Does HFT strategy on textbook and papers ever works?

Discussion in 'Automated Trading' started by tonyzhou, Apr 16, 2013.

  1. Or the HFT shops use their own, creative secret formula?

    1. Speed matters? If it is, the guy with highest speed wins and all others lose, from game theory point of view. So given the fact that multiple HFT shops exists, I think speed does not matter, or you should never put your strategy based on the assumption you are the FASTEST.

    2. If speed does not matter, and I want to equity market making, with passive limit order at 3 times spread, will it be profitable?

    3. Spread should be set according to order arrival rate and inventory. Assume order arrival rate is constant during the day, then the spread is only controlled by inventory, right?

    4. Does paper "High-frequency trading in a limit order book" even work? I tried, but seems does not work. So that gives the question, if the classic theory does not work, how come HFT shops profit?

    Anyone like deep discussion with me are welcome. Thanks.
  2. misaki


    It's like trying to compete in F1 (HFT) with a homebrew of information from Google search results and a few books you picked up in the library on combustion engines, wind tunnel testing and viscous flow, a Ford-supplied engine base (IB, QuantHouse), and a small team of former AAA service technicians (previously IT professionals).

    You have the occasional Koenigsegg that emerges out of pure inspiration and open source content, but more likely, success comes from inbreeding between large, successful and early entrants like Pagani/Lamborghini, McLaren/Mercedes, Lotus/Renault (Optiver/Flow, SIG/Jane St, DE/Spark/Two Sig, Citadel/Headlands/Teza), with a well-financed research team, privileged access to correct tools such as Microsoft MES, and/or financial backing such as from Santander, Etihad, Petronas.

    The reasons why someone would spend time publishing diffuser flow research that only gets them 30 citations - besides protect their current job as production engineer at Chevrolet - when the unverified claims of their results could land them positions as lead chassis designer for Ferrari or CTO for RBR, are questionable.

    That's the nature of the business and Microeconomics 1010.
  3. Being the fastest to a trade is only one subset of HFT. There are strategies out there that I'd personally categorize as HFT that do not require you to be first.

    My own definition of HFT is simply trading at a speed and volume that cannot done manually by humans and usually involves co-location and requires a very low commission structure (.001 per share or less). To me, it does not mean being the fastest possible trader out there.
  4. dealmaker


    Agreed. Beside speed, co-location, low commissions they also enjoy market-maker status thus they can buy and sell at the same time, sub-penny and get an early look at the order-flow.
  5. Not sure if you are being serious or not. However, you do not need to have "market maker" status. Anyone can sub-penny to the affect of sending mid point orders. Early look at order flow?

    Sounds like you are mistaking internalizers for firms that run HFT strategies.
  6. Let's say traditional market maker without high frequency. We can think they are scalper. So how can they make money? what strategy to set bid/ask to have high probability wins?

    Have knowledge in programming and math. Just no experience. So looking for strategies that I can DIY.

    Well, though different shops have different kind secret, but the fundamental should be the same. Like value investing, we know what to look at, it's all about different parameters you look at. But at least, even if the number you look at is not optimal, still you can earn money, though less.

    I just did not figure out how market maker wins. So far my tries are all failed. Sadly to say.

    Not to beat people in Getco, which inventory is 400M, revenue is 1B/year and profit is in the 100M. You see that cost eats everything. But the raw number is promising, 400M inventory get 1B per year.

    So just try to get my 400K have the same revenue level, say 1M/year. All the cost is my time and effort. Not too bad for the numbers for me if I can do that. With less fund, hope to get more return since no liquidity issue on me.

    Any though? Thanks
  7. I don't imagine anyone will tell you much. Those with strategies that work, tend to keep quiet. There is a fear, sometimes rational, sometimes not, that if you share too much, someone will figure out what you are doing and remove an edge. Even in some firms (not all of course), groups working on their own strategies don't like to share with other groups.

    There are probably a dozen or more papers written regarding market making. If you are really good at math, you can probably figure it out. While it may not work exactly as they lay out, it may point you in the right direction.
  8. Well, personally, I do not believe there are multiple solutions that can survive, but only one optimal solution. With so many shops, variants, the fundamental should be the same. If a shop use ad-hoc without any theory behind it, I just doubt it.

    In terms of fundamental, the basic is in micro structure textbook.
    But in reality, what taught in textbook is based on some assumption. If the underlying model is wrong, how can you trust the results?

    Let's say price movement. Is it based on random walk with natural contrary movement, or it should be martingale process? Different probability leads to different results. But if you put those results into reality, does it work? Probability not, since the assumption is just not fit the reality.

    Looking for sub-optimal solution as long as can make money. Still staring at how to improve and turn to the profit region.

    Fundamentally, spread is the revenue, and in a Nash equilibrium, all players should share the revenue across the fund they put it the game.
  9. hftvol


    PM me if you are interested in working hft strategies. I am happy to share.
    #10     Apr 18, 2013