Monte Carlo Simulation - Correct Assessment?

[NJ23]

I have backtested a strategy on two different symbols. For both of them time frame is different(I backtested them one after the other). One trades on X min time frame and the other on Y min time frame.

I wanted to run a simulation on how this would go when traded together. I couldn't find a direct method of doing so.

So, what I did was I took the % returns from Symbol 1 and the % returns from Symbol 2. Copied them together in an excel file. And ran a Monte Carlo Simulation of them together(So, the assumption is that they retain the position sizing method in their original testing and in the Monte Carlo the both trade at 1:1 allocation on a portfolio level). Can this method be an alternative to backtesting them together(which could be a real pain in the behind if we were to use AmiBroker/NinjaTrader)? Can the results of MCSim be considered for any decision making?

 

[Somatung]

I have backtested a strategy on two different symbols. For both of them time frame is different(I backtested them one after the other). One trades on X min time frame and the other on Y min time frame.

I wanted to run a simulation on how this would go when traded together. I couldn't find a direct method of doing so.

So, what I did was I took the % returns from Symbol 1 and the % returns from Symbol 2. Copied them together in an excel file. And ran a Monte Carlo Simulation of them together(So, the assumption is that they retain the position sizing method in their original testing and in the Monte Carlo the both trade at 1:1 allocation on a portfolio level). Can this method be an alternative to back testing them together(which could be a real pain in the behind if we were to use AmiBroker/NinjaTrader)? Can the results of MCSim be considered for any decision making?

So far a very interesting question :thumb:. Never tested that way on any ES, even I run a MCS program over here. But the MCS is not related in any way to the ES. Have subscribe to your thread because of the value of your question.

H C / Dan

 

[NJ23]

So far a very interesting question :thumb:. Never tested that way on any ES, even I run a MCS program over here. But the MCS is not related in any way to the ES. Have subscribe to your thread because of the value of your question.

H C / Dan

Thanks Dan. My guess is it should work. In any case, the worst would be predicted by the MC Simulation since it would take the trades for both the symbols and that too completely randomly.

 

[miss.epsilon]

have you tried matlab or R for mcSimulation?

 

[NJ23]

have you tried matlab or R for mcSimulation?

No, I haven't. I find Equity Monaco easier to use.

 

[rmike]

So, the assumption is that they retain the position sizing method in their original testing and in the Monte Carlo the both trade at 1:1 allocation on a portfolio level. Can the results of MCSim be considered for any decision making?

Hi NJ23,

There appears to be a basic flaw in the premise, or possibly the info provided is a bit ambiguous on a few key aspects.

The intent of running a Monte Carlo is to get a better fix on real world returns (or principal parameters of a backtest). This is done by computing statistical probabilistic returns distribution by manipulating trade (and underlying instrument) order in various permutations and combinations. Once one understands this premise, the raison-de-etre for running a Monte Carlo is readily apparent for a large universe of instruments traded with a limited capital

In your case, since you are looking at trading only two instruments and (moreover) with a fixed 1:1 portfolio allocation to boot. Ergo, all trades will be taken. The trades are not hampered by either the number of scrips or capital available (unless there is dynamic allocation). Hence the utility of a Monte Carlo analysis (simulation) is dubious (at best) in this particular case.

You can easily arrive at a good guesstimate by averaging out the ROE and adding the MAE of individual backtests for worst case scenario.

HTH

P.S - If your query's aim was different in intent then please feel free to disregard the info.

 

[mechtrader]

Hi NJ23,

As per me the Mc sim would do. MAy I know that it is a intraday strategy or a positional one.

THanks.
MT

 

[NJ23]

Hi NJ23,

As per me the Mc sim would do. MAy I know that it is a intraday strategy or a positional one.

THanks.
MT

It's an Intraday Strategy.

 

[mechtrader]

If its an intraday strategy then you can copy all the completed trades day wise in excel and then mix the results of two and make a combines time series.
You can run your analysis on that. That should give you an idea how both the strategies work together.

Thanks,
MT