MM in Montecarlo Analysis

Hi all,

just a question about the Montecarlo analysis. I use a risk fixed % of account MM method.

When the software makes the resampling, does it consider the MM method (% of risk for any trade)? Or just the final result of any trade espressed in $?

Thank you

Hello,

it keeps the original position size of each trade the same but randomizes trade order

Hello,

 

it keeps the original position size of each trade the same but randomizes trade order

Therefore for the risk fixed % position sizing the Montecarlo analysis isn’t reliable? Is it correct?

Because it could keep a trade at the end of the period when the equity (and also the size) is 10 times bigger (or more), put it at the begining of a random series and lose all the money, but that can not happen in real. 

Therefore for the risk fixed % position sizing the Montecarlo analysis isn’t reliable? Is it correct?

 

Because it could keep a trade at the end of the period when the equity (and also the size) is 10 times bigger (or more), put it at the begining of a random series and lose all the money, but that can not happen in real. 

Exactly. This is another feature that results in completely corrupted data because QA relies on $ values from backtests instead of % values. This simple change needs to be implemented urgently.

Exactly. This is another feature that results in completely corrupted data because QA relies on $ values from backtests instead of % values. This simple change needs to be implemented urgently.

From the mathematical point of view do you think the results are complitely wrong? Or  we can assume them approximately reliable? The worst cases and the best ones could also offset.

Reliable.

Sorry, but you’re incorrect. QA’s Monte Carlo is currently only reliable for strategies that use fixed lot size or fixed dollar risk. A profitable strategy with solid reward profile and proportional sizing will consistently yield positive biased MC results; though results become less corrupted as a strategy’s reward profile approaches 1:1.

Note I am only talking about a single strategy above, if you are working with a default portfolio output from QA consisting of multiple, proportionally sized strategies ran on a single balance, the results are corrupted to begin with. Crap in, crap out: If you are putting corrupted data into Monte Carlo to begin with, that is all you will get out of it.

This software has HUGE potential, but issues with strategies that feature balance proportionate sizing (ie the vast majority of strategies) need to be fixed.