And for every system for which 2 year sample is enough, there are 10 which turn to crap trading live, don't you think? I wouldn't be surprised if you had some of those during your research years. In any case, what do you think is relevant sample size for swing trading?
I can only speak from my own experience. I build only one system, and after a lot of problems I got it where I should be. My initial idea was very good but very difficult to put it in a working and manageable system. That's what caused for me a lot of problems.
So I never had other systems that might, or might not, have crashed. Systems that crash after a short time are just proof that the initial setup was wrong. I invested a lot of time in understanding the market, developing the needed logic. So I did not immediately start to program or test indicators, like most do. A good preparation is (more than) half of the work.
I never studied swing trading so have no idea about relevance of testing period.
I compare it with Tesla versus German EV's. The Germans will from start have a much better product then Tesla had initially as the preparation is far better. Tesla was more a trial by error. The Germans fully develop and test profoundly before producing. So risk of bringing a bad product will be much lower for the Germans.
