Skip to content

High Frequency Trading and the Online Retail Investor

June 16, 2010

I recently had a series of communications with Jennifer Gorton-a blogger from ForexIndicators.net about providing some unique educational content . When I was presented with the possible topic of high frequency trading it was immediately appealing to me because I feel that this is a huge factor in shaping the new environment of market dynamics. Long-time readers have often heard me opine about the dangers lurking in ultra short-term trading, after reading this article perhaps you will better understand what you are up against. Personally, I would never play poker against someone who could see my cards………

High Frequency Trading and the Online Retail Investor

by Jennifer Gorton

If you have a retail brokerage account and trade the equities markets, you might have heard about high frequency trading.  High Frequency Trading (or HFT) programs are use by large financial institutions and hedge funds. These funds use the current market order information that they receive from the exchanges to execute quick deals, known as flash orders. The exchanges will give clients that execute many trades on the exchange information that is not readily available to the average retail trader. The average retail trader does not trade nearly enough shares on a daily basis to pay for this information from the various stock and electronic exchanges.

These flash orders that have been characterized as unfair are orders seen by these institutions that pay the exchange money for data before the retail investor can see it. The computer programs that the big funds have used have these flash orders to make decisions to buy and sell before everyone else can. Most believe that it is unfair for a certain group of banks and hedge funds to be able to pay for information before others are able to obtain it. This leads them to get ahead of the market and take advantage of the small retail trader. To add to this, these companies are allowed to rent office space closer to the exchanges to place their HFT computers. Known as collocation, the institutions are able to get their orders transmitted faster because of their computers being located almost right next door. The HFT programs can make or lose money if the order is a millisecond or a microsecond too late. This is another reason why the average investor will not be able to compete with the programs and institutions.

These institutions will argue that they provide much of the liquidity that is available to the market. When the computer programs trade back and forth with each other, over the course of the day, they reduce the size of the spread. This is beneficial to everyone, especially the little guy who is trading from his online broker at home. When the spread is smaller there is a greater chance that the average investor will make a profit. This is also true when trading in the currency markets. Most HFT programs focus on the equities markets because they can buy the valuable data from the exchanges. Because the foreign exchange market is an over the counter market, there is no central exchange where institutions buy their info from. There are HFT programs that exist which analyze Forex indicators and make buy and sell decisions off of those. However, because of the volume traded daily in the currency markets is so high, it is hard to gain a clear advantage and get ahead of other buyers or sellers.

It is no secret that most big institution firm profit off of their high frequency trading programs. It is also an advantage for them to have access to market data before everyone else. The individual investor, however, is not trading enough size to effect the direction of the market and in the long run will see more benefits than downfalls from these HFT programs. It is a bit worrisome when these stock exchanges will allow those who want to pay more for information to obtain it. The theory behind free markets was that everyone had a fair chance to make money. It does appear that the exchanges are being bribed by their best customers, hedge funds, and banks to obtain this valuable information first. It might not directly affect the average investor – however, in principle, what they are doing is just not right.

13 Comments leave one →
  1. retailhftguy permalink
    June 16, 2010 2:41 am

    I am a retail daytrader and I’ve always read this website because it usually has exceptionally well researched articles. I’m very disappointed to find this one which seems to have not been fact checked at all. Quick deals are not flash orders. This is a basic flaw in the article. A flash order is when a certain exchange allows its members to offer the liquidity to fill an order to prevent it from being routed out when there is no liquidity available at NBBO. See http://www.sec.gov/news/press/2009/2009-201-factsheet.htm ; Even the SEC’s own description of HFT destroys this articles credibility. Most specifically, flash orders are an option that are offered by certain exchanges. All exchanges don’t use them. If you don’t want your order to be flashed, just route your order to a different exchange.

    Now, as for colocation: it takes time for a signal from California to reach New York. Thus a retail trader in New York has an advantage over one in California. A retail trader with a higher quality internet connection is going to have an advantage too. Colocation is just an end product of the speed of light. One question I have is as follows: why does even a 100 millisecond reaction time difference (which can easily be the round trip time for a signal from a brokerage, to a home computer, and back) matter at all to a retail trader? I trade just fine not worrying about 100 milliseconds at my house. I can trade several times a minute on a stock if I want and still not have to worry about the HFT’s 100 millisecond advantage. What exactly is the advantage provided by those 100 milliseconds if you are not actively engaging in market making? (And quite frankly, if you are actively engaged in market making, you can afford to be colocating. We are talking thousands of dollars a month, not millions).

