• Remove Haskell tag
• Clarify a range of possible arguments which might be assumed as correct
• Add more tests, including those with wrong arguments, not only with an id and different lists, but with non-list (like 2>>=id) and with more complex function.

What is supposed to be returned on (1,-1) or (-1,1)? I thought it might be zero, according to kata's description. Why am I wrong?

Might O(n^3) be a complexity of your algorithm because of tails (producing O(n) lists), map inits (inits for each list produces O(n) sublists) and map sum that sums each of O(n^2) sequence obtained after previous operations?

If it is, this solution is not efficient as it could be.

Could you argue why your solution is more efficient that this one, especially considering that GHC (and others) increases its ability of optimization every year.