Monte Carlo analysis

[Pat]

Would it be possible in the Monte Carlo analysis to completely randomise the order in which trades are taken during a test.

Something akin to generating a trade database, taking all of the trades and putting them into a hat and on each day randomly selecting how many trades are drawn out of the hat, which potential trades are available to be taken and based on available money, which trades are taken ? :?: The only thing that would remain fixed potentially is the period of the test and the trade length of each of the trades.

This would allow for a similar number of overall profitable and losing trades to occur in the future (backtest), but the order that they occur in to be different to what they did in the past. This could be a rough way of simulating out of sample data, apart from the normal rigors in database production of in sample design and out of sample testing.

The reason for my question is the current simulator, while excellent, always commences on the same day and the order of available trades is always the same (all be the trades taken are randomly selected). Which I feel could give the results some bias towards curve fitting in-sample.

Cheers,
Pat

[jdfagan]

I agree with Pat that the MC randomizer shouldn't be bound to only subset of trade database (those trades that have the same entry date).

I think it would be a powerful option to open up Monte Carlo analysis to randomizing the trade distributions across the entire spectrum of trades.

It does seem limiting that it only juggles trades that shared the same entry date and would thus bias the MC analysis to the past overall market conditions. Allowing full randomizations I would think allow the system to be fully stressed and see its full impact of having all R-Multiple trades juggled in the same bag for each trade pick.

Regards,

JD

[jdfagan]

Similar thread to this one FYI.

http://www.compuvision.com.au/phpBB2/viewtopic.php?t=667