08-15-2020, 12:46 PM
AI are typically very strong in Trike , so I am 90% confident that there is somewhat of a bug with the AI when I say this, and it is not just that they "aren't good enough" ...
So, basically the computer has a very strong preference for the first move - lets just call it move A.
On small boards move A is known to be winning for player 1, and probably it is also winning for player 1 on large boards as well.
However, the same computer has a very strong preference to Swap this move, when meta (swap) is implemented, which it is in Trike.
It makes sense for the computer to swap it , and it makes sense for the computer to play it , seeing that it is a strong move - but it doesn't make sense for it to play it, if it considers the fact that the opponent might swap.
In fact, if it were considering the possibility of a swap, it should realize that the opponent could put it in a losing position by swapping from that position.
Thus, it should consider this move to be not such a good opening move, from the perspective that the opponent can swap and steal that winning position.
This is why I believe the AI does not properly handle the swap rule , perhaps it doesn't consider the possibility of swapping when considering which first move to play - rather , it just plays the strongest available move, which ends up being bad for it later on, when the opponent swaps.
I dont know if this is a high priority for Devs to look into, but it essentially puts AI at disadvantage if they are moving first.
So, basically the computer has a very strong preference for the first move - lets just call it move A.
On small boards move A is known to be winning for player 1, and probably it is also winning for player 1 on large boards as well.
However, the same computer has a very strong preference to Swap this move, when meta (swap) is implemented, which it is in Trike.
It makes sense for the computer to swap it , and it makes sense for the computer to play it , seeing that it is a strong move - but it doesn't make sense for it to play it, if it considers the fact that the opponent might swap.
In fact, if it were considering the possibility of a swap, it should realize that the opponent could put it in a losing position by swapping from that position.
Thus, it should consider this move to be not such a good opening move, from the perspective that the opponent can swap and steal that winning position.
This is why I believe the AI does not properly handle the swap rule , perhaps it doesn't consider the possibility of swapping when considering which first move to play - rather , it just plays the strongest available move, which ends up being bad for it later on, when the opponent swaps.
I dont know if this is a high priority for Devs to look into, but it essentially puts AI at disadvantage if they are moving first.