Allow fitness as function of multiple periods
Closed this issue · 0 comments
jameschch commented
Recent findings are that Sharpe ratio of an immediately antecedent period are a good indicator of expected in sample performance. Also, scoring a set of parameters based on the aggregate of two time periods will counteract period fitting. Minor changes will be required for a DualPeriodFitness to be possible.
It's not clear at this time whether multi-period fitness should naturally follow from this.