Simple bitwise operator
number & 1 === 0 // even
number & 1 === 1 // odd

The function name could be something unique, rather than the name of a pre-existing builtin function. Working in IDEs will give unintended result.

I am very curious how your solution passes tests on your PC with the same data if it does not even solve the right problem?

Your solution would work with regular Python lists (well, almost, it would fail for some cases), but it will not work for generic linked lists.

Your solution tackling wrong problem is not a valid kata suggestion.

Code passes all pytests on my PC with the same data, but doesn't pass even one test on codewars.

It's becaus I don't take in account that:
"The linked list is defined as follows:
class Node:
def init(self, data, next=None):
self.data = data
self.next = next

Function should return root node, but in description not specific that.

nice kata, had me searching and overthinking before coming to a simple solution (clojure)

description amended

I rewrote the JavaScript tests suite, as well as enabled Node 18 + chai assertions. the test suite now includes edge cases (almost-sorted arrays, sorted arrays of length 2)

the error comes from your code, at this line:

int direction = (array[1] - array[0])/abs(array[1] - array[0]);

if array[1] == array[0], this divides by 0, thus raising a SIGFPE

as of now, TypeScript 4.9 is the only version available

duplicate of this issue

fixed

re-raised there as an issue

re-raised there as an issue

https://docs.codewars.com/training/troubleshooting/#c-c-signal-11---sigsegv

for some inputs, control flow can leave your function without returning any value, so your code effectively returns random memory addresses in those cases. this is visible if you add return "???"; at the very end of your function