While my approach passes the standard tests it fails often in the random tests. The test suite claims that for example for the difference of
A-B = [9, 12, 1, 1, 77, 34, 1, 1, 0, 73, 20, 25, 60, 42, 28, 24, 10, 11, 30, 35, 36, 0, 33, 18, 51, 26, 0, 51, 13, 25, 13, 10, 49, 8, 5, 25, 64, 12, 42, 1, 13, 57, 43, 0, 1, 39, 12, 10, 18, 1, 27, 1, 31, 7, 71, 61, 27, 38, 15, 42, 1, 11] the optimal solution is 21 thieves for k=77.

As far as I can see from solving this by hand and with a binary division and greedy approach it can be achieved with 20 thieves without breaking the rules. This also corresponds to the heuristic for the lower bound of the sum(differences)/k which is 19.5. Am I missing something or are the random tests only sometimes using the optimal solution as K01egA hinted at?

Yes, that was not clearly specifified and could only be derived from the example. To make sense as an outlier-rejection algorithm the order of evaluation must not matter. We therefore only consider the original data as a whole.

Oops, there was an off-by-one error that is fixed now. So the 3 is cut now correctly.
Thanks for raising this issue!

Thank you for the feedback on the beta of this Kata! The initical code was indeed necessary as part of the specification, so I changed the description now that it can stand alone.

I find the description still confusing. If I naively think of paste errors why can the output still contain "BBBB!BBBB!"?

Ok, fixed that as well. I somehow interpreted this as number of rolls per side, not total for all sides of the die.

I have added a show method to the test suite. The behavior should now be closer to the Java version.

This fun Kata deserves more translations. I put together a Python translation and would be happy about comments and feedback.

For this nice Kata I added a Python translation and the corresponding tests. I would be happy about any comments, improvements, etc.

I thought of this as well and found that dice can have two faces (https://www.scienceabc.com/eyeopeners/many-faces-can-die.html) but that depends if you follow their argument.

In any case, thanks for the feedback, I will work on some translations of the other Katas now. For the next one I will aim to be much more specific about the input and output parameters.

Especially the first one is quite similar. I must admit before authoring this Kata I did not really search for similar ones and then only in Python.

My focus was on the aspect of the transformation from the input data into a vertical histogram, so the third one is quite different. The best way forward is probably to add a translation to the first two. I don't think I can make this one unique without simply making it more complicated/specific.

Thanks again for the feedback, this is my first Kata ever.

Thank you for the suggestion! I added a test case so that it is easier to figure the rules for white-space. Now the user can see from both the example output in the description and the test case the width of the columns.