High Frequency Trading

Posted on November 2, 2012 by dlaidlaw

On August 1st, Knight Capital’s trading software malfunction caused wild swings in the quoted prices for about 140 common stocks. Since the initial story broke, it has been very difficult to understand exactly what happened since neither Knight itself nor the regulators and exchanges have released any public statements or reports detailing the specific problems.

 

Blooomberg reported on August 14th that the error was caused by a dormant software system that multiplied trade orders by 1,000 times the desired amount. If this were the case, an order to buy 1,000 shares of a company would be transformed into an order to purchase 1,000,000 shares of that stock. A few Internet blogs speculated that the erroneous trading was caused by the installation of testing software that began routing large orders

over the NYSE. Regardless, the unintended trading persisted for between half an hour and 45 minutes.

During that time, Knight purchased so much stock that negative price movements would have depleted Knight’s capital. Knight sold this massive position to Goldman Sachs for a discount to the value of the stocks given the size of the block. Knight suffered a net loss of over $400 million on this transaction and ended up selling itself to a group of investment firms a week after the problem trades.

 

Knight Capital’s trading problem was not an isolated event. During Facebook’s initial public offering earlier this year, trading in Facebook’s shares was halted throughout the day as the NASDAQ exchange could not process orders. Another example is the “Flash Crash” in May of 2010 which caused the market to lose almost 10% during the day.

The large volume of equity trading has lowered the costs dramatically for all investors. Trades that cost hundreds of dollars in the 1980s, now cost $10 or less. These lower transactional costs have benefited all investors.

 

However, the increase in the number of trading venues and speed at which these trades are executed have produced these recent problems. There are now as many as 50 trading venues including formal stock exchanges such as the New York Stock Exchange and numerous “dark pools.” Dark pools allow purchases and sales away from the exchanges for either lower costs or greater anonymity. High Frequency Traders buy and sell stocks within

milliseconds to capture a very small profit on each transaction. The automation of the trading process has caused these spectacular meltdowns within the market.

 

The Securities and Exchange Commission is convening high frequency traders to explore solutions that will keep the markets liquid (and inexpensive) without subjecting trading to manic swings. The most talked about solution involves requiring every trading company to have a “kill switch” that allows it to immediately cease all trading activity with the appearance of problems. We are certain that better market trading structures and regulations could reduce trading problems.

 

Regardless of potential fixes, short-term trading problems such as Knight’s experience do not threaten the integrity of the portfolios under our management. Stock prices within the portfolios may change price dramatically during the trading day. However, the problems presented by these trading glitches are usually resolved within minutes and hours. The impact of the flash crash lasted longer, but the market reached equilibrium within two trading days of the crash on May 6, 2010.

 

Even if we were trading a security that was subject to order issues, we would most likely be able to determine that trading was irrational and that we should not transact that particular trade. We execute our trades in individual blocks. There is no software algorithm that dictates our trades. Ben Connard places the majority of our trades electronically and monitors them to make sure that we are executing for an appropriate price. For instance, assume that we wanted to exit a stock that closed the prior day at $30/share. If we saw bid levels plummet to $20/share within minutes, we would not execute the trade. The only circumstances under which we would sell such a position is if we determined the reason for the crash was fundamental and that we wanted to exit the holding immediately. Similarly, we would not buy a stock that was spiking up above our purchase target.

 

Changes to market structure are coming since it is necessary to restore confidence in the trading platforms. However, the current issues are not a threat to the value of the stocks under our management since we are employing an intentionally low frequency trading style.