Tic Tac Toe Coding Challenge

[chilly_charly]

Kevin,

Nice infos you found on the web ! The similarity between the RPS and the TTT chalenge is striking ! That's exactly the conclusion I reached experimentally after playing a while with the ref and random players.

I observe that you choosed a "secure" strategy in your own KP2A vi, preferring to reach a draw against  a serious opponent than winning against other weak players. Now you'll have to find a way to detect rapidly a weaker player... 🙂

Another possible way to rank the different strategies would be to evaluate the performances in problem solving. For instance, compare the various vi players against CCref, using the following initial configuration, 
 XX
O  O
O  O
 XX

 The result is very interesting 😉 While Bruce random player manages to win a few games (when playing second), KP2A and CCref always reach a draw. Never lose... but never win either !..  As a matter of fact, with the proposed configuration, the first to play should always lose against an advanced player (can you demonstrate this statement ? ;)). Means that KP2A and CCref are quite far from optimality 😉 😄

So there is a lot of room for improvements and I would like to see some contributions from other competitors. Christian ? Ben ? Tst ? Shane ? JLV ? What are you preparing in the dark ?

[tst]

---

@chilly charly wrote:

I would like to see some contributions from other competitors. Christian ? Ben ? Tst ?

---

You can all relax. I'm not preparing some secret algorithm which will blow all of you out of the way.

Unfortunately, I know that if I even look at this I will be forced to keep going and I don't have the time. I started doing some checks when Bruch initially brought this up but haven't even got to have a working version. Looks like I'm going to have to sit this one out as well.

[altenbach]

---

@chilly charly wrote:

Christian ? Ben ? Tst ? Shane ? JLV ? What are you preparing in the dark ?

---

I actually submitted a very dumb but simple and fast algorithm to Bruce a few days ago. It is so dumb that it does not even look at the pieces of the opponent. 😮 Maybe I can convert it to 7.0, remove the diagram and post it so you guys with the fancy algoritms can be proud of yourself. 🙂

At the moment, I don't have any time to work on these things but I have some ideas for much stronger tic DAQtac toe code... 😉

[altenbach]

OK, as promised, here's my first draft (LabVIEW 7.0, passworded).

Have your players chew it up and spit it out! 😮 Have fun. 😄

Be warned, I'll be back !

[chilly_charly]

Christian,

Thanks for this most interesting entry ! Actually, I didn't understand immediately what was going on there ! :D:D

OK, back to serious things now.

After some tweaking of the CCref player parameters, here is what I have obtained : the Deep Toe player. I believe it might well win against any opponent : Deep Toe never loses a game and scores

- 198/195 (first/second player) against the random player

- 162/182 against CCref

- 102/123 against KP2A

In the O-ring problem solving (see (and study...) my previous post), it even scores 200/200 ! 😮

Since it would definitely kill the challenge, I promize solemnly not to submit it to Bruce 😉 :D:D:D

[shoneill]

"What are you preparing in the dark ?"

Er, nothing, I have the light on, I swear.

No seriously,  I haven't had time to even implement a Q&D version.

I hope to get around to it, but I'm up to my (insert body part here) in work at the moment.

Still, lots of free Brain cycles are being kept active by the ideas I have.

Shane.

[chilly_charly]

I'm deeply sorry : there was a bug in Deep Toe. Here is a corrected version. Enjoy !  :D:D:D

[cosmin]

hi all,

chilly you made a tricky player, when it loses it does not make any move, so in tester the opponent end up losing the game.

shame on you

[chilly_charly]

Damn'it... je suis découvert ! :D:D:D

Message Edité par chilly charly le 05-05-2006 09:06 PM

[altenbach]

---

@chilly charly wrote:

Damn'it... je suis découvert ! :D:D:D

---

It really never had a chance on my tester:

Of course Bruce's tester allows even the following player code to always "loose", the minimalistic Deep Toe:

Message Edited by altenbach on 05-05-2006 01:35 PM