Even simple algorithms almost often outperform expert judges, due to a reduction in noise.
Even simplistic models (using weighted averages) often are better predictors than expert judgments.
👉 Experts are noisy. Simple formulas are not.
This is actually probably a large part of the benefit of The Checklist Manifesto. Reduction of Noise when deciding what to do. That’s my note, not from this book. This same concept was also discussed in Thinking Fast and Slow.
You have to check that your algorithm isn’t ‘baking in’ prejudices (either directly or indirectly). See Math & Fairness.