Why are polynomials of even powers better for Pollard's rho?

175 Views Asked by At

Taking all $C(900,2)$ combinations of the first 900 prime numbers, I defined $N = pq$, where $p$ and $q$ are a combination of primes. Then I factored $N$ using Pollard's Rho, counting how many iterations and how many successes were found, where a success is finding a non-trivial factor and an iteration is a computation of the GCD algorithm.

Notice how even powers have much better performance while keeping about the same number of successes. Can you explain that?

iterations  sucesses  polynomial
 784491     44081     x^2 + 1, x0 = 2
 831938     44098     x^2 + 1, x0 = 3
3380606     44505     x^3 + 1, x0 = 2
3176223     44511     x^3 + 1, x0 = 3
 566286     43757     x^4 + 1, x0 = 2
 581990     43765     x^4 + 1, x0 = 3
5912991     44646     x^5 + 1, x0 = 2
6050589     44622     x^5 + 1, x0 = 3
 486181     43630     x^6 + 1, x0 = 2
 507281     43707     x^6 + 1, x0 = 3
7514619     44717     x^7 + 1, x0 = 2
7884300     44727     x^7 + 1, x0 = 3
 528307     43822     x^8 + 1, x0 = 2
 528177     43809     x^8 + 1, x0 = 3
3139981     44455     x^9 + 1, x0 = 2
3027514     44430     x^9 + 1, x0 = 3
 552764     43801     x^10 + 1, x0 = 2
 568909     43823     x^10 + 1, x0 = 3
8623590     44741     x^11 + 1, x0 = 2
8463129     44743     x^11 + 1, x0 = 3
 400465     43396     x^12 + 1, x0 = 2
 401493     43316     x^12 + 1, x0 = 3
8497977     44733     x^13 + 1, x0 = 2
8047224     44761     x^13 + 1, x0 = 3
 588730     43910     x^14 + 1, x0 = 2
 634799     43911     x^14 + 1, x0 = 3
1686343     44145     x^15 + 1, x0 = 2
2091292     44238     x^15 + 1, x0 = 3
 479840     43619     x^16 + 1, x0 = 2
 506561     43579     x^16 + 1, x0 = 3
8919461     44766     x^17 + 1, x0 = 2
8673652     44747     x^17 + 1, x0 = 3
 448822     43469     x^18 + 1, x0 = 2
 461822     43562     x^18 + 1, x0 = 3
8782917     44766     x^19 + 1, x0 = 2
8466664     44768     x^19 + 1, x0 = 3