I have written some tests in #293 and the py.test fails on this particular example during the travis build.

 def test_mean_boolean_error():
  >       assert mean_boolean_error([1,1], [0,0]) == 1
  E       assert 0.5 == 1
  E        +  where 0.5 = mean_boolean_error([1, 1], [0, 0])

This test works fine locally as it should because mean_boolean_error([1, 1], [0, 0]) is indeed 1 . What could be the source of this error?

@norvig is this the expected behaviour?