Beware the Alpha Sharks! High-Frequency Trading and Its Impact on Markets
Maureen O’Hara, professor of finance at Cornell University, and member of the Systemic Risk Council has a warning to traditional investment managers, i.e., low frequency traders (LFT): High-frequency trading (HFT) is now the norm, and it is not going away. Continuing to pretend that HFT has no effect on your investments and on your alpha is at best naïve and at worst ignorant. Need proof?
New Laws, Including Breaking the Speed of Light
Both the Chicago Mercantile Exchange and NASDAQ are building a new faster link between their data centers in Chicago and New Jersey to “beat” the speed of light. Wait, I thought the speed of light was a universal maximum? It is, but you see by constructing two 60-foot towers the additional height travels a shorter distance than if the connection were at ground level due to the curvature of the earth. Doing this earns each transaction a faster execution time of — wait for it — 4 milliseconds!
Exchange attention is directly proportional to the amount of trades executed on their exchange. Tall towers will result in more orders and thus, more profits. So HFT is not going away, and this is why you have to understand it.
Importantly, as O’Hara pointed out at the 2013 Financial Analysts Seminar, “getting HFT” is not just about obsessing over speed, as the media so often does. No, she argues markets are forever different, yes, due to the speed of trading, but also the nature of market making, and even basic concepts like liquidity are dramatically different in an HFT world. O’Hara emphatically states that this does not make markets worse, just different than before.
One important advantage discussed infrequently is that HFT bases decisions on a large amount of information. This processing speed of Big Data is leading to the development of new fundamental information, and combined with the lightning-quick execution, you are at a permanent disadvantage. You might ask, “How weird could it be, really?”
Your Clock is Wrong
Since the dawn of time, humans have based their time measurements in chronology. That is, for people, time is a measurement system used to sequence events. But machines think in terms of cycles and are therefore event based, not time based. Machines complete a cycle at various rates depending on the amount of information involved in a particular instruction. So machines are volume based. Their world is measured in jobs done, not in the regular rhythms tapped out by time.
As it happens, HFT relies on machines, and thus it relies on measuring time in terms of events. This volume-time is challenging for us humans. But for a “silicon,” it is the natural way to process information. Why is this important?
If you use a volume clock as opposed to a time clock to measure returns — say look at trades every 1,000 shares — return distributions are much more well behaved. Goodbye fat tails. Goodbye the leptokurtic middle. Goodbye skew. Because HFT machines are looking at the world through a volume-clock lens, your actions are predictable; the distributions allow for extremely accurate predictions.
Ever lose money because a savvy trader anticipated your actions? Thought so. Now imagine an army of machines all looking at the world through their lens and programmed to take advantage of your naïveté. In other words, machines can see your clumsy human trades from hundreds of milliseconds away. Goodbye alpha!
Alpha Sharks
We are all swimming in silicon waters, whether we want to or not. Now there are “predatory algorithms” or “algos” looking to feed. These algos are a special type of informed trader. Rather than possessing exogenous information yet to be incorporated in the market price, they know that their endogenous actions are likely to trigger microstructure mechanisms in either your trader or your algorithms, and all with a highly predictable and foreseeable outcome.
Examples of these alpha sharks include:
- Quote stuffers: Here the algo overwhelms an exchange with messages with the sole intention of slowing down competing algos in order to gain a slight trading advantage.
- Quote danglers: These algos send dozens of quotes that force traders trying to complete an order quickly to chase a price against their interests.
- Pack hunters: These are algos designed to find other predator algos that are each hunting independently. Once these fellow algos are identified they form a pack in order to maximize the chances of triggering a cascading effect.
SHARK REPELLANT
So what can a low-frequency trader do?
- Stop putting in round lot orders as it signals to algos that you are a GUI (graphical user interface) trader.
- Stop pretending that you don’t need to understand HFTs and that it has no effect on you and your investors.
- Develop statistics (e.g., VPIN) to monitor HFTs activity and take advantage of their weaknesses.
- Use “smart brokers” that are specialized in searching for liquidity and avoiding a footprint. LFTs must become invisible to repel the alpha sharks.
- Do not target a participation rate. This is a trade order in which you instruct the broker to “only buy 1% of the volume today.” Instead, determine the optimal execution that fits the prevalent market conditions.
- Trade in exchanges that incorporate smart circuit breakers and matching engines.
O’Hara predicts a time in the very near future in which firms will employ “silicon analysts.” These experts will be highly proficient traders trained to understand how algos work and to help you get about your business without losing portfolio limbs to the alpha sharks.
This year, the 2015 Financial Analysts Seminar will be held in Chicago on 20–23 July. Learn more about the agenda and speaker details on our website.
Please note that the content of this site should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute.
Photo credit: ©iStockphoto.com/jangeltun
Great post Jason! Agree with all of the above. The full Volume Clock article is an extremely worthwhile read for anyone who transacts in equity markets!
