The reference solution incorrectly handles too many decimals (the result of the last test is always wrong).

Too few random tests.

NUMERALS should be explicitly provided to the user, and not hidden in the preloaded.

Reference solution should be a part of the fixture.

I added extra cases that the previous solution would fail for:

    Test.assert_equals(alphabet_war('jztwj'), "Left side wins!")
    Test.assert_equals(alphabet_war('twjzt'), "Left side wins!")
    Test.assert_equals(alphabet_war('jmtsj'), "Right side wins!")

Awesome thanks!

Thanks for your feedback. I've added some randomized tests. Tell me what you think of it.

Thanks! If you don't mind, please provide more test cases, with some randomized. I know that this is a exact translation of whats already there, but this is an old kata and the existing test suite is not nearly complete enough.

On top of the least candidates, candidates that received no votes at all should also be removed.
Also I didn't have to manage the complete tie.