"Universality" of your trading system?

Discussion in 'Strategy Development' started by gmailer, Mar 11, 2012.

  1. gmailer


    Hello guys,

    How do you deal with "universality" of your systems? Do you think system should be good at all markets(or at least as many as possible) or can it be good only at one?

    I have few trading systems designed for forex. But I have following problem - some of them look great on EUR/USD but completely suck on GBP/USD.

    Should I worry about that? My intuition tells me that system should work good on many markets - this means it is not "tailored" for one market/curve. But my experience shows me that my systems tend to look different on different pairs. I use 5000 periods - intra.

    What is your opinion on that? Is it normal that system works in different way on different pairs? Or should I worry that system is not as good as I thought?
  2. Since we can't even expect any one strategy to work all the time on even one market, I feel it's too much to ask for one that works on all at all times. Each market has it's own particular rhythm I think, and even that changes with time. My suspicion is that strategies adapted for the particular market and it's current condition will outperform a more generalized strategy. But I'm still learning so..... FWIW.
  3. vikana

    vikana Moderator

    I've actually seen that many times. GBP is just different for reasons I don't understand.
  4. It's ok to design systems for a single instrument, and not worry much about if they don't work on others, imho.
  5. My opinion: I see wildly different price action over different stock sectors and have also noticed the GBP difference in currencies.

    Some markets are too noisy for trend following strats, but they're great for range algos. I don't see the logic in discarding a strat that works well in one area because it won't work on both. You have to target your service to the right market.

    This can also go for times and time frames, as well. One of the classic tactics of trend following was a volatility filter to turn the system off if the market was too wild. The demo for Trading Recipes shows this in action.

  6. gmailer


    Thank you all guys. So it looks like I worry too much.

    My concern was following - if system does not work on many markets(different environments) how can it survive on one market for a long period of time in future? I mean my market can change in future - right? So if it works on many markets it could survive changing environment on my market(EUR/USD).

    System is trend following - breakout above/below highesthigh/lowestlow. I thought that such a very basic concept would work in similar way on different markets - but apparently it does not.
  7. This may or may not be a valid assumption, but there are other ways to calibrate your trend following system. Not all successful organisms are cockroaches (able to survive in almost any environment), but few elegant animals can survive outside of their niche habitat. The key is to design your animal for a set of conditions that exist in abundance, and tend to change slowly enough for natural selection to keep you ahead, before going extinct.

    Trend algos like lots of signal, not a lot of noise. However, a noisy looking chart may actually be much cleaner on a different time frame. You are allowed to target your service to the best market. If you chase the same solution and market as your competition, it might get crowded......... Only so many wolves can survive in one forest.

  8. You should optimize your system monthly to semi-annually.

    You then should create a backtest on your data where you have done this, but because most are very limited in their trading data, I like to use the most recent 2 years and optimize monthly.

    Because my systems work on every symbol, I can understand the consternation of wondering if it works here why doesn't it work there, and vice versa. The answer would be curve fitting.

    The question's easier to solve with multiple symbols to collect optimized parameters on to see other symbol's characteristics, but at least make sure you can use the model on something else.

    I might be spoiled on my systems, but I understand which ones they work best on, and even if you did have a universalist trend following algo, you'd still decide which markets you ought to trade anyway since capital is typically limited.
  9. gmailer


    Thank you both guys for valuable input. So from what you all are saying following picture comes up for me:

    It does not have to work on all markets. It can work on one market only- but I should stay ahead of possible changes. Try to look into future how it can change and know what to do with it.

    My idea would be now to try optimize system for other markets(because I do not optimize my system now! not in a sense of doing it in automatic testing, by computer) to see what parameters need to be change and HOW in order to make it work on other markets.

    This way I could learn more about my system(I am thinking about particular one) and use this knowledge in future for my market(EUR/USD) and possible changes. Is that correct thinking?

    2. BTW. How often market conditions can dramatically change, so system should be optimized again. How often you do it? How often you notice changes? My system works on 1h charts but it makes like 1-2 trades a months only...reversal. I wanted to make it as stable as possible.

    Optimization is the word I do not like...I would not like to optimize too often.
  10. jcl


    The Forex markets are rather similar. In my limited experience with trading systems so far, if a system works reliably with one currency, it should also work with most others, but might require different parameters. At least I don't have a reliable system here that would only work with one currency.

    How often you must optimize depends on your strategy. The markets change permanently, but some strategies can adapt and don't need frequent optimization.
    #10     Mar 13, 2012