SQX_136_win_final_20221223 [SEVERE BUG]: SQX builds invalid strategies that are difficult to detect. SQX is looking into the future (stockpicker engine)

Voted for this task.

Voted for this task.

Hello, thanks for reporting the issue but for me it is not clear. 

Can you turn on the detailed log and show me which signals are not valid ? 

You can do it by selecting Trading options -> Store stockpicker log. After backtest go to Results->Stockpicker log and check entry/exit signals + open positions for every day.

Sure, I have attached the detailed logs for both portfolios (TestGroup and TestGroup_modified). It looks like TestGroup is making more trades than TestGroup_modified according to the logs (at least one more trade). But I don't know which signals are valid and not valid. It could be either portfolios that is valid, I don't know. Furhtermore, when I wrote about the difference in trades (12 vs 17) I was looking in the List of trades in Retester.

I have tested a lot more. Maybe this scenario can help put light on the issue:

I took the data from the TestGroup and shortened the data for every stock by 2 months, so the last day was 2022.10.31. I made a stock group using this data and called it TestGroup_shorter.

Then I retested the same strategies as before using Start day = 2022.09.30 and End day = 2022.10.31. When I use the original stock group called TestGroup then the Retester shows 9 trades in total (attached as Portfolio Testgroup normal)). But when I switch to the new stock group called TestGroup_shorter then the Retester shows 13 trades in total (attached as Portfolio TestGroup_shorter). The interesting thing is that both log files seem to be identical.

It looks like if there is more data beyond the End day, then the trades are different compared to when the End day is the last day in the data set. I don't know which one is correct.

All files are attached.

Does it make any sense?

I just found out that the log files only contains the trades from a single strategy and not the whole portfolio it seems. So please disregard my comments about the log files.

Do you need other files from me?

HI again. I slept on it and realized that is it difficult for me to explain what the problem is. So here are 2 screenshots that explains it all:

One screenshot shows the stock group TestGroup. TestGroup uses data from 2017.01.03 to 2022.12.30.

The other screenshot shows the stock group TestGroup_shorter. TestGroup_shorter uses data from 2017.01.03 to 2022.10.31.

So both stock groups contain the same data except TestGroup_shorter which has 2 months less data than TestGroup. I filtered out 3 strategies that show the issue.

On both screenshots the retest period is 2022.09.30 - 2022.10.31. The stock group TestGroup shows 5 trades in total but TestGroup_shorter shows 9 trades in total. I would expect the number of trades to be the same. I don't know which one is correct or if they both are wrong.

It would be nice if the issue was fixed for both long and short trades and for all indicators in the TA library.

Does it make sense now please?

Regards

Voted for this task.