Thanks for the detailed post. The backtest performance of Strategy B in 2016 is like this: but in 2015, it's like this: I didn't test it for that many years like you do, since it's a intraday strategy. Based on the two years testing, I might not think the overall performance is good enough. But it does being good in 2016. Can I assume it's a short-term strategy that is supposed to not work across many years? If we look at only 2016, the max drawdown is like 2000. So based on your idea, whenever I had a drawdown greater than 4000, I would think of stopping trading this strategy. Am I correct?
I can not see why. If I could, I would be able to code it in to conquer that. For example, "when this happens, stop opening".
You may want to go back and do an analysis of the market when the strategy did and did not work. Measure several statistics and compare the two different time periods. I would imagine the strategies failure is embedded in some metric.
Each year trade differently, check out weekly data of both years, I just posted charts few days ago somewhere based on each were different. If you are Scalping for ticks, for me doesn't matter much what longer term charts are showing as I trade noise, but for like 60 minute systems I have, very much different based on sustain trends either don't last or reverse and I get losses. When I backtest, I do it over 10 years, because you are showing differences in automation, and what if 2 of past years it loses each week, you have no way of knowing.
I think your overthinking it, or over worries, when u develop a system either automatically or manually its gonna happen that your also developing an infrastructure where it can be applied to other instruments, so while the system u develop takes 50 hours lets just say kr 50 mos or whatever, and it stops working on es lets say, the next time u go to develop something it wont take as long because u shouldve and usually its i heritied that u developed a formula or a system where the input can be changed and u still get an output so dont worry too much if something stops working, focus on building the infrastructure to where its mobile to other products and time frames and insteumenets etc
hi SimpleMeLike.. I'm sure everyone handles it differently. But, what I do is .. let's say I based it off of 10 yrs (I like to at least go back to 2007-8 to include that environment). And you have a max drawdown from the tests. You can add some to that to be conservative. Now you have some idea of where you might start thinking of unplugging the thing in case things start to happen that were not accounted for. Remember.. you only took a small sample of what can happen.. and that was based off the past.. So.. now that you have some guesses of drawdowns.. this could help to hang-on during "normal" drawdowns and not get depressed or scared at every move against your equity curve.. ha
Support and resistance aren't very hard to code. In general, almost anything related to classical technical analysis is relatively easy to code.
Thanks pauljherrera, This is good to know as I have good experience with trading levels (support and resistance) price action on intraday small timeframe.