when things go wrong...monte carlo

[killATwill]

how do traders here use monte carlo simulations to simulate the sort of drawdown they should expect with their trading strategy. how do they use a monte carlo sim to tell them when something is broken (their system is performing too well or too poorly)? also, would you recommend any tools for calculating?

thanks

 

[jrkob]

See the attached spreadsheet, you can use it as a starting point.

Variables are in red: win%, average win, average loss, and number of simulations to run in the Monte Carlo process.

Input your variables, then run the macro "Macro1".

In the Sheet2 spreadsheet, it displays the probability that your strategy "breakeven or better" after 20 trades.

The same way, you calculate the probability of a "-15pts" drawdown or worse after 20 trades.

Of course this is not only merely an example (20 trades only) but it is also very theoretical.

Personally I haven't found that kind of exercise extremely useful, because of the volatility in the win%, average win and average loss. Just me though.

A few things to keep in mind:
1) you will need more trades than the 20 I put in for the results to be meaningful.
2) you will need more than 1000 runs for the results to be meaningful.
3) Excel random numbers are not exactly random. Do a search "random number" under my name and you'll find why, as well as alternatives that have been suggested to me.

 

[nononsense]

Quote from killATwill:

how do traders here use monte carlo simulations to simulate the sort of drawdown they should expect with their trading strategy. how do they use a monte carlo sim to tell them when something is broken (their system is performing too well or too poorly)? also, would you recommend any tools for calculating?

thanks

Doesn't make any sense unless you have a sufficiently correct probabilistic model for your market. Nobody has ever come up with someting like that.

If not: Garbage In, Garbage Out.

 

[jrkob]

Quote from nononsense:

Doesn't make any sense unless you have a sufficiently correct probabilistic model for your market.

Absolutely agreed. And that makes that kind of tool not extremely useful. It's very theoretical.