hand_rank | Implements fast poker hand ranking algorithms in C | Awesome List library

After a lot of digging, I think I have found the source of the bug. To begin with, your logic was correct: by calling set(), you removed duplicates, so the result should be the same with or without duplicates.

In the original code, if the second test case is commented out, using set(hands) would pass the other cases (at least in my environment). Thus, the problem might be with the second test case.

If you isolate the second test case out, and run the program with set(hands) for, say 20 times, the tricky thing happens: most of the time the test case fails, but in a few times it actually passes!

This is crazy! The same code gives different results. The correct output for the second test case is ('TD', 'TC', '7C', 'TH', 'TS'). But the max function would also return at least two other outputs ('TD', 'TC', '7C', 'TH', 'TC') and ('TD', 'TC', '7C', 'TD', 'TS').

If you inspect the return value of hand_rank() when called on all these three hands, the results are the same (7, 10, 7). In other words, these three hands are identical in their ranking. Thus, which one to return as max depends on whether it is a list or set that is passed to a max function.

If a list is passed, as in the case of hands, the first occurrence of the biggest value is returned. In this case, the index of ('TD', 'TC', '7C', 'TH', 'TS') is 9081, index of ('TD', 'TC', '7C', 'TH', 'TC') is 9627, and index of ('TD', 'TC', '7C', 'TD', 'TS') is 9102. Therefore, when hands is passed, the correct answer ('TD', 'TC', '7C', 'TH', 'TS') is always returned.

If a set is passed, since there is no internal ordering, the returned value could vary, which is exactly the behavior we have observed. For further testing, you can try this: