Trillium Fine Nothing to do with High Frequency Trading (HFT)

On Monday, September 13th the Financial Industry Regulatory Authority (FINRA) announced sanctions against Trillium Brokerage Services, LLC and several of its employees totaling $2.26 million in total fines. (News Release) Trillium has been fined $1 million with another $1.26 million levied against nine traders, the Director of Trading and the Chief Compliance Officer.

Everyone is mis-interpreting this case as a blow to HFT, it's not

The interesting piece of this sanction is how it has been portrayed by FINRA and some prominent bloggers out there, particularly Zero Hedge, as a fine against a particular HFT strategy. FINRA's news release states that the fine is for the use of "an illicit high frequency trading strategy and related supervisory failures". But, upon reviewing the facts of the case, these were not high frequency traders.

Perhaps the crux of the problem is the lack of a standardized definition for what HFT entails. At the very least, any definition of HFT should include the concept of automated strategies. HFT strategies are programmed ahead of time and executed by blackboxes, or computers. This case does not involve HFT but typically very astute financial bloggers are mistaking it as the first major blow to HFT. Well, for better or likely worse, it's not.

Trillium traders were manual, proprietary traders gaming stocks by layering the bid or offer with orders they never wanted to have filled so they could garner fills on orders on the other side of the book. The traders were creating the appearance of buying or selling interest for price improvement on their orders.

Let's create a hypothetical example: say a trader wants to have 500 shares filled on the bid at $10.05. Using the strategy in question, he puts in his limit order bid for 500 shares at $10.05. Then he begins offering relatively large orders on the offer to induce other market players to hit his bid at $10.05. Say the inside offer is $10.07, this trader comes in at $10.09 showing the market his intention to sell 10,000 shares. The other colluding traders join the offer by offering stock at $10.10, $10.11 and so on to create the appearance of significant selling pressure.

Traders outside this ring watching the stock immediately react by hitting $10.05 to exit their positions or to enter a short in the stock. The Trillium trader in question receives his fill for 500 shares and then cancels his 10,000 share offer. Subsequently, he repeats the entire process on the other side of the book to attempt to exit his position at higher prices.

Nowhere in any of this process was there computerized trading employed. On the contrary, the manual trader is actually trying to game HFT! As the action states these traders engaged in this strategy "to take advantage of trading, including algorithmic trading by other firms". High frequency traders (computers) are nearly always the first to react to changes in the order book. Many other traders react as well, worried that the 10,000 share offer will push the price down. Yet, in this case, and in the case of what happens over and over all day long in stocks across the board, the offer is "fake" -- there is no intention to have it filled by the trader entering it.

FINRA charges that Trillium traders used this strategy 46,000 times over time, likely over weeks and weeks, not in a few hours as Barry Ritholtz incorrectly assumes (Note: I heart Rithotz but he's wrong here). Also, his solution would not solve this problem, namely mandating that all bids and offers last for 2 seconds. While this would raise the level of risk to the manipulating trader, it would not wholly stop this practice from happening.

Yet, some HFT do this all day long, just differently

Now, I am not saying that there aren't high frequency traders out there manipulating the tape. There most certainly are. What I would say though, is that it is much, much harder to prove. What makes this case easy to prove is two things: 1) it involves a colluding group of nine different traders, and 2) their orders were "often in substantial size relative to a stock's overall legitimate pending order volume".

A high frequency trading system is one entity and uses very small orders. The manipulation is to profit from very small spreads and, when manipulative, involves quote stuffing, not large quotations. The only way to battle this problem within the HFT arena, as Sean Hendelman and I have been arguing, is to tax order cancellations (or messaging traffic as Senator Kaufman argues, a very similar concept). As I'm finishing up this article I've just been IMed Felix Salmon's post today and he's exactly right.


Brandon R. Rowley
"Chance favors the prepared mind."

*DISCLOSURE: Nothing relevant.

Where is HFT Headed? World Research Group Summit

I spent the afternoon off the trading floor attending the Buy Side Tech: High Frequency Trading Summit put on by the World Research Group. Our CEO and Managing Partner of T3 Capital Management, Sean Hendelman, was featured on two panels discussing a variety of aspects of high frequency trading development and the future of HFT.

I arrived late after trading the morning with the panel discussion on "Assessing the Growth of High Frequency Trading in Futures, Options, and FX" in progress but heard a few interesting tidbits worth repeating. One panelist noted how HFT strategies based on speed now have so many competitors in US equity markets and options that major profit potential is becoming increasingly limited in terms of the latency game. The future of the industry lies primarily in trading across asset classes and across global markets. Global currencies offer significant arbitrage opportunities as the landscape is still quite fragmented with many different exchanges across the world. For example, there are only 4 fiber-optic lines running from London to New York and being on the right line with the highest speed is crucial and speed arb is still possible in that space.

