It seems like you've got good number of trades to make it statistically significant. My guess is slippage assumptions will be your main culprit. You absolutely will have slippage on your stop loss orders. Likely on entry too, depends how you enter. If LMT - test with price going slightly below your entry, at least couple of cents. If STP or MKT - you will 100% see slippage. If your model breaks on 0.1% slippage per side I would recommend to run it life on a small size to get some real numbers and incorporate into your test to get more precise results. PS. If you hold overnight - 20% account size position seem to be too risky. Intraday on SP500 might be alright. Good luck! Val
Many thanks for all opinions. I have not started paper trading yet (had some time-consuming trouble in other fronts). I hope to paper trade from some date in September to the end of this year. Will post results.
what is the core logic of the trades? if it's only quantitative then they are just fixed rule based trades, and backtests mean nothing for many reasons. How can they be assessed? i had traders with much "better" backtest stats, and then only to find out exactly why they wont make money going forward once i was told the rules.
I have reworked the strategy, correcting a bug in the sizing algorithm and somewhat altering the rules. Most importantly, I plotted the trade amounts and did not like what I saw at all. As the maximum amount was 20% of the portfolio value at the moment (position sizes are not adjusted later), often 4 positions left little room for others, resulting in numerous little trades sometimes as small as a fraction of the percent. 20% was actually reached only in a small minority of trades. This certainly introduced an element of random luck and made slippage much more likely. In the new version, the maximum amount is just 5% and the 37141 trades are made instead of 5744. Consequently, the position size histogram looks much healthier: https://ibb.co/r41j5Sc This simple metric should be one of the core stats IMHO. The performance stats of the new version: https://ibb.co/bWhcSL5 They are somewhat lower than those of the first version but I am much more confident in the results. As expected, the new version tolerates leverage better. Changing 100 -> 50 in the Account Margin field (2x leverage, 50% money borrowed) results in a fairly modest increase of the maximum DD from 10.29% to 14.89% (the Net Profit raises 4.2 times). I believe that this is due to decreased variance of the trade results. Paper trading will start tomorrow.
Out of curiosity, as i understand you are using AB, how are you going to execute this large multisymbol intraday strategy with AB since it has no autotrading engine?