This Is the End


Markets, Risk and Human Interaction

July 17, 2009

Goldman and High Frequency Trading

There is a funny attempt at logic going around in trying to explain Goldman’s high earnings:
1. Someone has been charged with stealing code from Goldman, code used for high frequency trading.
2. Goldman made tons of money last quarter.
3. Therefore, Goldman made its money from high frequency trading.

This argument not withstanding, I doubt that Goldman is making much of its money from high frequency trading. For one thing, high frequency trading does not have a lot of capacity. For another, why bother with high frequency trading, an area where there are relatively low barriers to entry and where you have no particular comparative advantage, when you have a screaming money-making franchise that is close to unassailable?

I even wonder if the code really was for a high frequency trading operation. Prosecutors want to paint the most extreme picture possible. So if you listen to them, you will come away thinking the code not only made untold millions for the firm, but if put in the wrong hands it could destroy the Western world. And the person accused of the theft, Sergey Aleynikov, would have had an interest in exaggerating the value of his work to potential future employers – at least before he was apprehended. Granted it might have been valuable for a high frequency shop, but it is more likely is that this code was one component of a broader trading operation, a way to efficiently execute trades, to add value to other systems. We do know, for example, that Goldman – like others – has substantial infrastructure for automated trade execution algorithms as a part of its market making and brokerage business.

More interesting than what this alleged code theft might tell us about how Goldman made money is that it highlights that we have no clue how the firm really did make money. And, more to the point, that the regulators have no clue.

Imagine if early into the current crisis the New York Fed’s Division of Bank Supervision had performed a routine analysis to see where the banks’ profits were coming from. That question would have led to the burgeoning structured products markets. The next questions would have been – or at least should have been – whether those profits came from cutting corners in terms of risk or compliance. Or, whether the scent of yet larger profits might lead to future corner cutting. I gather that such an exercise never took place. (I also wonder why there hasn’t been more finger pointing toward the Division of Bank Supervision, but that is a different matter).

Well, we missed on that one. But maybe it is worth learning from the past and begin asking those questions of the banks as part of the supervisory process. Banks have plenty of ways to make money through questionable means and through imprudent risk taking. Come to think of it, if we ever get to the point of having hedge fund regulation, maybe the regulators should ask the same sorts of questions there, too.

I am in the middle of writing a novel that begins in the midst of the 2008 crisis. In the novel there is an investment bank where one of the trading units gets requests from its clients to price their illiquid inventory. (This is an exercise that occurs in real life, because the clients have to mark to market, and for some assets there is no market. So they go out and get bids from a couple of banks, and then mark at the average of these two prices). This trader puts in incredibly low-ball prices. One bank prices a security at $92. He prices it at $50, leading to a mark to market price of $71. The trader knows that with such a low price, the client will be forced into liquidation mode. The trader positions his book for the forced sale that he helped precipitate, generating big profits from his scheme. This is fiction. But if we have learned one thing over the past couple of years, in the world of finance truth can be stranger than fiction.


  1. "For another, why bother with high frequency trading, an area where there are relatively low barriers to entry and where you have no particular comparative advantage,"

    First, there are not low barriers to entry with HFT when you are talking about trading huge lots every 50ms. This requires a cutting-edge server colocated with the exchange's server running advanced code. Second, they do have a competitive advantage, as it's pretty clear they're frontrunning. (Take a look at their massive volume HF SPY trades immediately before the S&P spikes, and tell me that's not frontrunning.)

  2. If a firm is front running, that is not part of what I would call HFT, that is something that comes out of their franchise. The franchise is giving them the information to be able to front run. They might use high frequency methods to execute based on it, but that is a different matter.

    And granted two guys in a garage can't do it, but the infrastructure investment for HFT is one many firms can readily make. A lot of what is necessary in terms of infrastructure is available from vendors.

  3. Hi Rick,

    Please excuse my ignorance as I know little about finance/trading, or IT for that matter, but perhaps you can answer a question for me.

    As I understand it, today many financial institutions seem to be little more than data refineries. Over 90% of money exists in the form of data and trillions get 'pumped' around the the globe every year. Some companies pump the data faster and/or can 'read' it more quickly, so they steal a march on competitors (with inferior infrastructure/software) by trading automatically - technology enables them to be 'faster' and more 'precise' and take the best price.

    I've often wondered to myself 'is it the IT guys/companies that make the big bonuses'....for example, if programs are automatically trading to maximise value, or low latency enables trading ahead of rivals, what skills do traditional traders have that require huge bonuses? Isn't much of Goldman Sachs profitability coming from tech capability as opposed to any financial 'skill'?