Hi Michael,
Thank you for kind words. Thank you, too, for pointing readers toward the full Volume Clock article.
With smiles,
Jason
If HFT has so much impact on the mkt then why does the mkt trade off the same numbers I calculate each day….what a bunch of hoooey designed to rope in retail traders and keep them confused.
Hi Scott,
Thanks for taking the time to respond and to add your contribution to the discussion. You asked me separately via Twitter what is my source for this story. The woman that chaired the investigation for Congress and the CFTC into the “flash crash” who delivered the details of the report. Then she delivered what she believed actually happened, because it was her opinion that the politicized report diminished and obscured the actual effects of HFTs on the events. As a part of her presentation she had exact trade-level data with the millisecond timing of trades, including bids and asks, and she very aptly demonstrated the effects of HFT during the “flash crash.” Separately, during the same presentation she examined many other market anomalies taking place on a daily basis that can only be explained by people abusing trading networks through some of the abusive tactics I describe in this post.
This presentation was delivered at the Financial Analysts Seminar and in the audience were those that used to serve on the NYSE who were expert at understanding trading tactics designed to extract basis points here and there for gain, again through the tactics described here. Last, the FAS is Socratic format and in the audience were several ethical HFT firms themselves who also monitor market activity because they believe that bad apples can spoil the barrel for everyone.
What is the source of your skepticism?
Yours, in service,
Jason
I don’t care about any flash crashes….REPEAT–MARKETS STILL TRADE OFF THE SAME CALCULATED NUMBERS AND TIMES…so why do I care about whether someone else can execute a 4 millisecond trade between the cash open and the first turn calculated to be 50 mins later at 9:20 CST…. I traded it today w/o any problems…again what a bunch of…I gave those turn times to a 30 yr broker at Goldman Sachs this morning before the open and guess what?? The times still worked to within 5 mins…Will be glad to send you the email. Scott Craven
Hello Scott,
My original article was published in 2013, and there are many examples of “flash crashes.” Here is a link to one covered by MIT just last week: https://www.technologyreview.com/s/602586/algorithms-probably-caused-a-flash-crash-of-the-british-pound/?utm_campaign=internal&utm_medium=homepage&utm_source=top-stories_3&set=602580
Yours, in service,
Jason
Thanks for reading my post and responding, Jason. When you get the “flash crashes” as you call them it is because several cycles are ending at the same time so you get a quick move….It’s like Gann said, cycles are at the heart of market moves…you just have to know the lengths and the starting points…Every mkt move is tied to a previous high or low…period
Can somebody please mail/post the link for the full volume clock article. Sound like an interesting read.
Hello Pratik,
There is not just one “volume clock” article – there are many dozens. If you are interested in a deep dive on the subject I recommend going to a search engine and entering as your search term “volume clock trading.”
Barring that, here are several pieces that may interest you:
The Volume Clock: Insights into the High Frequency Paradigm: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2034858
Exploiting the volume clock: an agent-based model of high and low frequency trading: http://www.antoinegodin.eu/node/45
With smiles,
Jason
Thanks Jason
Please excuse my own naïveté, but do the three “alpha shark” practices above not sound like market manipulation? If not, why not? If they are, why are they permitted?
Kind regards,
Hello Stephen,
Yes, a very strong case could be made that they are market manipulation. Exchanges globally are contemplating ways of dealing with these issues but nothing concrete has been done, yet.
These trades happen so unimaginably quickly and in such gigantic volumes and across so many platforms globally that to unwind just one example of manipulation can take weeks, and yet it is happening every day according to some HFT watchers.
One problem with an exchange changing their rules to make the market manipulation described above more difficult is that the HFTs account for so much of the volume and profit of the exchanges. So if an exchange implements rules to make it more difficult for HFTs to trade then they will just go elsewhere, taking their volume and fees elsewhere.
Remember the context, too, that the HFT firms do not care about the underlying asset, just the movement of prices, so they trade globally and simultaneously. So to stop the regime shopping exchanges globally must agree across the board to changes. As you know, consensus is very difficult.
My guess is that as soon as non-HFT traders, GUI-traders as HFT firms call them, wake up to the full magnitude of what is happening – hopefully not after a crisis – that they will demand action on the part of the exchanges.
Thanks for your question,
Jason
I am struggling to understand the comment about “not” trading in round lot numbers.
1. If I trade larger than 100 then sometimes my brokerage breaks it up in round lots + remainder. So is the suggestion to trade x number of 99 shares? That would incur some hefty brokerage fees.
2. Why is this an issue for me. It is unclear from the article how I will be impacted negatively.
thanks
Hi Danny,
Great question. When you trade in round lot numbers (read: human-ish), like 10, 20, 25…100…150, and so forth you indicate to algorithms that you are a human trader; that you will trade slowly; and that you are most likely uninformed as to the way their algos will manipulate your price.