    So what is this article actually about? Finding a new boogyman to explain someones own trading failures. It certainly feels a lot better to blaim your losses on computers doesn’t it?

    • david varadi permalink*
      June 16, 2010 2:56 am

      hi there, thanks very much for the intelligent response—the article as with everything is a point of debate and certainly I am not aware of the true facts personally (especially since this is not my article). I would encourage your continued participation and welcome all discussion.

      best
      dv

      • retailhftguy permalink
        June 16, 2010 3:15 am

        I’m aware its not your article. (Sorry for thinking my comment was deleted, had some weirdness with the comment disappearing). You would not believe the personal attacks that come out though if you defend HFT. At this point its almost more of a market psychology thing then a real debate about the merits of the approach. You read the articles on zerohedge or what not and it is clear that the objective is not to study the problem, its to do away with HFT. And quite frankly there are lots of people who should be afraid of HFT. They just arn’t the retail investors.

        Algorithmic trading (which may or may not be high frequency in nature) is slowly replacing a good number of the trading operations that used to be performed manually by people. The war against HFT isn’t because of the its merits on the markets. People are against HFT because it is making traders obsolete the same way factory workers were made obsolete to a large extent by robots. Even concerns about high speed trading on news — that has been going on for years, just by people. Now we don’t need people.

        The debate you see coming here about HFT is just the beginning of the debate you will see as computers increasingly take over what are traditionally thought of skilled labor jobs. Wait till we see the uproar when computer programs do fundamental analysis and portfolio design on fundamentals. I don’t imagine it is more then a few years off.

  2. TopTick permalink
    June 16, 2010 10:19 am

    I largely agree with retailhftguy that HFT doesn’t matter to retail investors. However, I have two complaints when the HFTs pat themselves on the back to say “we provide liquidity, replacing the market makers and specialists of yore”:

    a) flash orders are the opposite of liquidity. A flash order is a right of first refusal, which takes liquidity rather than providing it. In most business deals, a right of first refusal (an option) is granted in exchange for some value (option premium). My broker might get that premium in the form of a rebate from the exchange — whether that works to my benefit is not clear.

    b) in the mini crash of 5/06, some HFTs were quoted as saying they “turned off their computers when it looked like it was getting crazy”. That’s not providing liquidity, either. (In retrospect, those that stayed with it and provided bids made a fortune.)

  3. bgpl permalink
    June 16, 2010 1:55 pm

    staying clear of the debate on the HFTs, i think there would be value if the experts writing and reading this blog could comment on how HFT is affecting the market in terms of trading strategies – MR vs TrendFollowing vs momentum (not in terms of actual order flow but in terms of which of the above tends to be exhibited as a side effect or due to HFTs).

    A disclaimer is that i know nothing about HFTs so this could be completely silly. TO me it seems that one effect of HFTs is that they exacerbate trend following on a microscopic time scale. I reasoned that this may be the case because HFTs would stand in front of the order flow and “provide liquidity” in front of an existing order (maybe a fraction of the penny less). this probably will show up as increased volume activity in the direction of microscopic-time trend.. Other algorithms which dont look at microscopic time but look at minutes instead could key off this and conclude that trend following tendency is dominating and push it further in the same direction (up or down).
    Of course in a longer time frame (hourly or daily – ironic that a day is considered a long TF) MR will come into play as the market cannot continue in the same direction.

    Thus one effect of this seems like it will lead to more TF in microscopic time frames and more mean reversion in hourly or daily timeframes.

    Anecdotally, the market ranges are higher now than before and higher this decade than previous decades.. Daily follow through and MR also has been far more dominant this decade.

    It will be great if some of the knowledgeable experts can chime in, if there is some correlation or if it is completely orthogonal.

    Please dont flame me for ignorance😉 I am trying to understand the HFT effect on trading regimes, and not on market microstructure.

    • retailhftguy permalink
      June 16, 2010 3:53 pm

      My experience trading in the HFT heavy stocks is that you can expect more mean revision everywhere and that trend following will lose you money very quickly. The immediate implication is that HFT liquidity provisioning would increase trending.
      Not all HFT (and probably not even most) is designed to stand in front of the order flow. If one type of HFT dominates there is going to be profit to be made by the other types. Excessively standing in front of order flow will make those algorithms vulnerable to mean revision.

