“This kind of market”

On Tuesday September 1, 2015, the New York Stock Echange invoked once again the Rule 48 for the four times in a week, and it may be assumed that September 1 (or at least the first minutes of trading) was, again, a new profitable moment for electronic markets makers (EMM) like Virtu or Optiver. While I’m not a specialist of market making, I have spent the past weeks reading academic papers (Chakraborty & Kearns 2011, Benos & Wetherilt 2012 or Menkveld 2013 for instance) about what I vaguely considered to be a “not-so-sexy-activity”: fueling the markets all day long, quoting both buy and sell prices for thousands of financial products around the world (I was trying to understand if and how market makers exploit statistical arbitrage, as “firms that engage in statistical arbitrage may indeed have strategies that have market making behaviors” [Chakraborty & Kearns 2011]). What happened on Monday August 24, 2015, when a storm coming from Asia hit the US markets (the Dow Jones shed over 1,000 points in early trading, the worst drop since October 2008), is perhaps a good opportunity to look at EMM’s activities (and profits) during such an extreme day.

For some observers, Virtu or Optiver are the parangon of the high-frequency trading firms who make money at other people’s expense (moms and pops, mutual funds, etc.) by playing some “latency arbitrage” games, thanks to technological edges (colocation, fast networks, etc.). I don’t want to debate “latency arbitrage” here as “arbitrage” between NBBO/SIP and direct feed prices only concerns the US exchanges, and Virtu operates on 229 different markets wordlwide. “We truly are a market making firm”, Virtu CEO Douglas Cifu says. “Bloggers want to pretend they know what we do but they are just making up conclusions. We aren’t the fastest and do not engage in latency arbitrage using a slow SIP.  That’s just nonsense.” He’s right on one point: a lot of other HFT firms are definitely faster than Virtu. Even if the firm owns microwave networks in the US (from New Jersey to Chicago, and from Washington DC to both NJ and Chicago), other networks (such as the McKay bros one) are definitively faster (so are McKay’s clients). In Europe, it seems that Virtu may have tried to build their own microwave network between Slough and Basildon, and between Slough and the future Hibernia landing station in Brean, but this was prehistory – they don’t have proprietary microwave networks in Europe, unlike Optiver, regarded as the fastest in this part of the world.

Not being the fastest didn’t prevent Virtu to have a good day on August 24, as we learnt in this Wall Street Journal article by Bradley Hope titled “Historic Profits for High-Frequency Trading Firm”. Although I understand that celebrating profits is quite natural, I was a little bit embarrassed with this statement made by Cifu: “Our firm is made for this kind of market.” The statement was quite equivocal, as if Virtu was celebrating the fact they made a lot of money at the expense of all the trading firms who lost money on this volatile day. I was not alone. I had a chat with the CEO of a hedge fund who summarized my views: “It is one thing for Doug Cifu to be proud of his firm’s performance on a tough day in the markets. It is a whole other can of worms for him to be trumpeting how much money he makes on the worst kind of day for every-day traders who have a very strong long-bias trading the markets. It is terrible PR for HFT and is the kind of thing that leads to politicians getting behind a Financial Transaction Tax (FTT). He is bragging about how much he loves trading days that feature most people getting losing lots of money”. Those words may be harsh but they come from a CEO “of a firm that had its best day on Monday [August 24]. So I have thought this through from a perspective that is somewhat similar”. “Best day” means a “a nine figures”. I’m not a trader and I don’t have a total assets under management of 2 billions, but I tend to agree with this comment.

The issue is not about the fact HFT (or non-HFT) firms made tons of money on August 24. The issue is that this kind of statement may discredit the whole HFT (or PTF) ecosystem, and high-frequency traders are already sufficiently criticized, they don’t need to be “trashed” more than they are (Virtu knows a lot about that, as they postponed their IPO after Michael Lewis’ Flash Boys release last year). That said, Cifu’s answers to CBNC on August 25 were far more interesting and detailed. The key here is to understand what is “this kind of market”. HFT firms are not monolithic: there are “HFT” firms employing a dozen of guys and doing statistical arbitrage in the US only with zero overnight positions; there are “HFT” firms with 30 people trading equities, options and currencies in the US, EU and Brazil (or elsewhere); and there are “HFT” firms like Optiver with 800 employees, making markets all around the world and holding €5.000.000.000 overnight non-directional positions (“We always aim to be delta flat”, says Willem Sprenkeler, Optiver’s Manager Corporate Affairs.) These “HFT” companies may share technical features (colocation and so on) but they don’t necessarily do the same work, and above all those “HFT” firms are not all “designated market makers”, formerly known as the “specialists” – the guys on the floor who lined their pockets thanks to very wide spreads (see the classic Christie and Schultz 1994 article to learn how some Nasdaq specialists avoided odd-eighth quotes before decimalization). 

