By the way, maybe my algo is wrong? I calculated the distances between i-th and (i+1)-th poitns and the distances between i-th and ((i+n/2) mod n)-th points if n is even or between i-th and a middle of segment, which is constructed by ((i+(n-1)/2) mod n)-th and ((i+(n+1)/2) mod n)-th points if n is odd. If these distances are equal, then "YES"
Oh! You are right, my program outputs "YES" on some special tests, but the answer should be always "NO"! Thank you!
Now I've got AC. Simple algo with checking if points lie on a circumference and Point[i].x = x0 + R*cos(a0 + 2*k*PI/n), Point[i].y = y0 + R*sin(a0 + 2*k*PI/n) is true for sure. I just wanted to implement my "exotic" algo there :)))) But it was wrong...
I think that most Ac are by the chance, varing eps. If impossible to solve from the fist submission the problem is incorrect. I think right uderstanding the problem is following: Let P=<P1,...,Pn>-given n-polygon and r(P)=min(max dist(Pi-Qi),on all ideal n-polygons Q=<Q1,...,Qn>) if(r(P)<=1.e-5) YES else NO. That is optimization problem.
P.S. I glad to say that under above consideration I got AC immidiately, while using chaotic eps-using had 12 WA 18.