certify Algorithms before deployment Government/Regulators need to simulate Financial Regulation before deployment