This is more of an intermediate or advanced question but I’m not sure where to post it.
Suppose Trader A develops a robust, ‘one-size-fits-all’ trading system that can be used across multiple pairs. It’s been tested on 10+ pairs since 2001 on the daily timeframe, and sample size is over 2000. The profit factor is around 1.25, which isn’t super-hot, but Trader A decides to trade this system because of its robustness, which boosts his confidence.
Trader B has a different philosophy. Instead of creating a ‘one-size-fits-all’ omni-trading system, Trader B will develop a portfolio of multiple trading systems, with each system optimised for each currency pair. The portfolio will end up consisting of 10 trading systems, with each system based on a much smaller sample size of 200. The average profit factor from these systems is 1.25.
Cumulatively, Trader B is trading on a sample size of 2000, just like Trader A. However, Trader B has also diversified the ideas and concepts that he is trading, which adds an element of robustness that Trader A lacks (who is trading on a single idea). Trader B’s weakness, however, is that each system in his portfolio is based on a small sample size of 200, and optimising a system for each currency pair may lead to excessive curve-fitting.
Is Trader B’s style of system development and trading viable? What about Trader A? Thoughts please.
I think that Trader B has the weaker system because a sample size of 200 is quite small - and if it only covers 1-2 years then is very specific to the current market. The overall market varies considerable from year to year, even month to month. This is also a weakness of Trader A because many strategies that fit the market in 2001 do not work now, unless someone has ‘curve fitted a strategy.’ I think I would go for trader A because he is less likely to blow up (even though I would not necessarily expect a profit) as a small sample size in B is a weakness despite the different systems. Trader B is also trading different pairs but may be relying on the same concept/correlation that can just as easily fall apart for multiple currencies.
Actually another problem here is that trader B who is depending on “diversification” from different pairs is actually still trading inside the currency complex. If trader B was trading completely different sectors it would be different. But actually your basically all trading inside the same sector, here currencies. This could be the same in self directed 401ks where people think having stocks from the SP500, Dow and NASDAQ means diversification, when actually it is not that diversified as they are all US based stocks. However system vs system correlation does become a factor which i will address.
Lets take out # of samples and measure on degrees of freedom, which is much more accurate measure of robustness. # of samples is what creates the degrees of freedom relative to constraints or rules. which is why in general more samples are better but its not the samples that are creating the robustness on their own. Assuming both systems have 95% degrees of freedom (my personal minimum) then comparing on the same currency they would likely have the same amount of robustness. This could be checked in a monte carlo simulation to determine if that was infact the case.
The most robust method for trader B would be 10 systems in totally different complexes, fixed income, equities, commodities, currencies etc. But also 10 systems that are non correlated to each other as well. Which means if 2 systems would have a <.15 correlation, against each other. so the sectors are diversified, the trading systems them selves are independently uncorrelated. now your cooking with gas. If you can have a net 0 correlation portfolio irregardless of samples but maintained a 95% degrees of freedom which could be verified in a monte carlo simulation. I would select that trader.
However given the example above unless the systems are extremely diverse, because all pairs are still inside the currency complex. then his portfolio will not have a low correlation coefficient on balance. in which case trader A has the advantage. if trader be was able to have extremely diverse systems all trading inside the currency complex that had net neutral correlation coefficient then I would select trader B. It is possible just more difficult to do it that way, having different sectors makes it much easier.
bottom line: select trader B if his systems portfolio on balance has a low net correlation coefficient and can maintain 95%+ degrees of freedom through a monte carlo. select trader A if not.
I agree with MH on diversification 100%. When it comes to a trading strategy, since your example is all about forex one strategy should work on all currency pairs. A good strategy will work on all pairs, just like a good equity trading strategy will work on all equities. You will not have a different technical approach trading Apple and Google, the approach will be the same the signals will differ.