+1 888 340 6572
MultiCharts Project Management
previous_open_issue.png
Go to the previous open issue
previous_issue.png
Go to the previous issue (open or closed)
star_faded.png
Please log in to bookmark issues
feature_request_small.png
Open Feature request MC-1713

New Optimization Mode: Random Sampling of Input Parameter Values/Monte Carlo Simulation

action_vote_minus_faded.png
3
Votes
action_vote_plus_faded.png
next_issue.png
Go to the next issue (open or closed)
next_open_issue.png
Go to the next open issue
Description

I have been using genetic optimization quite a lot to find the best parameters for 
my system and it works great. My understanding is that by design it outputs a 
subset that represents the "best" parameter values of all possible combinations, 
and this is obviously quite useful.
What I'd like to also be able to do is have it produce a random sampling of all
possible combinations (i.e. not just the "best") so that I can then evaluate the 
risk of selecting the "wrong" parameters. For example, in most cases the full 
optimization is just too big to even be run in a reasonable amount of time much 
less be further analyzed in a spreadsheet.
This (I think) should be at least a conceptually simple modification of the 
Exhaustive mode optimization (or perhaps a modification of Genetic optimization 
with neutral selection bias) where instead of running and outputting all possible 
combinations, the optimizer would randomly sample from all possible combinations 
(no replacement) and output just the randomly selected ones. The user should be 
able to control the sample size via an input for the desired % of total population 
of simulations or a specific number of simulations (which should then be kept as 
the default value for future optimizations until changed).
UPDATE: I have read through request MC-803 "Built-In Monte Carlo simulation tools" which is currently under review and would like to reframe this request in that context. I'm not quite sure what exactly is being requested by the OP because no spec was given, however it appears from the comments that Ninja Trader's MCS module is the target. So I went over to NT and read through their help document and unless I'm overlooking something it appears that NT's MCS simply randomizes the order of trades. I'm not sure I can see the value in that (I'm open to being corrected on this if I've missed something). What I am looking for is randomization of system input parameter values. For example, all moving averages from 2 to 20 days and all trailing stop values from 0 to 5% and all profit target values from 0 - 10%, etc. etc. What I think this would show again for example is this: of all randomly chosen inputs for a given system, symbol, time period, a moving average of 10 days, a stop of 2.5% and a profit target of 5% on overage produces the highest P&L or lowest risk, etc., etc. Comments welcome.
P.S. what is with this ing editor?! It keeps inserting HTML formatting garbage and removing CRs!

Steps to reproduce this issue

N/A

Comments (0)
History
Issue basics
  • Type of issue
    Feature request
  • Category
    Usability
  • Targeted for
    Not determined
  • Status
    Not Reviewed
User pain
  • Type of bug
    Not triaged
  • Likelihood
    Not triaged
  • Effect
    Not triaged
Affected by this issue (2)
People involved
Times and dates
  • Posted at
  • Last updated
Issue details
  • Resolution
    Not determined
  • Severity
    Normal
Attachments (0)
There is nothing attached to this issue
Commits (0)
There are no code checkins for this issue
Duplicate issues (0)
This issue does not have any duplicates