The meaning of “this kind of market” may be found in a study made by Greg Laughlin last year, Insights Into High Frequency Trading From The Virtu Initial Public Offering (you can download here a slightly more updated version than the first version I gave to the Wall Street Journal). Laughlin looked at the documents associated with Virtu’s IPO filings to understand how the firm “had only one losing day in its six years of operation” (a revelation that was commented a lot). Beyond the profitability (Cifu once said that “51% to 52%” of Virtu’s trades are profitable, knowing that one profitable trade is a ~$0.01 profit per share), Laughlin estimated that Virtu trades ~160 millions shares per day (i.e. on normal days), “which put Virtu at ~3% of the US stock markets”. He compared the daily volume distribution in the SPY S&P 500 ETF with Virtu’s daily profit distribution. “Virtu’s profit histogram has a modestly sharper peak than the distribution of SPY trading volume, but to zeroth order, these are similar distributions. This supports the idea that Virtu is primarily a market maker with daily profit tied nearly linearly to total market volume”:

“Virtu’s profit histogram has a modestly sharper peak than the distribution of SPY trading volume, but to zeroth order, these are similar distributions. This supports the idea that Virtu is primarily a market maker with daily profit tied nearly linearly to total market volume.” On August 24, 2015, the volume of trades in the US was three times higher than during a “normal” day, meaning that Virtu may have traded ~500.000.000 shares on that day. If profit is correlated with volumes, so it was expected that August 24 should have been a good day for Virtu. A good reason to celebrate. “Total stock market volume on August 24 was ~15B shares, which is roughly three times the average value during the time period covered by the histogram” says Laughlin. “So that would imply a ~$7M profit (with asymmetrical error bars -3 and +10).”

“It’s likely that the spreads on the most heavily traded equities were greater than one penny for significant chunks of the volume (e.g. the first 30 minutes)”. Optiver’s Willem Sprenkeler confirms: “We had to widen the bid-ask spreads accordingly from time to time; due to the very high volatility that day, the risks for a market maker were also a lot higher. It will be no surprise that being a market maker we had a good result at the end of the day.” What’s more, a lot of exchange-traded funds (ETF) were trading far off of fair value based on the prices of the underlying securities. “That means there were likely plain vanilla arbitrage opportunities that a cool-headed, technically proficient market participant could have exploited. On the other hand, one would have to have had a very clear understanding of all the limit down rules for the various equities involved, so they may have deemed the ETF/underlying arbitrage as too risky when the opportunities were at their greatest. A market maker would face serious risk if they hedge a trade and then either the trade or the hedge is broken by the exchange after the fact”, says Laughlin. Cifu adds: “If we are making a market in an ETF and are lifted we will remove components to hedge our risk”.

Market making is a tough job. “The likely fact of the matter is that a lot of participants impulsively decided to pay a huge premium for immediacy when things were looking scary, and technically proficient market makers with a full view of the market were there and were willing to provide that immediacy to them. You never get a good deal on fire insurance when your house is on fire” concludes Laughlin. It appears there is more to explain EMM profits on “this kind of market”: volatility. “On a very volatile and high volume day our percentage of passivity will actually increase significantly”, says Cifu. “We make markets by placing passive bids and offers at many levels and only remove to hedge our risk.” The percentage of trades a market maker “make” (being passive) is correlated with a profit margin. “This is because any time that you cross the spread to get in or out of a trade, you are paying for the liquidity. In the extreme case where there is no price movement and the price bounces between bid and offer, the maker always earns one spread and the aggressor always loses one spread”, according to Laughlin. All of this is seems to be accurately described in the Chakraborty & Kearns 2011 paper:  “Perhaps the hallmark of market making is the willingness to always quote competitive buy and sell prices, but with the goal of minimizing directional risk. […] In this sense, pure market making strategies have no ‘view‘ or ‘opinion‘ on which direction the price ‘should‘ move — indeed, as we shall show, the most profitable scenario for a market maker is one in which there is virtually no overall directional movement of the stock, but rather a large amount of non-directional volatility.”

This is the “kind of market” Virtu CEO talked about: a huge volume as demand for liquidity is high, inducing a lot of trades with wider spreads than usual with the help of non-directional volatility (corollary: on very calm days, market making may be a boring activity, and not so profitable). The new electronic specialists are capable to make money on exchanges designed to avoid the “nefarious activities” of ultra-fast HFT firms: Virtu is the best customer of IEX and on August 24 the company made its biggest day (“by a long shot!”) on the platform (since this Monday was also the biggest day of IEX, that would be interesting to have numbers about the participation of EMMs on IEX on such a day). That’s the proof some market makers don’t need to be the fastest or to use “latency arbitrage” strategies to make a profit (as IEX was designed to be faster than all the – HFT or not HFT – trading firms). There is a legend about the first “floor specialist”: the guy had his legs broken, he was in a wheelchair, unable to move from a pit to another, so he decided to “specialize” in one product only, without moving from his favorite pit. Times have changed, and space has expanded: floors have nearly all disappeared and “electronic specialists” need to deal with hundreds of exchanges worldwide. Those new algorithmic specialists make much less money per trade than their old human predecessors, but ostensibly market making may be a profitable activity – at least on days when they have “this kind of market”.