Incomplete win structure of the minimax with alpa-beta pruning integration

[Alpengreis]

Hello

I have the problem as described on following wegpage:

https://roboticsproject.readthedocs.io/en/latest/ConnectFourAlgorithm.html

"
Incomplete win structure

During stress testing, it became clear that the algorithm would not make a winning move if there were two or more possible winning moves available. This is presumably because it could not decide between equally weighted branches, and therefore made the ‘next best’ move. This problem did not impact the algorithm’s success rate, however, because as soon as the human player filled one of the possible winning spaces, the algorithm would win the game using the other."

Is there some hope for a fix (for ex. randomly choose a winning variant)?

EDIT:
By the way: it's not only if there are multiple possibilities to win, as I found out ...

Kind regards,
Alpengreis