HHC 2017 RPL Programming contest information and results thread

[Claudio L.]

(09-18-2017 01:54 PM)David Hayden Wrote:  So although the directions say that you should match the input decimal as closely as you can, the scoring favors fewer terms over higher accuracy.

Yes and no. I actually came up with an algorithm that minimizes the denominators but still guarantees 1e-12 accuracy to avoid all penalty from error. Once I got to that point, my "pseudo-score" (using 0.6,0.99 and 0.71828 only) dropped from 6900 to 519. With that score, even a small 1e-11 error added 100 points, ruining my efforts.
I'd like to see how it all plays out with Gene's pool of numbers, but I found the weights to be quite balanced once you reach a certain level of optimization.