I have had repeated and inexplicable deviating backtesting results in the past. I thought for a while about what could be the cause and have now found a workflow that makes it repeatable. Something seems to be wrong with the optimization function. Below is the workflow with the results/deviating backtesting results.
First the settings and the first backtesting result.
I then carried out an optimization with the following settings:
I then changed the ConsecDown to 4 days in the strategy.
The old backtesting result with 3 ConsDown is displayed, although the backtest ran over 4 ConsDown.
I then set the strategy back to 3 ConsDown.
Now the backtest gives different results than in the first backtest shown above (also with 3 ConsDown). Despite the identical strategy, the results suddenly diverge.
In addition, the results do not match the results of the optimization run. Here is the optimization run again.
The problem seems to be related to the optimization (the results seem to be shifted or something similar). I hope the workflow shown is comprehensible. I have tested it with finantec optimization package and without finantec optimization package. The results were the same.
I hope you can find out the cause and fix it.
First the settings and the first backtesting result.
I then carried out an optimization with the following settings:
I then changed the ConsecDown to 4 days in the strategy.
The old backtesting result with 3 ConsDown is displayed, although the backtest ran over 4 ConsDown.
I then set the strategy back to 3 ConsDown.
Now the backtest gives different results than in the first backtest shown above (also with 3 ConsDown). Despite the identical strategy, the results suddenly diverge.
In addition, the results do not match the results of the optimization run. Here is the optimization run again.
The problem seems to be related to the optimization (the results seem to be shifted or something similar). I hope the workflow shown is comprehensible. I have tested it with finantec optimization package and without finantec optimization package. The results were the same.
I hope you can find out the cause and fix it.
Rename
The strange thing is that the deviations are retained. If you save the optimized strategy and call it up again later, the results shown after the optimization run are retained. If you recreate the strategy with new blocks and then run a backtest (without having previously optimized it), the old results reappear. Optimization seems to be changing the strategy.
The Strategy is run with whatever settings are in the sliders on the Strategy Settings panel, can you check those?
In fact, that is the "mistake". I didn't know that. If I set the Building Blocks to 4, the Strategy Stettings show 3 and vice versa. Why is that? It's a bit confusing. Does this mean that I always have to check in the Strategy Settings whether all the controls are set exactly as they are in the Building Blocks? Couldn't that be matched? I mean, why is it different?
Now I know why it was like that. Thank you very much for your quick help. Perhaps it could be matched in future (so that the settings change the same on both sides, or is there a reason why this is not the case?)
The settings you specify in the Building Block is the default value. But changing the default doesn't change whatever current settings you had set in the sliders. There are arguments for both ways I suppose, but we felt we didn't want to possibly change the values out from somebody so we decided to keep them separated.
All right, I understand. Thanks again for your quick help!
Your Response
Post
Edit Post
Login is required