      As for increased daily ranges: I think this is directly a function of increased investor/hedgefund churn. When commissions dominated trading costs, there was less position turnover. Now that commissions are going to zero, it can make economic sense to trade large positions on a faster time frame which increases daily ranges.

      The reality is that if you are a HFT firm you want to remain market neutral, and it is going to be very difficult to move the market while staying market neutral. If this was not the case and HFT firms were actually moving the market substantially there would be a gigantic fortune to be made trading against them. The fact that no one has done this leads me to be pretty skeptical that HFT can move the market substantially.

  4. June 16, 2010 10:35 pm

    Any claim that HFT does not have a negative impact on a retail trader is making the assumption that all retail traders submit market orders only. For any intelligent retail trader (those not crossing the spread blindly), HFT takes money from his pocket. Trading is a zero sum game. By definition, if HFT programs are making a ton of money, others trading the same securities are losing a ton of money. HFT programs only provide liquidity when there is an advantage to providing it.

    All that said, there is nothing wrong with HFT, provided the market data playing field is level. It’s simply the result of competition. I lose some of my potential profits every day due to HFT, but I am not opposed to allowing the practice.

    Finally, the comment,

    “Now, as for colocation: it takes time for a signal from California to reach New York. Thus a retail trader in New York has an advantage over one in California.”

    makes no sense. A retail trader by definition will be using a retail brokerage platform, which means any orders submitted will travel through the brokerage servers (which could be located anywhere), and will not go directly to the exchange.

    • retailhftguy permalink
      June 16, 2010 11:24 pm

      I don’t cross the spread and I *love* HFT. I can execute all sorts of more complicated strategies then I could otherwise because of the mass of liquidity available. But the key word here is more complicated. If you are losing money because of HFT then adapt your trading strategies to take it into account. You do not have a right to make easy money as a retail daytrader (and I’m assuming daytrader because from an investing pov the spread is usually irrelevant).

      Market Data is a relatively level playing field — it is not that expensive to buy better quality data. It is now within the realm of the retail trader to pay for significantly better data than is available from most brokers. I think most brokers provide horrible quality data, but it is my opinion that the advantage between say $200-300/mo data and data direct from the exchange is negligible. Presumably if you are a retail trader you can afford $200-$300/mo. If you can’t you don’t have enough capital to be trying to capture the spread. (Just as I assume that most retail daytraders have some form of redundant internet connection even if it is just a 3g mobile card or something for emergencies; if you can’t afford it, you are under capitalized).

      Sorry about the colocation comment, that wasn’t as clear as I meant it to be. It has been my experience that most brokerage servers are in the NY/NJ/Virgina area. Yes the signal has to go from the brokerage to the exchange, but my signal still has to get to the brokerage (so the analogy would be colocating in my broker’s building instead of the exchange’s building). If my brokerage’s server’s are in New York, and I live in California, I will be able to trade more effectively by moving to New Jersey. This is a simple ramification of the fact that it takes time for data to travel across the country. Yes, the signal still has to travel from the brokerage to the exchange, but I can control the distance from my computer to the brokerage. Which ISP I use can matter just as much (if very quick execution times from mouse click are important to you as a retail trader, you are going to be paying just as much attention to your ISP’s latency as that kid down the street who plays quake all day). I’ve saved 60ms latency before by changing ISP’s. How is that any different than the colocation game?

  5. bgpl permalink
    July 13, 2010 3:05 pm

    no one is probably reading these comments any more, but i saw some relevant information so i thought i would post:
    http://www.zerohedge.com/article/scientific-proof-high-frequency-trading-induces-adverse-changes-market-microstructure-and-dy

    Without taking sides in this argument I thought the following was interesting:
    “One answer could be that HFT is the only type of trading that can exhibit trades that are reactive and exhibit feedback effects on short timescales that traditional trading generates over longer timescales.”
    (similar to my comment in the thread above)
    and the paper has of course more details.

