Closed ArjunSahlot closed 3 years ago
Please do a ELO test between this branch and v1.0.0
Out of 102 games, v1.0.0
won 35, this branch won 33, and they drew 34, resulting in this branch having an ELO of 7.8 less than v1.0.0
. But, I think they are roughly the same because there might have not been enough test data and the random move done at the beginning of every game might have worsened one of the engines' chance of winning.
Describe changes
Add a pseudo-random number generator. Random numbers are mainly needed for hashing, and they don't have to be very good. They just need to cover a wide range and have a roughly even distribution. Fixes #79
Testing info
Builtin C++
rand()
: Multiple collisions at 1M random number generations. NewRandom::random()
: No collisions at 100M random number generations.More testing details here.