M3 · Twenty Questions
Random Forests
Build a forest and explain the extra randomness on features.
Resampling the rows is not enough. If one column is strong, it wins the first cut in almost every resample. So the trees come out alike, and a crowd of lookalikes votes like one voter.