The more the merrier. I think ultimately it will depend on the availability of computational resources. To get good statistics, you will need 100-500 polymers, but absolutely not lower than 30. You can also consider doing multiple simulations.

Thanks Abdallah... I was doing simulations of Poly Lactic acid which has 1803 atoms in each chain. I will take 30 chains then.

Sonu,

What are you trying to simulate?