Given sufficient notoriety and money involved, it would be possible to just hire someone to essentially run a training model of alpha zero against moves specifically selected to be likely to be made by Magnus, and then all you really need is memorization of key scenarios (which for a good chess person should be no problem) to identify the right move to make.