1 Topic by tradingblueocean67 (edited by tradingblueocean67 2024-04-05 07:10:08)

  • Registered: 2024-02-23
  • Posts: 25

Topic: Very low or zero output from Montecarlo

Hello Everyone,

since a week I am experiencing a very low output from the montecarlo test.
Apparently the parameters are quite "easy to pass", but i do not get why yet.
All available indicators set + 1 entry rule.

it is about 0.9 and 0% after at least 1o hours generating.

Anyone experiencing the same issue?

Thx
Vincenzo

Post's attachments

Montecarlo 1.png 133.03 kb, 1 downloads since 2024-04-05 

Montecarlo 1.png 133.03 kb, file has never been downloaded. 

Montecarlo 2.png
Montecarlo 2.png 11.84 kb, 1 downloads since 2024-04-05 

Montecarlo 3.png
Montecarlo 3.png 30 kb, 1 downloads since 2024-04-05 

You don't have the permssions to download the attachments of this post.
  • Popov
  • Popov
  • Lead Developer
  • Offline
  • From: Bulgaria
  • Registered: 2006-11-30
  • Posts: 7,895

Re: Very low or zero output from Montecarlo

"Randomize backtest starting bar" is the bottleneck, most probably.

Imagine your strategy makes a smoothly rising balance curve, reaching $1200 profit.
Let's say Monte Carlo cuts off the first 1/3rd of the bars, meaning the strategy will make roughly $800 profit.
The Monte Carlo test will fail in such a case.

Such Monte Carlo settings will favour strategies that make most of their profit near the end of the data series.

Try using a Profit Factor instead of a fixed Profit number.

Popov's Website

  • Registered: 2024-02-23
  • Posts: 25

Re: Very low or zero output from Montecarlo

Good point, thx!
BR
Vincenzo

  • Registered: 2024-02-23
  • Posts: 25

Re: Very low or zero output from Montecarlo

Popov wrote:

"Randomize backtest starting bar" is the bottleneck, most probably.

Imagine your strategy makes a smoothly rising balance curve, reaching $1200 profit.
Let's say Monte Carlo cuts off the first 1/3rd of the bars, meaning the strategy will make roughly $800 profit.
The Monte Carlo test will fail in such a case.

Such Monte Carlo settings will favour strategies that make most of their profit near the end of the data series.

Try using a Profit Factor instead of a fixed Profit number.

Hi Popov,

I have also changed pair and run the exact same condition and the results is completely different, provinding more then 100 strategies passing the MC validation with exceptional net profit.


Might it also be that some pair comes very week out of the generation/validation and therefore do not pass the MC?
While the same setup for a different pair it is much more afficient and stable ?