Montecarlo return versus avg annual return

StrataSearch will find what you tell it to, but what should that be? Will the same criteria work across different Sectors? Different time periods? Here we discuss the ins and outs of OneClick Searches.

Montecarlo return versus avg annual return

Postby Schutten » Tue Jan 18, 2011 3:23 pm

Hello,

When I compare the monte carlo returns in the performance report I see a big difference. Montecarlo returns is about 8% higher then average anual return for all portfolios.
What does this imply?

Regards,
Dennis
Schutten
 
Posts: 42
Joined: Thu Oct 25, 2007 6:17 am

Re: Montecarlo return versus avg annual return

Postby Overload » Wed Jan 19, 2011 9:52 am

When you look at portfolio-based results, only some of the positions may be making it into your portfolio. For example, suppose there were 25 buy signals produced on January 15th, but your Portfolio Size is set to 5. In this case, only those 5 positions would be included in your results, and there would be no evaluation of those other 20 potential positions. Yes, it is true that the Portfolio selected those 5 positions first, but it can also be said that if the conditions were only slightly different, you may have purchased any of those other 20 positions. In any case, your Portfolio Size = 5 results include only those 5 positions, not the other 20.

This is where the Monte Carlo Analysis is helpful. In a Monte Carlo Analysis, StrataSearch runs a unique analysis where the same number of trades are pulled into a Portfolio of 5, but it randomly selects them from the entire pool of positions rather than the strict portfolio selection. StrataSearch does this 1000 times, sorts the annual returns from lowest to highest, and then places them on the Monte Carlo Analysis chart.

The worst return on the far left side of the chart tells you the worst annual return you could have expected, using all of these alternate portfolio selections. Likewise, the return on the far right tells you the best return you could have expected. The average return is in the middle.

In general, you should be wary of your results if your Average Annual Return is significantly higher than the Monte Carlo Average Annual Return. In such a case, a fortunate (lucky) portfolio selection may have created returns higher than what you can expect in the future.

Likewise, it can be considered a good sign when your Average Annual Return is lower than the Monte Carlo Average Annual Return. In this case, your portfolio selection may have been unfortunate (unlucky), and future results may have a higher likelihood of being better than what your portfolio-based results imply.

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

Re: Montecarlo return versus avg annual return

Postby George2010 » Wed Jan 19, 2011 7:03 pm

Pete,

If I understand how this works, a small portfolio selection from a larger number of buy signals may be unfortunate (unlucky) or fortunate (lucky), especially if you use "Alphabetic" ranking.

However, it seems to me that it may be worthwhile to try to find a better ranking formula than the presumably random "Alphabetic" choice. As I'm sure you know, some people use "symbol rotation" systems where the primary system component is a "ranking" formula. If one believes there may be some merit in the rotation approach, there may be times when you are specifically attempting to design a system where the Average Annual Return is significantly higher than the Monte Carlo Average Annual Return through the use of a custom ranking formula. When I see a system where the Average Annual Return is significantly lower than the Monte Carlo Average Annual Return, I've wondered if a better ranking formula than the default proc(close,10) descending approach would be worth pursuing.

Thanks,
George
George2010
 
Posts: 23
Joined: Thu Sep 30, 2010 11:31 am

Re: Montecarlo return versus avg annual return

Postby Schutten » Thu Jan 20, 2011 8:31 am

I agree as well that a big benefit would be a more dynamic ranking selection mechanism. Or at least results could be improved with using a good ranking algorithm.
The rotation strategies become more popular every day. This does not necessary mean that this is good, but I have looked at some and I think that there is something in there what would help Stratasearch. I already contacted Pete about it, but it seems that implementing a dynamic ranking system is very complex.

Regards,
Dennis
Schutten
 
Posts: 42
Joined: Thu Oct 25, 2007 6:17 am

Re: Montecarlo return versus avg annual return

Postby Overload » Thu Jan 20, 2011 10:03 am

Just to be clear, StrataSearch currently allows you to enter the customized ranking algorithm of your choice. Alphabetical is used by default when creating new Setups, and "proc(close, 10)" was entered into a number of the pre-installed setups, but the rank formula can be changed to any formula of your choice. And while it isn't possible to automatically test different rank formulas in the same way that you test different trading rules in the AutoSearch, you can currently enter variables into your rank formula and test a variety of parameter sets automatically.

The ranking feature is fully available for position entry. The same feature, however, isn't currently available for position exit, and that's why this feature isn't available as a tool for rotational trading. More specifically, the ranking feature can be used to enter positions when they are ranked in the top X symbols, but it cannot be used to exit a position when it is no longer ranked in the top X symbols.

While a rotational trading feature isn't specifically available in StrataSearch, one user provided a programming trick that can be used in some cases. Read the following thread for more information:

viewtopic.php?f=3&t=824

While the trick in the above thread may have some heavy overhead since it requires the specific coding of each symbol in the sector, it can be used for rotational trading in StrataSearch (both entry and exit).

George, getting back to your suggestion that ranking can help the portfolio selection, I agree. That is indeed the purpose of the Rank feature: to improve the portfolio selection. Nevertheless, I think one needs to be careful of putting too much emphasis on ranking as a supplement to the Entry String, since this can lead to curve-fitting. It's for this reason that a comparison of your Portfolio Results to your Monte Carlo Results is a good idea. If your Portfolio Results beat your Monte Carlo Results, that's great, and maybe even ideal. However, you should be prepared for the possibility that the Monte Carlo Results, not the Portfolio Results, are what your system may ultimately produce. In short, strive for the best portfolio selection, but be prepared for any portfolio selection.

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

Re: Montecarlo return versus avg annual return

Postby George2010 » Thu Jan 20, 2011 1:09 pm

Pete,

Thanks for the comments, they make sense to me.

George
George2010
 
Posts: 23
Joined: Thu Sep 30, 2010 11:31 am

Re: Montecarlo return versus avg annual return

Postby Dacamic » Thu Jan 20, 2011 1:26 pm

Portfolio bias is very good at creating both false positives and false negatives. Thus, I find it best to mitigate its effects during system development.
Steve
Dacamic
 
Posts: 457
Joined: Wed Nov 30, 2005 12:40 pm


Return to OneClick Searches and Scoring

cron