Hello Guest

Consistent Seed In Simulations

  • 2 Replies
  • 81 Views
*

AlsoGeese

  • **
  • 5
  • +0/-0
    • View Profile
Consistent Seed In Simulations
« on: October 12, 2020, 10:53:44 am »
I'm trying to do multiple simulations with different settings, to then compare which works better. However every simulation I run it uses a different Server Seed which then of course gives me different results that I can't compare.

I understand that I could take the .csv file out and then manually do the testing in Excel, but it's substantially faster with better control and easier to do it in the Bot.

Is there a way to make the Simulation use the same seed every single time it does a run through?

*

Seuntjie

  • *****
  • 508
  • +14/-2
    • View Profile
Re: Consistent Seed In Simulations
« Reply #1 on: October 12, 2020, 01:07:01 pm »
No, and your approach to use the same seed is, scientifically, a very bad way of looking at strategies, because you'd be building a strategy to overcome that specific data set.

A better approach would be to do several simulations (10+) per strategy and compare an average of the results.

*

AlsoGeese

  • **
  • 5
  • +0/-0
    • View Profile
Re: Consistent Seed In Simulations
« Reply #2 on: October 12, 2020, 02:38:19 pm »
That's true, I hadn't considered overfitting. I was more just looking to see if there was potentially a way that I could compare results quick and dirty to get things going, but you're right that I'll be much better off taking a much larger set and seeing how I go. Thanks for the point.