Should a trading system work in many different markets ?

Discussion in 'Strategy Development' started by jrkob, May 16, 2004.

  1. jrkob

    jrkob

    I haven't been posting for quite a while.

    I have read so many times here that for a trading system to be good, it must work on as many different markets as possible, to prove its robustness, unless there is a very good reason why it should work on one market only.

    I have developped a trading system for my local stock index (not in the US). After what I think has been very diligent testing, test of robustness, etc etc etc... I have traded it for the past five months. Results in live trading have been perfectly in line with expectations, and even better because volatility has been higher so far this year and my system is supposed to capture that kind of market conditions. Don't get me wrong: I am NOT rich, because I trade small for the time being. But results are in line.

    I ain't no computer genius, I use Excel as my unique tool, and my system is what most of you guys would consider as indecently simple.

    I cannot explain WHY I can't make that system work on other markets though. But for the time being I don't plan to stop trading it. Let's see...

    Anybody else here trading a system that SEEMS to work on one market only ?
     
  2. prophet

    prophet

    I agree with what you’ve read. An edge that exists across multiple markets is likely more robust and stable over time than a single market edge.

    Did all your initial testing and development involve that one market? Or did you test multiple markets and only find one that was profitable?

    I spent a few months developing a system on NQ data only. When it worked, it only worked on NQ and provided an actual trading lifetime of 3 months before the equity chart went flat. Fortunately before it died I started testing ES, 6E and YM. This revealed which system aspects should be common across different markets and which aspects need to be specialized per-market, and how to adapt them in a robust way. This led to much better systems overall.
     
  3. Markets are in constant change. Find out exactly why your system is working on your current market and don't worry about it again until it stops working on it, then change to one that it will.

    This has happened to me a few times now.

    In order for a system to be legitimate it merely needs to make you money. PERIOD.

    Enjoy the ride and be happy. :)
     
  4. There has been much discussion about this over the years as you have correctly noted.

    I agree that the most robust systems, those that can last a lifetime, will capture the inate supply and demand and fear and greed patterns present in markets throughout time. I have heard from some of the seasoned traders here that you want a "mitten fit" strategy rather than a "glove fit" strategy for longevity, that is something that is not overoptimized to the current situation. That is not to say however that you can not develop systems that work well in only certain markets in certain situations. There is nothing wrong with that as long as you have a plan when and where to implement your strategy. In fact it can be a very profitable way to trade. It sounds like your system fits into this second category. That's fine, just be sure to start developing your next system to stay ahead of the curve.

    In terms of the difference in markets I think it would greatly benefit you to start comparing the two markets in a statistical fashion. In this way you can see and understand why one market may be profitable for your system while another is not. The tools that you want to use for these are simple. Try comparing autocorrelation, daily return histograms, high low time of day occurances, maybe even Hurst Exponent analysis. As a first step you are likely trying to see how trendy or how mean reverting a market is. If your trend following system works in your primary market but not another market exhibiting the same statistical trend friendly properties then you may be in trouble. If however your trend following system works in your primary market but not in one exhibiting more mean reverting statistical properties then you know the reason. Most markets are roughly the same statistically but they all have slightly different flavors. See how yours stack up. Good luck.
     
  5. Samson, How long do YOU wait until your system stops working to switch? At what point do you lay the system to rest, after how much$ or percent?

    Michael B.


     
  6. I think this is flawed thinking so an extent. Different markets behave in different ways. A simple breakout system may trade very well on financial markets, but get chopped to death in the S&P.

    A system that is robust across all markets with the same parameter settings may inspire confidence in the system. But I think different markets need to be treated in different ways (by markets I am referring to sectors like currencies, indexes, softs...). The underlying mechanics behind corn is drastically different than that of the S&P, or even gold.

    Don't take this as an encouragement to overoptimize to fit a market. Overoptimization is not a good idea. But you may find better results if you intelligently optimize for different sectors.
     
  7. Well said.

    For example I find the YM much different than the ES. And I havn't words for your index in your part of the world, Holy Volatility Batman!

    Michael B.

     
  8. Electric this is a very good question and one that I have struggled with for sometime now because of it's complexity.

    What I am trying to do now is to stay on top of many more markets and just pick the ones that meet the optimal criteria to suit my methods on any given weekly/monthly basis.

    So far this is helping but it is still a work in progress.
     
  9. Samson,

    I have made a promise to myself I will not post what sounds good, but rather what I know.

    Let me know when you figure it out ok? As I truly cannot "put it in a can"

    Michael B.

    See my Journal "YM in a Can"

     
  10. jrkob

    jrkob

    Thanks a lot guys for your input.

    Prophet: my system was designed for that market specifically, because I have known that market for a very long time. In short, I studied that market first, THEN I designed the system. I prefer to do it that way as opposed to design a system first, and then check if it makes sense or not.

    Samson: I know WHY my system works on that particular market. What I haven¡¦t managed to do is comparing it with other markets and spot the subtle differences. Can you guys tell me what characteristics of a market I should compare ? For instance, volatility, occurences of price shocks etc etc (opmtrader already gives me a few things to check)¡K
    I think I agree with you that as long as the system makes money¡K I¡¦m happy.

    Now what you guys seem to suggest is that my system will work as long as the nature of that market doesn¡¦t change too much, and that when my ¡§luck¡¨ runs out, I will be time to change the system, or adapt it to the new conditions.

    I think this definitely makes sense, I will keep that in mind.

    I agree with opmtrader on the overoptimization issue. I currently don¡¦t trade my system with the most profitable historical PnL, but with those that bring me the largest number of profitable months. This is because from a psychological point of view, I know that I would have a lot of difficulties trading a system that brings 4 or 5 months of consecutive losses. Even if in the end I make more money. My system has 5 degrees of freedom.
     
    #10     May 17, 2004