If PLAY were automatic you would not be able to recall more than just back to the last move. Unless you would throw significant amounts of memory at it, which is better used otherwise 
Using two modes of PLAY however could be an option, in theory at least. I would have to evaluate if it is feasible in practice though.