"If you are not moving forward, you are falling behind." (Nightingale)
I understand it as meaning that we must keep improving.
The improvement of EAs is always on the edge of the trap of curve-fitting, so
it is very difficult. I’m iterating to avoid falling into this trap.
・If small adjustments drastically change performance, reject
・Check whether the profit structure remains intact even when the parameters are shifted by ±10–20%
・Validate using past data divided across multiple periods
・What matters is not the best performance, but "works reasonably in any period."
Maintaining these standards properly and conducting backtests is
very important, I think.
Avoiding curve fitting—this may be the most crucial skill in EA creation
.
I’m thinking of creating something like an "Optimization Policy Document" or "Optimization Procedure Document."
Until next time.
× ![]()
