Show HN: C4 – Connect Four, with AI

[kenrick95]

http://kenrick95.github.io/c4/demo/

[kenrick95]

How I build this project: It was mid-2013 when I wanted to apply what I've learned from the MOOC CS188.1x Artificial Intelligence offered at edx.org. Around December 2013, I joined a bootcamp held in my university (Nanyang Technological University, Singapore) in which some of the speaker came from Mozilla. Here, I learned a lot of stuffs on JavaScript (they also introduced us to HN). Later, I thought of applying what I've learned to a game. I then started the project and finished it during January 2014.

[atrilla → kenrick95]

Awesome! It's wonderful to see these kind of side projects for deliberate practice.

I also expect to build some of these at http://ai-maker.com/

Thanks for sharing it and congratulations!

[mrborgen → kenrick95]

Nice job, do you recommend the Artificial Intelligence MOOC? Is it fun?

[kenrick95 → mrborgen]

It's amazing. Although quite time-consuming, the projects are nice: it was building an AI for Pac-man to maximize its score in various scenarios.

The next offering is starting soon:

https://www.edx.org/course/artificial-intelligence-uc-berkel...

[smilliken]

We did a friendly programming competition at MixRank last year writing AIs for Connect Four.

It's really a great game for playing with AI strategies since it's so simple. You can borrow our game runner and write your own: https://github.com/smilliken/aigames/tree/master/connect-fou...

You can use any language since it just talks over pipes. There's a few bots in there you can play against.

[smilliken → smilliken]

Also, if you'd like some light reading, Victor Allis proved a winning strategy in 1988: http://www.informatik.uni-trier.de/~fernau/DSL0607/Masterthe.... It's a great read, I recommend it.

[RMarcus]

I wrote a C version of a connect 4 AI and used emscripten to build a JavaScript / AngularJS version. You can check it out for comparison: http://rmarcus.info/blog/2014/12/23/connect4.html

Edit --

Github is here: https://github.com/RyanMarcus/connect4

Looks like my AI beats your AI. ;)

[guidopallemans → RMarcus]

Just beat yours at 5 moves ahead

[RMarcus → guidopallemans]

I've been playing at 8 for a challenge. Can't win yet, but I can draw. I left it at 5 as the default because it works well on tablets. :)

[cl42]

Not to be a pessimist, but Connect Four is a solved problem: http://en.wikipedia.org/wiki/Connect_Four Wouldn't that mean the best AI should always win if it's first (alternatively, you should always win if you're first)?

[kenrick95 → cl42]

I'm aware of the fact that this is a solved problem. When I experimented by making it my AI vs my AI, it did not use the optimal first move. This shows that my AI is not perfect :)

[vapter]

Nice work! I was actually planning to do a similar project in the near future. However, it troubles me that I was able to win when your AI didn't "block" me when I put 3 aligned circles. Shouldn't that be the very first rule for such an AI?

[kenrick95 → vapter]

Normally it tries its best to avoid that situation, but sometimes your scenario happens and I did not know why.

[austinhutch]

Finally got a win, took me about 10 rounds. Really cool thanks for sharing.

[dirkk0 → austinhutch]

Had a win in the first round but that was pure luck. :-)

[philbarr → dirkk0]

Had a win in the first round because I played Connect 4 aggressively against my brother for years when I was a kid!

[mlamat]

I won in the 2nd round.

[lwgdhn]

Every once in a while a new game pops up on HN and I always pay the price in time. Not complaining!

[hitchhiker999]

Note to future readers: This is how we ended up training our new robot overlords wasn't it?

[bluehazed]

Wow, I really suck at Connect Four.

[kenrick]

Wow another developer named kenrick :)

[kenrick95 → kenrick]

Hi, another Kenrick :)

[noso]

Really nice work and very cool

[JamyDev]

Nice, I just beat it :)

[kenrick95 → JamyDev]

I admit that the AI is not perfect. :)

[tromp → kenrick95]

My Fhourstones solver and a perfect playing applet may be found at http://tromp.github.io/c4/c4.html

[unstabilo → kenrick95]

No need to apologize, even if it's perfect, the first player can always win by playing the right moves.

[stefs → kenrick95]

a perfect AI would be rather boring anyway!

[alacritythief]

I want to note that you can win every time by dropping wherever first, then on your next moves dropping every piece on top of the AI's pieces.

[inglesp → alacritythief]

Unless I've totally misunderstood, I don't think you're right.

For instance, on your first go, drop in column 0. Then, the AI drops in columns 1, 2, 3, 4, leading to this:

    . . . . . . .
    . . . . . . .
    . . . . . . .
    . . . . . . .
    . X X X . . .
    X O O O O . .

[alacritythief → inglesp]

It could have been fixed after my post, I won 8 times in a row and it seemed to be some kind of fluke.

After trying it now, the AI actually tries harder to beat you, and my previous method is not effective anymore.

[kenrick95 → alacritythief]

Just a clarification: The latest commit is around a month ago. https://github.com/kenrick95/c4

[wuliwong → alacritythief]

Interesting, this method has lead to the game freezing a couple times. And another time it seemed that the computer just skipped it's turn.

[ljk → alacritythief]

only got a draw by this method