Ml

My test set was a replay of my training data

While training a model for my PhD, I found my "test" set was a replay of part of my training data. The fix was a one-line change. The part I can't explain is why the honest numbers came out better.