A study in ScienceDirect argues that slave-making ants prefer larger, better defended host colonies

Slavemaker colonies showed increased raiding activities when the slave to slavemaker ratio inside the slavemaker colony was low. Slavemakers did not favour host colonies with more pupae, but preferentially attacked colonies with more workers. These represent riskier raiding targets, but as larger colonies usually contain more brood in the field, the increased benefit may necessitate fewer raids, decreasing the total risk during a raiding season. However, confronted with two host colonies that showed more distinct benefit to risk ratios, their decision shifted. Thus raiding behaviour and decision making in P. americanus are affected by a combination of external and internal stimuli.

The simple formula looks like: fewer attacks on other colonies results in less risk.

The attack events chosen then are ones estimated to have the highest potential reward per attack — enslave the strong. Makes sense to me if I accept that enslaving the weak brings more risk/cost than reward.