Week 5 · Day 3 — Random Forests & Feature Importance

Theme: Machine Learning Tryouts 🤖⚽ Time: ~75 minutes Tools: Python, scikit-learn, pandas, matplotlib Goal: Upgrade to random forests, understand what stats actually predict wins, evaluate properly with cross-validation

Warm-Up: Why One Tree Isn't Enough

Yesterday's decision tree was good. But it has a weakness: it's overconfident.

Train it on one dataset, get one opinion. Change one match in your training data and the tree can look completely different. It's like asking one scout to judge every player — their biases dominate.

The fix: Ask 500 scouts. Each one studies a different random sample of games. Then take a vote.

That's a Random Forest.