Comment by cmonnow

it's called bias vs. variance tradeoff, or over-fitting, in stats/machine learning lingo.