- ago
When I launch optimization of a strategy that uses an indicator from NeuroLab, the results are different each time. For example, I launched it once and launched it a second time with the same parameters.
But when running a backtest, I get a completely different result.

In the settings - Favor Stop (Pessimistic)
0
243
Solved
15 Replies

Reply

Bookmark

Sort
Glitch8
 ( 11.87% )
- ago
#1
Does your Strategy have NSF positions?
0
- ago
#2
Yes, it was. I unchecked the box, but the error remained.
0
- ago
#3
Without indicators from NeuroLab this has never happened before.
0
Glitch8
 ( 11.87% )
- ago
#4
There's no "error" here.

An NSF indicates there were trades that could not be taken due to Non-Sufficient simulated Funds.

Unchecking the box does not change this situation, it just changes how they are handled behind the scenes.

For consistent results check the box again and decrease position size or add margin factor until NSF is zero.
0
- ago
#5
This did not change the situation. Strategy with a single symbol, margin factor - 9
0
Glitch8
 ( 11.87% )
- ago
#6
Let’s set up a zoom support call so i can see what’s happening, it’s not clear enough to me from the discussion. Can you write support@wealth-lab.con to set one up?
0
- ago
#7
Unfortunately, I speak English badly. Let me try to show everything using the screen recording. I made a simple strategy and, on the example of this strategy, I will show everything in the recorded video.
https://drive.google.com/file/d/16eq90S54LvRdJEa3k9kWo7UFyvM3J0wX/view?usp=sharing

Different results arise when starting optimization. When Run Backtest, everything is stable
0
Cone8
 ( 6.30% )
- ago
#8
Something appears to be amiss with the optimization results - they are varying and therefore suspect. The individual backtest results, however, are consistent. For example, each time you ran an individual backtest with parameter value "45", the result was 887.83% profit (shown in the status bar), but this result changed wildly in the Optimization Results table, and I don't have an explanation for that.

Maybe Glitch will notice something about the Optimization values that I didn't.
0
Cone8
 ( 6.30% )
- ago
#9
After creating a simple strategy that uses NNPredictor and running a few optimization, the problem is clear - don't trust Optimization results that uses NNPredictor . We'll investigate ASAP!
0
Glitch8
 ( 11.87% )
- ago
#10
It looks like the NN is getting disrupted by running in parallel. Try Exhaustive Non Parallel. We’ll need to address this in at least the Neuro Lab help.
2
Best Answer
Cone8
 ( 6.30% )
- ago
#11
Good call! Exhaustive non-Parallel works fine with NNPredictor.



1
- ago
#12
"Try Exhaustive Non Parallel." - This helped.
What does it mean to use this type of optimization (non-Parallel)?
0
Glitch8
 ( 11.87% )
- ago
#13
"Parallel" means it can run multiple optimization runs simultaneously using different CPU cores. This is introducing some side effects in NL.
1
- ago
#14
Thank you!
0
Cone8
 ( 6.30% )
- ago
#15
Fixed for Parallel Optimizers in the upcoming NeuroLab release cycle. Will be released when we release WL8 Build 105.
1

Reply

Bookmark

Sort