Improvements using Expectimax

[SafeMemoryZone]

Your current minimax implementation assumes that the dealer plays optimally, which prevents it from exploiting suboptimal behavior by the dealer. By switching to an algorithm like expectimax, you can achieve significantly better results in the long term. However, keep in mind that alpha-beta pruning does not work with expectimax, so you’ll need to devise your own optimization techniques.

This is only a suggestion of course.

[redstone59]

While the algorithm itself is interesting, I don't see the harm in re-searching whenever the dealer plays non-optimally. I know it's not computationally efficient, I still believe minimax with alpha-beta pruning would be more efficient than expectimax (however, I haven't really read into it too much!)

If you want to make a fork and try writing it for yourself and benchmark it, be my guest, but I haven't really updated this bot since the initial Steam release, and it's not even fully compatible with it yet, so you'd have a bit of extra writing to do.

[Anna-Spanner]

Your current minimax implementation assumes that the dealer plays optimally, which prevents it from exploiting suboptimal behavior by the dealer.

As far as I'm aware the dealer isn't a smart/learning AI. it is deterministic following specific logic, no? So useing expectimax would be an excess usage of resources whichn would decrease the accuracy