  4. Very timely article. You are right that the barriers to entry in the HFT game are not that high.

    You don't need Goldman's connections to do millisecond level trading. With very small amount of capital, someone with intermediate or high programming skills, moderate knowledge of finance and financial systems can build a basic trading system (I know from personal experience).

    I don't quite understand the charge that Goldman was intercepting trades and trading ahead of them. As far as I know, Goldman is a market maker and could use client order flow to front-run -- very illegal.

    I'll be surprised if they were stupid enough to place their programs between exchanges and their customers...assuming that it is even possible (very, VERY, unlikely). If they did intercept such orders, how could they then generate their orders and transmit them to the exchange before the order they intercepted hit the exchange.

    There are just a whole set of things that don't make any sense.

  5. Hi Rick,

    I thought A Demon of Our Own Design was one of best books I've ever read. As a close market watcher, one thing that stood out last fall was the 'lack of a crash' --- it was a crash but it was kind of a controlled ooooze more than a break. In your book or elsewhere, I would love to hear your take on the nature of what happened in Sep-Nov 2008 in your book --- from a mechanics/system perspective.

  6. Rick,

    What you described in your novel is a common occurrence in OTC derivatives. However, usually if you call up the outlier counter-party and kindly ask for a tradeable two sided market, they'll wise up. If not, get a couple more marks from other counter parties and threaten legal action...they will back down. Unless the majority of your counter-parties are marking at the same level (which implies your position is truly seriously submerged), the scenario you painted is not plausible.

    I suggest that you use the VIX future settlement in Nov or Dec 2008 (I forgot which one) to bring your point about market manipulation across. Someone was long many thousand contracts of 50 call options on the VIX future. On settlement date, the person manipulated the VIX by quoting pre-open on settlement day 1 cent bids and ridiculous asks on far out of the money SPX options that no one even bothers to make a market on, thereby boosting the VIX from the 40s to just above 50. The trader then pockets a princely sum from his 50 calls. Real life is indeed stranger than fiction ;)

  7. From William Cohan's House of Cards (pp. 336-7): "'... the marks are 99, 98, 97. A week later ... Goldman Sachs sent, by e-mail its April marks on our securities...'. As a counterparty to trades in the funds, Goldman was obligated to report its thinking.... 'Now there's a funny little procedure that the SEC imposes on you, which is that even if you get a late mark, you have to consider it,' [one Bear Stearns executive] said. 'Suddenly we get these marks. ... They give us these 50 60 prices. What we got from the other counterparties is 98. ... We have to repost our NAV. And now we go from minus 6 to minus 19 ... and that is game fucking over. by the way, the firm that sent us the 50 made a shit pot of money in 2007 shorting the fucking market.'"

    Truth, according to the Bear Stearns executive, is indeed stranger than fiction -- or at least on a parallel track with Rick's forthcoming fiction.

  8. Hi Rick,

    What you say in your article, that you do not believe Goldman are making their money from High Frequency trading, is correct.

    Goldman have no high frequency trading teams in London (they had a medium frequency team, but they all left in 2008) and they have no high frequency trading teams in New York. They do have a team in Asia, but the PnL they generate is nothing relative to GS's results for the last quarter.

    Few of the investment banks really have a big presence in HFT as of 2008, when the the banks were unable to pay bonusses to these profitable teams, and most of these teams relocated to hedge funds, started their own hedge funds, or moved to dedicated high-frequency trading firms such as GETCO, Optiver, Kyte Group etc.

    Again this is GS conspiracy theorists getting over-excited and making a load of unsubstantiated conjecture.

    The banks who have big high frequency prop trading teams are BNP Paribas, Societe Generale and Barclays Capital.

  9. In my previous comment I said, "I've often wondered to myself 'is it the IT guys/companies that make the big bonuses'"

    After reading this article

    if they aren't they certainly should be. It appears we are getting close to 90% of trades being done between computers. (Which is a bit worrying because it's pretty clear that finance does not know how it works. How confident is anyone that the right data is going to the right place at the right time?)

    Anyway, can anyone direct me to an "idiot's guide" so I can try and understand precisely what human traders do that requires they are paid huge bonuses?

  10. Regarding the previous comment on IT bonuses and HFT, people sometimes are getting confused between high frequency trading as a money making strategy versus algorithmic trading, which is intended to efficiently execute trades. The forner decides what is mispriced and buys or sells accordingly, the latter takes existing trading decisions and executes them to reduce transaction costs.

    Many firms provide their clients with algorithms for trading which are mechanistic, rule-based approaches. They might be useful for reducing transaction costs, but cannot be unleashed on their own to make money. They simply facilitate trades that would be done one way or another.


Note: Only a member of this blog may post a comment.