Another highlight was the idea of the high frequency trading style becoming a strategy into which investors would want to put their money within a portfolio. Like the desire for exposure to various asset classes within a portfolio, it is feasible to think that in the future there could be a demand for an exchange-traded fund that invests in HFT strategies. There is certainly a natural attraction for portfolio managers to a non-economic, no-risk strategy.

The last interesting comment I heard was out of Mark S. Longo from theoptionsinsider.com who noted the distortive impact of non-economic volume particularly in the options market. While many market participants use "unusual" trading activity in options as an indicator of possible future movement, many arbitrage strategies in the HFT universe may be creating false signals. Dividend arb or fee arb are two examples that can cause a surge in volume that is entirely non-directional in strategy. He finished by mentioning that investors would be wise to look more toward open interest and other readings along with unusual trading volumes.

The next panel discussion was entitled "Build or Buy? Strategies for Determining the First Step in Implementation" and Sean was one of four participants with another very intriguing guest being Adam Afshar, President of Hyde Park Global Investments. Hyde Park creates robotic artificial intelligence programs designed to self-adjust their trading to optimize results, a fascinating concept to begin with.

The primary thrust of this discussion came from Sean and Afshar telling interested observers to focus on the strategy before considering the build or buy decision. Broad consensus between panelists seemed to be that buying was the best option when starting out because of the time to market and lack of expertise but this will likely grow into a hybrid operation over time as you demand higher control over your data and execution. By Afshar's estimates it would take a firm $5-10 million to fully setup low-latency execution in-house involving direct feeds, co-location, etc. Yet, Sean stressed how important it is to have a strategy that works first because despite seemingly common opinion that HFT shops simply setup and make money, it is actually extraordinarily difficult to find profitable strategies as the vast majority of strategies fail and the ones that do work can go out-of-favor very rapidly.

Next up was a half an hour presentation by Matt Samelson of Woodbine Associates titled "The Impact of High Frequency Strategies on Spreads and Volatility on Highly Liquid U.S. Equities". The presentation was a summary of the $3,750 "ground-breaking study" available from the firm with their basic argument being that spreads tightened in 2/3 of the 39 most liquid stocks throughout 2008-09 and therefore the "traditional" market participant is better off, not worse off, as HFT has grown as a share of trading volume. While this presentation purported to defend HFT against attacks, it accomplished nothing in terms of engaging in the contemporary debate. While T3 Capital runs HFT strategies and welcomes defense against much of the misinformation out there, this presentation was sorely lacking only working to regurgitate old arguments. It's as if he were a philosophy professor that taught Descartes' Meditations and simply didn't bother to acknowledge the circularity objection to the "I think, therefore I am" statement (even though I don't believe this is a crippling refutation but that's another discussion). The current debate has moved far beyond his presentation.

The anti-HFT crowd believes traditional market participants are being disadvantaged in illiquid securities, not liquid ones. And, I won't be one to join the stereotyping, most of the "anti-HFT" crowd are not anti-HFT per se. They are against strategies that they believe hurt the retail trader/investor and there's clearly merit to many of their thoughts on the problems with the current market structure. The primary issue within the liquid security universe is not the spread but the overall true trading cost as the ECN fee structure encourages an absurd level of liquidity provision in stocks not needing any. Samelson also explained a bizarre statistic of "realized spread" in which they measured where a stock was trading five minutes after an HFT trade and claimed that the majority of the time the stock had gone in favor of the counterparty to the trade. The use of five minutes is highly arbitrary and is completely irrelevant in the HFT world especially rebate traders who may have been in and out of the stock multiple times by the time five minutes has elapsed.

The second panel discussion with Sean was titled "The Drive for Zero Latency: Optimizing Existing Systems for High Frequency Trading Strategies" moderated by Jayaram Muthuswamy, Professor of Finance at Kent State University. Muthuswamy offered a stimulating academic perspective on the current race for low latency. While the concept of the limiting factor in latency ultimately reducing to the speed of light was mentioned, the discussion shifted towards the various areas of possible latency reduction. It is not simply the distance and speed of execution, it is also quote speeds coming in and the speed of interpretation, the complexity of the algorithm code and its decision-making speed, routing speeds, and regulatory hindrances among other things. Sean also pointed out the importance of low latency during times of market stress where inefficiencies are high and HFT can find substantial opportunity. These volatile times, like an FOMC announcement for example, act as a perfect test of the capabilities of an HFT system.

Ultimately, Muthuswamy finds HFT to be a source of incredible progress. He maintains this belief even while he mentioned an email from his friend Eugene Fama, often regarded as the father of the efficient market hypothesis, where Fama stated his beliefs on the growth in HFT in one line: "excessive HFT can be deleterious to market efficiency". Yet, even with the incredible growth, Muthuswamy believes we are only "scratching the surface" of what is possible particularly in statistical arbitrage between any and every market and asset class around the world. He hinted that in his speech tomorrow he will attempt to hypothesize what the ultra low latency game will be dependent on ultimately: the complexity of the underlying code.