Is it possible to set up experiments that would be completely reproducible by seeding all random number generators and fixing spawn points so that the bots would act exactly the same each time given the same settings? I was having this conversation with my supervisor so I just said I'd ask if you think it's possible. I was thinking myself that you could get close to the same performance but that there is too much variability to reproduce the game exactly. The purpose of this would be to reproduce past games and performances on different machines.

Any insight would be great.

Thanks,

Frank