How do I define a blacklist for DoS prevention?

Many hosting and web security companies offers nowadays protection against Denial of Service attacks, by giving their users the chance of defining a list of blacklisted subnets. My question is really basic. How could I know the subnets to be blacklisted before undergoing an actual attack. I guess post-attack blacklisting is not a great strategy, but even knowing the IPs of my future attackers is kind of unrealistic.