Fewer parameters make better models.

Best Systems I have have over 70 nodes in the ranking system.
Best part: all equally weighted, not optimization! Either in or out.

Combine with the right buy and sell rules (not more than 5 alltogether!) and things get interesting :wink:

Best Regards
Andreas

When you select factors non-systematically, having too many factors is detrimental, which increases your likelihood of overfitting.

When more factors come from simply screening with simple and systematic methods (e.g., A.Y. Chen) or systematically randomly generated (e.g., Bryan Kelly), then more factors precisely reduces the risk of uncertainty, as Yuval's example above.