Curve-fitted portfolios and portfolio optimization

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Curve-fitted portfolios and portfolio optimization

Post by alp »

When testing do you avoid curve-fitted portfolios completely, or instead admit some portfolio optimization?

By portfolio optimization I mean choosing parameter values based upon the actually traded portfolio, even though a larger portfolio or even completely different markets might be used to validate them.
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

Sometimes I "develop" (design and optimize) on a 200-300 market portfolio, but only trade a 100-200 market subset. The extra markets have the effect of giving the Suite of systems more price scenarios to chew upon. The extra markets also increase the ratio (#trades / #system_parameters) which gives me a warm tingly feeling of increased robustness & decreased probability of overfitting. I don't know whether it actually helps, but I do know for certain that it feels like it helps.

Some of the extras include overseas markets with complex regulations that I have absolutely NO INTENTION of ever trading (examples), some of them have uncomfortably low liquidity (however I choose to define liquidity), below the liquidity I would ideally prefer, some of them are unbearably difficult to get real time or EOD price data for, and some of them are actively, vigorously deprecated (despised, really) by my broker(s). But hey, they're real markets traded by real people using real money, so I may as well include their price histories in my development and testing. Throw as many different scenarios at the Suite as possible.

(reference)
alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Post by alp »

sluggo wrote:Sometimes I "develop" (design and optimize) on a 200-300 market portfolio, but only trade a 100-200 market subset. The extra markets have the effect of giving the Suite of systems more price scenarios to chew upon. The extra markets also increase the ratio (#trades / #system_parameters) which gives me a warm tingly feeling of increased robustness & decreased probability of overfitting. I don't know whether it actually helps, but I do know for certain that it feels like it helps.

Some of the extras include overseas markets with complex regulations that I have absolutely NO INTENTION of ever trading (examples), some of them have uncomfortably low liquidity (however I choose to define liquidity), below the liquidity I would ideally prefer, some of them are unbearably difficult to get real time or EOD price data for, and some of them are actively, vigorously deprecated (despised, really) by my broker(s). But hey, they're real markets traded by real people using real money, so I may as well include their price histories in my development and testing. Throw as many different scenarios at the Suite as possible.
I sort of do the same way even though I try to get a feel of the character of the markets I will be trading and try to respect them, by optimizing on a portfolio of markets of similar characteristics. Simultaneously I compare results and parameter values in different portfolios and scenarios, by adding extras, "weird" or "wild" markets, especially markets which have proven quite difficult to trade using the same system or parameter sets, in order to detect signs or clues of overfitting.
Post Reply