Why: Prevents a few super-fit individuals from taking over early (premature convergence) when raw fitness gaps are huge.

How: Rescales fitness using population stats so selection pressure adapts to diversity.

Common formula (with ()):

where is mean fitness and is the standard deviation.

Effect:
When is large → strong down-scaling of outliers → reduced selection pressure.
When is small (population similar) → little scaling → normal pressure resumes.

Pros: Maintains diversity; smooths selection pressure over time; simple to implement.

Cons: Adds a hyperparameter (c); if mis-set, can under/over-pressure selection; still needs elitism or other safeguards.

Use when: Early generations show big fitness variance or you observe premature convergence with roulette/fitness-proportional selection.