OJ Board

No it's exactly the same... Your solution holds holds the width and height fixed, whereas the other solution holds the coordinates of the top-left cell fixed. Exactly the same thing, different approach.

It's greedy. (i.e. O(N))

Yes it seems to be able to handle (1, 2, 3) cases correctly. I'll pm you my e-mail as you said.

I must be very unlucky, as I compared my program with a bruteforce implementation of the solution (3^n). I ran 100 random tests, perfect matches for each test (n = 13 for each test, otherwise the bf binary would take ages to finish). This is weird :\ And I don't think it's an off-by-one error in ...

I was wondering if someone could point out the error in my DP algorithm (I get WA):

let sum(n) be the sum of the candies #1..n of the input

let c(n, a, b) be the min difference (as defined in the problem) of the total candy weight of the 3 children, right before candy #n is given, at that point ...