Jason Carroll of Hudson River Trading, a member of the Modern Markets Initiative (MMI), has framed out some thoughts explaining how professional high frequency trading firms do in fact take on and manage risk in today’s modern markets:
One of the frequent criticisms of HFT is that HFT firms don’t take any risk. The most frequently cited example is Virtu’s statement that in 1,238 trading days it was profitable in all but one. I’d like to explain why this isn’t as preposterous as it sounds.
A traditional trader or market maker from a decade ago was not automated. He or she traded by walking into a pit or up to a trading post or possibly by clicking a key by hand as he sat at a NASDAQ Workstation waiting for someone to try to trade with him. Frequently he traded a few products or symbols. The number of trades he made in a day probably measured in the range of 10-100, and his success rate on those trades was probably 60-70%.
He probably worked for a firm that, in aggregate, employed dozens to hundreds of traders covering a broader array of products. In addition to covering many more products, these traders also probably covered an array of trading ideas — from shorter-term market making activity to longer-term inter-product arbitrage strategies. Any individual trader’s daily Profit or Loss (“PNL”) was probably far more volatile than the firm’s aggregate PNL, because when you sum diverse and independent sources of PNL, you end up with a smoother aggregate total.
Many automated trading firms (or HFTs) conduct very similar activity, except they do more of it, for smaller rewards, using computers. An HFT firm may manage to trade 100s to 1000s of times a day in 1000s of products and symbols in different asset classes and markets around the world. Our success rate is closer to 51-52% than 60-70%, and the amount we stand to make or lose in any single trade is also smaller than what a manual trader faced.
Consider these two contrived companies:
Firm 1 employs 10 human traders, and each human trades 10 times a day. Each trader is right 60% of the time, and each trade makes or loses $500. Given that data, Firm 1 will trade on average 100 times a day, and make $10,000 per day.
Firm 2 employs 10 people who develop automated trading strategies, and each employee is responsible for 10,000 trades a day made by those strategies. The strategies are right only 50.5% of the time, and they stand to make or lose $10 per trade. Firm 2 will trade on average 100,000 times a day, but will still only make $10,000 per day.
For an individual human trader described at Firm 1 to lose money, he’d have to be wrong in more trades he was right. So he’d have to be wrong in 6 trades and right in 4 trades. Statistics tells us the chance of losing at least 6 trades out of 10 when your win percentage is 60% is only 16%. So about 1 in every 6 days, each human trader will lose money.
Here’s a calculator that you can use to check my math: http://stattrek.com/online-calculator/binomial.aspx. For a deeper explanation of the statistics behind the math, I’d recommend reading some of the links on that page.1
For Firm 1 overall to lose money on a given day, it would have to be wrong in at least 51 trades. The chance of losing in 51 trades out of 100 when you’re right 60% of the time is even lower than 16% — it’s 1.7%. That means Firm 1 will lose money about 1 in every 60 days, or about 4 losing days a year assuming the firm doesn’t trade on weekends and holidays.
For Firm 2 to lose money on a given day, it would have to be wrong in at least 50,001 trades. The chance of being wrong in 50,001 trades or more out of 100,000 trades, when you expect to be right 50.5% of the time, is very, very small — it’s .077% or approximately 1 in 1292. So that means that this firm will lose money in one day every 5 years.
Is Firm 2 taking risk? On any given trade it’s taking on a lot of risk. It has a far smaller chance of success on each trade than the human traders do. But because Firm 2 makes so many trades, it achieves a much more stable PNL profile than the human trader.
In general, investors should appreciate that automation has allowed firms to compete more aggressively to participate as market intermediaries. After all, the better such firms are able to manage risk, the smaller the reward is they need to capture in order to viably provide short-term liquidity.
 For the stats geeks among you, I acknowledge that this analysis relies on the assumption that the outcome of all of the trades is determined independently. That’s probably not the case either at Firm 1 or Firm 2. Weakening that assumption may change the scale of the comparison, the effect I demonstrate from making a much larger # of trades still exists.
For more information on this topic, please also read a previous post from the MMI entitled “No Trading is Without Risk.”