You can solve it yourself from scratch in a few minutes by running my Fhourstones benchmark [1]. Other interesting connect-4 tidbits may be found on my connect-4 page [2].
If you tell the optimal algorithm to only think 3 moves ahead, it's too easy, and if you let it think 4 ahead it becomes very difficult (but not quite impossible).
I've thought about winning algorithms for this game before. The best I could come up with is roughly "get the highest locations you can while still defending against 3-in-a-row".
Basically when it’s the computer’s turn it evaluates every possible move and assigns an attack and defense value for each move. The attack value is based on how many in a row and whether it can win on next turn, and the defense value is how many opponent’s chips in a row and whether it will lose on opponent’s turn. And if 2 or more moves are equal, favor the move closest to the center of the board.
I think first is an advantage since the opposite side has to react to your moves, a bit like chess.
A strategy that works for me is building a tower. In that tower try to construct 2 rows of 3 length, can be one horizontal and one diagonal but doesn't have to be, that when you go to one side of the tower, it would give you 4 in a row 2 times right after another circle is placed on top of the first one. While doing this also making sure the opponent isn't doing the same below your rows.
Doing this makes it impossible for the other to defend as when you're building on the side of the tower, it gives the opponent 2 options:
- Defending the first row and placing a circle, allowing you to place another on that circle and making the second 4 in a row.
- Or not defending and you can obviously finish your first in a row.
Following this strategy I was able to win both as player 1 and 2.
The fastest I could beat this AI was going first and playing 7,4,4,3,6,7,5. The computer always plays the same moves and has to block a connect, but gives up a connect.
I've found that the only way to win is to go first, just like in tic tac toe. Going second can produce a draw at best, against an optimal opponent. This is just my experience though, not a strong assertion.
This is not true, I've played twice as the second player and won both. The seem to likes setting up 3 diagonal, during that you are free to setup your own lower trap with two intersecting diagonals. So far doing that always seems to win.
Basically when it’s the computer’s turn it evaluates every possible move and assigns an attack and defense value for each move. The attack value is based on how many in a row and whether it can win on next turn, and the defense value is how many opponent’s chips in a row and whether it will lose on opponent’s turn. And if 2 or more moves are equal, favor the move closest to the center of the board.
It's fairly difficult, but not perfect, and you can win as the second player.