Fitness is based on a number of different things combined together to form your fitness score. Jared posted his notebook in the Alpha Stream Scoring Criteria and it explains the different criteria for fitness scoring.
Potential candidate fitness functions were judged by the following:
- Independent and Deterministic - functions should not rely on a distribution of community performance
- Positive Percentage - functions should have results ranging from 0 - 1.
- Generous Initial Skew - alphas in profit should achieve near 50% score.
- Statistically Significant Activity - Algorithm should be the cause of the performance.
So, if lets say the equity you are trading gets de-listed at some point and your algorithm stops trading that equity, your fitness is going to go down because any gains your algorithm will see is due to standard market movements and not your algorithm.
You can read further about de-listing events and how to capture the event by looking at this Delisting Events Algorithm example which is talked about in the References for Handling Data