As an example, once the algorithms are in place they may quote dangle your brokerage’s trader. This is kind of like a situation in which there is one very desirable doe among some stags. When the doe does something flirtatious she pays attention to which stag emerges from the crowd to follow her. She then walks several paces away and sees if he follows. If he follows then she goes even further away. Each time making the stag’s interest more obvious. If the doe were a HFT then the price you pay can end up being much higher in order to consummate your trade.
A part of why I wrote the post is that most brokers do engage in the practice you describe. They will have to adapt. I wouldn’t be surprised that once the HFT issue is better understood by the popular press and the “little guys” that brokerages will begin competing for business based on their savvy execution of trades.
Good luck out there,
Jason
Kindly excuse my own naïveté, since I am a CFA level 1 student, given the fact that all these elements and concepts are new to me, Articles about HFT make me inquisitive and I possess a desire to practice HFT myself. I would want to know what are the prerequisites for doing HFT as an individual. I guess it requires huge money and exorbitantly priced machines. I tried searching the online engines about ‘how to be a high frequency trader’ but was welcomed with articles that explain what high frequency trading is and the pros and cons of it.
What level of knowledge does one need to have and is having a superb computing background a must for HFT because I am not much into computer programming and all that techie stuff.
Your article is indeed an insight giver and throws much needed light on this matter. In circumstances like these if we aren’t aware we are dead. Thanks for this one mate.
Warm Regards,
Chaitanya Pol
Hello Chaitanya,
I am not an expert on “high frequency trading” despite this blog post, but will try and answer you as well as I can answer you. I know several folks who work in the HFT space and, I am sorry to say for your sake, that they are all mathematicians and also very innovative when it comes to their understanding of computing. I think it would be difficult for you to enter the HFT space without these skills.
I know you are reading CFA’s Enterprising Investor from your recent comments on the blog. Coming soon is a post I have written about how to get started in the investment business. Your careful assessment of your own interests and aptitudes is essential for your success. When these are in alignment with a work opportunity then success is easier. That you dislike mathematics and computing is very valuable knowledge and helpful for you in discovering what style of investing would most resonate with your consciousness.
Best wishes for success!
Jason
I fail to see how HFT would benefit de-facto LFT. LFT to me implies true *investing* whereas HFT implies *trading*, which are completely different beasts. If an LFT is being patient, waiting for a good business to be purchased as a good price, timing the market is irrelvant. If the ask goes up $0.01 because of an algorithm and my margin of safety goes from 20.00% to 19.99%, guess what, I still ample MoS to work with; i.e.: my risk is still mitigated.
HFT should only concern those who are interested in timing the market. Alpha comes in many forms, and for LFT–investors, not traders–this means purchasing a good comapny at a fair price.
Hello tendim,
Thank you for sharing your views about HFT. I think you conflated two mutually exclusive models here. Namely, the speculator vs. investor model with the HFT vs. LFT model. What may help resolve the mystery is to think of HFT vs. LFT as robotic algorithm vs. human trader. Human traders include both speculators and investors, or if you prefer, short-term vs. long-term investors. Computer algorithms trade so much faster than any human speculator that if you use a time scale to measure the difference in execution the human speculator is now a long-term investor based on magnitude.
The very point you make about long-term investors (time horizons of 3-5 years +) and value investors’ margin of safety is the point I routinely make to uber critics of HFT. So please know that I am in alignment with your overall point.
When I was managing money professionally my turnover was 15%. With 40 holdings that is just 6 trades a year in which I am competing with alpha sharks. However, I will point out that if you lose a basis point on every trade and you notice no appreciable difference in liquidity it is still bleeding alpha. There are many sources of alpha not well appreciated by the public, but one that is manageable is trade execution. Bleeding any alpha via trading dilutes the sole advantage of the active human investor: diligent discovery of investment opportunity.
I will also point out that to make your point more contrasting you assumed a low level of alpha loss – a basis point – when, in fact, some of the illegal algorithms can cost human traders up to 5% on trades. Even in your model that is 25% of your margin of safety, and that marginal portion may be the 25% you need to earn your required rate of return.
I am guessing that if this happens to you, even on just six trades a year you would notice and be offended; especially if that 5% loss on 15% of your portfolio is the difference between top decile and second decile performance. And these losses to an algorithm’s nefarious effects likely slip below the radar screen of the exchanges and regulators. Currently, you have no recourse available to you other than prosecution, yet the costs of factual discovery are difficult given the massively high volumes traded, and then you still must prove your case (#brutal).
I would characterize the current time in the markets as transitional. That is, it is a permanently changed landscape in trading (HFT is here to stay) and human traders (even at deep value investment houses) and human regulators have yet to adapt to the new reality. That was the point of this piece; now grandfatherly in that it is almost a year old.
With smiles,
Jason