Can you beat my Connect4 bot?

<< < (3/7) > >>

SCM:
What is Minimax and A* searching algorithm?

na_th_an:
http://www.aihorizon.com/essays/basiccs/trees/minimax.htm

http://en.wikipedia.org/wiki/A_Star_Search_Algorithm

wizardlife:
Quote from: "na_th_an"

Minimax and A* searching algorithm. That owns and can make your program unbeatable.


Yeah, except that it needs to run on an RCX: 16 MHz and 32k total memory. More like 16k practical once you download the kernel and program.

The way this works is it evaluates the value of the board that each new piece represents by only checking for groups and threats around that, and so creating any number of multiple pieces in a row is 'good', as is blocking the opponent's muliples... but it only counts them if one end of it is 'open', otherwise there's no point.

Agamemnus: Good plan, but I don't have the processor power for that kind of analysis on this platform.

SCM: I know about the 1432 win and it's varients. I thought I had that corrected... perhaps I posted an older version by mistake. The problem before was that it was still creating child nodes off of the win situation... so it considered letting you win worthwhile since that would let itself win the next turn.  :roll:  I'll check that on the current one and post back.

Dj Dennie:
You want your connect 4 bot to win every time? Well, you're lucky that somebody managed to figure out a perfect solution to it. This means that the player who goes first can always win...

Have a look: http://www.tomski.com/archive/000092.html

or here's a direct link to the pdf document: http://www.tomski.com/archive/connect4.pdf

wizardlife:
Quote from: "Dj Dennie"

You want your connect 4 bot to win every time? Well, you're lucky that somebody managed to figure out a perfect solution to it. This means that the player who goes first can always win...


I've looked through Victor Allis's stuff already. It's great for a human player, but completely impractical to try to program. There was only one Javabot that I saw that claimed to be using that system, and source wasn't provided. Anyhow, here's a slightly updated version. Here I'm showing what 'values' the program is establishing for the moves. (of both players)

http://www.angelfire.com/wizard/pigeoncarrier/Connect4a.exe

Navigation

[0] Message Index

[#] Next page

[*] Previous page