We study delegation of stake to pools from a game theoretical angle. We argue that pools in PoS protocols are different from those in PoW protocols and derive a lower bound on the share of reward that the pool runner should keep for itself to guarantee the safety of the whole protocol in the presence of malicious players. The lower bound is surprisingly easy. It is equal to the (suspected) share of the malicious players in the whole system.
For the rest, check out the paper: