Tackling NOISE: The Key to ATS success

Discussion in 'Automated Trading' started by asap, Dec 4, 2006.

  1. asap


    :eek: Long Post!

    I am working on an automated money management system and this concept has arise from the idea of trading out of an engine that produces random signals in a couple of extremely liquid instruments, ie the ES, ESTX50, EUR, etc.

    You must be LOL by this time, but you know, this is part of an endeavor to understand how much edge a trader gets from the typical pattern rules (chart, pairs, and other pseudo arb strategies that are so common these days in the hedge fund industry) and how much derives solely from adequate money management.

    While it is true that most liquid instruments clearly show certain patterns that could be translated into an edge, the fact is, however, these patterns are only valid when signaling powerful trends that are always backed by market sentiment factors. So, in fact, I might argue, market sentiment produces such signals, while the signals itself are worthless.

    My goal, though, was to analyze short term behavior in these contracts and more specifically understand whether some of the typical pattern analysis would contribute to sustain some edge in these time frames. The answer was NO. Most of the patterns the trader is trained to detect, will not work in this environment, because most of the movement is due to noise trading. From time to time a trend backed by sentiment may form, which will lead to another trend and then the noise resumes. The average noise-to-trend is way up, in fact is so high that is very difficult to keep any good pattern based ATS afloat in noisy environment. Most likely the bot will do great as long as the trend keeps on and bleed during the rest of the time. I am sure all of you have seen this before.

    There is plenty of bots out there that are specifically turned on and off when the manager thinks or at least whishes the conditions are met or not. However, predicting when the conditions are met is as difficult as predicting the S&P closing price for tomorrow, because the market is a complex adaptive system that constantly evolves. IMO, the success of any bot out there is pretty much about profiting from noisy environment because, most of the time, that’s what it gets (I am talking about bots that do intraday trading, because the longer term ones might bypass noise by trading every once in a while).

    So, how to become profitable in a noisy environment then?

    According to my analysis, the eventual edge that any ATS (or manual trader) gets from this trading derives mostly from money management. In particular, using 10 year historical data, I have found that, with a complete random signal generation method (i.e. using random trade generator engine provided by excel), one would end up above breakeven provided it used some baseline rules:

    - never use trailing stops - the noisy market is always chasing desperate to exit orders and most if its movement follow the trail of these stops.

    - use pyramiding - most prof traders will tell you shouldn't do this and that's because you are probably risking too much in each lot. However, if you enter just one fifth lot size and then pyramid it until you reach the full lot, you have much better odds to break even or end up winning.

    - use a profit target exit - that will allow to limit the exposure while assuring your equity curve keeps raising. This works with pyramiding very well, since it allows to rapidly stop the bleeding or recoup losses due to the added leverage.

    - stop loss - might be important to draw lines in the sand. Say, if something surfaces in the market that chances the balance of market forces, you would benefit by having some threshold upon which, you need to reassess the situation. However, this is shouldn't be triggered unless there's really a shift in market forces, otherwise it will interfere with your trading setup and most likely sunk your overall expectancy.

    - sterling ratio (Rp/max.dd)- manage your equity curve on a drawdown basis. Add risk when you're up, reduce risk when you're are close to your max drawdown. This makes the slope your equity curve to behave properly.

    Let me know your opinions. :eek:
  2. coaster


    I find this extremely interesting because I've wondered for some time if just by using proper money management one could place random trades and make a profit.

    So far you've just backtested this, right? Do you have any plans to trade it real-time? And do you care to share any stats on how your method did? Thanks...
  3. pyramiding, you mean average up or average down? if you do just random entries you win in noisy markets and you lose in trends so isn´t that 50 / 50?
  4. basis


    This is complete crap.

    What you're saying is that markets trend. There is no other way that a rule using random entry can work unless "up gets more up", and vice versa.

    So, if you want to talk about markets trending, let's see some autocorrelation stats and runs tests.
  5. asap


    I am just doing these tests to understand how much edge derives from pattern recognition and how much is pure luck. This is akin to the alpha measurements in long equity fund industry. It is quite obvious that some of the so called trader edge comes from pure luck rather than outstanding pattern recognition capabilities or code. To achieve a constant edge in any market, one should constantly adapt its strategies to ever changing conditions and have in place a powerful money management framework. In some cases, as in these tests, one can survive with nothing else than a pure random signal process, provided the MM framework is tuned to reap rewards and manage drawdown. Here's one of tests using the EUR/GBP pair. It uses a tight profit target, trading 100k euro lots and pyramids up to 3 times. It exits as soon as the the overall position reaches the profit target. Includes commission and has been ran against a real FX market (in simulation). Obviously that without profit target and pyramiding, the equity curve would have been quite different. IOW, the edge here resides purely on money management. That's why a superb money manager will most likely survive and prosper, while the best pattern recognizer might go bust.
  6. asap



    I mean both, average down and up.

    I know many traders out there and almost everyone tells me that they have a 50/50 business. However, some of them make lots of money while others go bust.
  7. bolter


    hi asap,

    Very interesting topic - thanks for posting it. Unfortunately, I disagree with just about everything you've said, but I'm always open to having my thinking and ideas challenged. However, I'll have to wait until the markets close to respond to your points.

    Now if we can just keep the lunatics at bay this could be a really useful thread.

    Good luck with it,
  8. All trades start negative. Commish.
  9. Hey ASAP,

    I subscribed just to see where this thread is going.

    Most of your basic concepts are pretty good (but a few of them are 6 of 1, 1/2 a dozen of the other), I'd like to see exactly how you implement them in your overall trading plan to consistently get the win over time.

    Looking forward to it.

  10. coaster


    I don't believe in luck. "Chance" would be a better word. Yet random events will favor one over another simply due to normal distribution. But there's no predicting that. Professional gamblers make money solely on money management. Traders should be able to as well on a strictly 50/50 system, I think.

    Looking at your graph, it's interesting to see a flat equity curve during a trend and a rising equity curve during sideways chop. That's what one would expect. So there may be something to this.
    #10     Dec 7, 2006