No k-mean for me, but Gauss helped for the last test ;)

I still can't get the last test working with k-mean.
Oh well...

No problem :)

Oh, now I see. Well, if you would mention the two last points in your first comment, I would not object. :)

It's not "would benefit". It's mandatory as kata best practice because without random tests:

• It's impossible to check if author solution is valid
• People can pattern-match inputs and pass the tests

And before you ask, no, hiding expected values is not a solution. That's security from obscurity, obscurity is fake difficulty and anyone who thinks that's a good way to have anti-cheese/anti-cheat should feel bad of themselves (and, unsprisingly, it makes us harder to debug our code, not easier, contrary to what you said in one of the posts below).

I mean, just look at every newer kata, they all have random tests. It's not about how you feel, it's about CW policy of kata best practices. So whether you agree or not it's still mandatory.

(Also, I'm beginning to raise issue for older katas because there are plans to revamp all the old ones in the near future.)

This kata would benefit from random tests, I agree, but I don't see it as an issue.

Needs random tests.

Make sure you've chosen language version to be Node v6 to get ES6 features. It's Node v0.10 by default for some reasons.

I tried to overcome this kata with a gaussian approximation, but sadly can't get the final tricky tests without "inproper finetuning".
I'm going to switch to a suggested k-means algorithm for my next try.

I expected advanced requirements, therefore started a more complex solution... found out midway that it was not needed, so I interrupted it. Unfinished solution.