If there are an equal number of naughty and nice it should return 'naughty'. In the example you provided there is 1 naughty ('broke...') and 1 nice ('got...') while the other 2 don't fit the pattern. Without seeing your code I can't offer suggestions but it worked for me in Ruby.

Language?

Also, are you sure you are not modifying the initial array?

So Strings that start with the letters b, f, or k are naughty. Strings that start with the letters g, s, or n are nice.
and if there is an equal amount of bad and good actions, return 'naughty'.

why did this fail?
Testing for [got someone a new car, broke a vending machine, learnt a new algo, pet a cat, yawned]
✘ It should work for random inputs too - Expected: 'naughty', instead got: 'nice'

I think the exercise is bugged, remeber you said that all other strings are neither.
it passes 48 tests and fails 4