MC returns

Questions directed to our Support Team. Responses can be expected within 24 hours.

MC returns

Postby Kevin_in_GA » Sun Jul 08, 2012 4:53 pm

Monte Carlo annual returns versus "standard" backtest returns - how often have people seen thae Monte Carlo returns be better than the standard returns (which I am assuming are based on the selection rules used for trades). If the MC returns are higher, is that good, bad or irrelevant?

My current systems almost always have better MC statisitcs than the standard. I personally think that is a good thing, since you usually can't predict the order of trades getting filled on a busy day.

Thoughts?

Kevin
Kevin_in_GA
 
Posts: 171
Joined: Wed Aug 04, 2010 10:14 am

Re: MC returns

Postby Overload » Mon Jul 09, 2012 9:16 am

There are pros and cons with Monte Carlo, but keep in mind that it is nothing more than 1000 results based on randomly resequencing the trades that were actually produced by the system. The results are then sorted and displayed in the MC chart from lowest to highest.

The Monte Carlo Average Annual Return is the average of all of the 1000 returns, but I actually prefer to look at where the underlying system falls within the chart. If your underlying result falls way on the right side of the chart (well above the MC Average Annual Return), then you know that the sequence of trades selected by your underlying system is an important factor. This may be a red flag, since you'll need to find out whether this sequencing is valid or simply curve-fitting. In other words, the results of your underlying system may be optimistic, and given the alternate sequencing that you may see in the future, your results could actually be closer to what the MC Average Annual Return suggests.

On the other hand, if the Average Annual Return of your underlying lies way at the left (or lower) end of the MC Average Annual Return spectrum, you know that the sequencing of your system has actually been poor compared to the average. This provides room for optimism, that the system may actually do better than the results suggest due to the likelihood of more favorable sequencing.

I'm not sure if it's better or worse whether the results of an underlying system fall at the low or high end of the Monte Carlo spectrum. It simply gives you information. If your system falls at the high end of the spectrum, there is reason for pessimism that your system will perform as well as the results suggest. If your system falls at the low end of the spectrum, there is reason for optimism.

In a nutshell, the theory behind Monte Carlo is that the Monte Carlo Average Annual Return may be more indicative of future returns than the underlying system's Average Annual Return, giving sufficient time to gravitate toward the mean. But, like I said, there are pros and cons. It's only one way of looking at your data, and there may be valid reasons why your system's results are significantly different than the MC average.

Pete
Overload
 
Posts: 2248
Joined: Wed Nov 30, 2005 12:14 pm


Return to Technical Support

cron