GA (Genetic Optimizer) is penalized?
Author: lepete
Creation Date: 6/22/2018 6:13 PM
profile picture



I really like the GA in WL Pro, and have been observing that GA results in some trading platforms are not penalized. For example, for 30 days of data for 1 min bars (from Fidelity), if I set a range of 2;20,1 for an EMA of lows and highs, and range of 0.01;5;0.01 for a price "delta" with a calculated trailing stop for lows and highs that send orders if the trailstop is breached, the WL results are quite good with many intra-day trades.

However, for some other platforms I have seen, that for the same asset, parameter ranges/intervals, and bars, that the GA will notice that price went up straight over the entire 30 days and make 1 trade on day 1 and exit on day 30. Isn't this a penalization issue?

In our own GAs for biological data we have seen that when genes 0 and 1 are used within chromosomes to identify which variables to use for a regression problem, say that 1/MSE is fitness, chromosomes with more 1s, i.e. regression variables used for the model, have better fitness only because you can predict anything perfectly with an infinite number of features. We never got back to this however, to determine how to penalize -- I suppose fitness = 1/(#features * MSE)?

At any rate, why would a GA for trading make 1 trade on day 1 and exit on day 30 since price went up continuously every day on a 1min chart?

profile picture


The GA developer has left the company. Sorry but I'm not familiar with the GA internals to the extent required to answer your question.
This website uses cookies to improve your experience. We'll assume you're ok with that, but you can opt-out if you wish (Read more).