PDA

View Full Version : Fitness Function


DeltaLover
12-16-2013, 09:21 AM
One of the most critical components in the development of a trading system (TS) is the creation of the fitness function (FF). The fitness function allows us to compare the effectiveness of two distinct betting strategies and select the one we think is best, by using a simple numerical comparison. The proper algorithm to be used for the FF should consider many more attributes of a TS in top of its final PNL to allow for more accurate comparisons. Max draw down, daily fluctuations, win-lose votes and just a few of the parameters we need to consider.

Using any form of pattern recognition or a finite state machine to spawn trading signals, I have found that the role of a good FF extremely important, especially to manage the related risk properly.

I am curious to see if anyone in this board has any interest in this topic of has already some ideas to present and discuss.

DJofSD
12-16-2013, 10:03 AM
That topic is new to me.

I found this: http://en.wikipedia.org/wiki/Fitness_function , however, there is not anything relevant to finance or investing.

Do you have any links that are germane?

DeltaLover
12-16-2013, 10:32 AM
This link is up to the point. A FF as I said above answers to the question of which of two models is better. What is important in this case are the criteria to use to reach the final decision. It can be something as simple as selecting the model with higher return or something more complicated considering more parameters, like the maximum draw down or the win / lose ratio. Forming a proper FF will allow us to create better models as we will be able to describe their behavior in more objective manner.

RaceBookJoe
12-17-2013, 06:47 PM
Trading models/backtesting is great to test the actual criteria of the "system" you are testing. It does a good job of that but where it stumbles a bit is that the modeling/backtesting cant take market sentiment/psychology and any number of current issues going on. So while a trade may setup with your criteria, whatever that is, but your trading accuman may say " no, not this time ". Just something to keep in mind.