    • retailhftguy permalink
      July 14, 2010 5:09 am

      I still subscribed to the comments actually. I read the paper when it came out and … there was one very concerning factor. An increasing number of small transactions were assumed to be directly because of HFT. The entire paper was based on this assumption. Unfortunately there are two causes for the increasing number of small transactions: 1) HFT 2) Algorithmic trading as large transactions are broken down into smaller orders to be executed throughout the day instead of printing as large blocks (This is NOT HFT by any conventional definition). The paper then builds on this 1 assumption for all of its conclusions. Unfortunately this conclusion is wrong and therefore everything that follows isn’t supported. Furthermore the paper failed to rule out the change from block orders to algo orders as being itself responsible for the change. It is possible that the changing way institutions place orders may itself be responsible for the entire effect observed in the paper. The math just does not support the conclusions of the paper. (And it would be noted that the paper was just submitted to the arxiv, not peer reviewed, and there are real problems with bad papers getting put on the arxiv; the first obvious sign of trouble was someone publishing from a one person lab).

  6. bgpl permalink
    July 14, 2010 10:09 am

    good points, thanks !
    but i thought the paper did mention algorithmic execution of large blocks.
    For example:

    “Another cause may be the nature of HFT strategies themselves. Most HFT strategies can fall into two buckets Lehoczky and Schervish (2009):

    (i) Optimal order execution: trades whose purpose is to break large share size trades into smaller ones for easier execution in the market without affecting market prices and eroding profit. There are two possibilities here. One that the breaking down of large orders to smaller ones approximates a multiplicative cascade which can generate self-similar behavior over time Mandelbrot (1974). Second, the queuing of chunks of larger orders under an M/G/1 queue could also generate correlations in the trade flow. However, it is questionable whether the “service time”, or time to sell shares in a limit order, is a distribution with infinite variance as this queuing model requires.

    (ii) Statistical arbitrage: trades who use the properties of stock fluctuations and volatility to gain quick profits. Anecdotally, these are most profitable in times of high market volatility. Perhaps since these algorithms work through measuring market fluctuations and reacting on them, a complex system of feedback based trades could generate self-similarity from a variety of yet unknown processes.

    Since firm trade strategies are carefully guarded secrets, it is difficult to tell which of these strategies predominate and induces most of the temporal correlations.

    However, your points are very valid in terms of publication. Maybe i need to read the paper more carefully but the math made it not an easy read. As you mention, can start with a wrong axiom and go down the wrong path with a lot of precision.

    thanks for your comments and insight, again.

    • retailhftguy permalink
      July 14, 2010 10:33 am

      I saw that… and I thought that was talking more about short term (<1 min) breaking up trades/smart order routing/etc. (Actually this is another good point I didn't think of before… smart order routing alone might be responsible). I don't understand how something along the lines of a complicated vwap (volume weighted average price) algo would usually be considered as high frequency trading. I mean its definately algorithmic, but usually HFT seems to refer to market making, etf block creation/destruction, news, stat, or latency arb. Given many accumulation algos can run for days lumping them in this category seems like a major mistake.

      "The last thirty seconds of trading were excluded from the NASDAQ time series since
      on some days huge spikes in share trading and the ending of trading often skewed the wavelet coefficients and distorted logscale diagrams despite trading throughout the day following the traditional pattern."

      He also left out the last 30 seconds of trading because allegedly it is too noisy but because there are closing auctions/large block trades here ignoring this data is going to significantly skew results. There are definitely some methodology issues here.

      "For all selected stocks, there was a marked decrease in the
      average number of shares per trade shown collectively in figure 2. This decrease in
      trade size was used as prima facie evidence of the influence of HFT on the selected
      stocks. Next to be addressed was whether the short-term correlation structure of
      intraday stock trading has been significantly affected by HFT and related strategies."
      Going back to this, this seems so questionable because there it does not differentiate between of the algo/hft difference. A decrease in average trade size can be directly attributed to REG NMS and smart order routing. When I place a 2000 share trade on Interactive Brokers it might get split up to say 5 different venues. Even though I placed one logical order it gets reported as say 5 orders. So you might end up with 5 400 share orders instead of 1 2000 share order. Take out this assumption and the entire paper collapses.

      In fact you could easily formulate a theorem:
      Any fragmented market will have fewer big orders and more small orders if trading through NBBO is not allowed.
      Proof: Think about it and get a easy inequality.

      I'm going to keep studying the math because it seems interesting but the conclusions of the paper don't seem very strong with that assumption removed.

  7. thok permalink
    August 26, 2011 1:05 pm

    hft is like an aftermeal spike on our bloodsugaR. IT doesnt hurt much. everything returns to normal.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: