20 replies to this topic

[souske88]

I fell asleep during my 4th tournament! There are quite a few good guides to beat tournments 1-10, so if a program were to do that...(Tourny 11+ may be too hard).
Of course there is the problem of what if the opponent makes a weird move that is unexpected? Dunno, thought it may be a cool program, the trophy isn't bad looking either.
Just a suggestion

[jcrdude]

I'm sure that with the right generic startegy, a program could make it a decent amount of the way.

With REAL logic, though... we might well have a program that could go into Level 11 and beyond... the problem is coming up with that advanced logic ><

[Oaken]

I'm sure we used to have an auto player for Cellblock

I'm pretty sure it's possible as people have made topics about it in the past....and from what I remember, nobody said it was impossible.

[ShadowLink64]

Maybe I'll take a crack at it if I have time soon. Algorithms are fun to devise sometimes, lol.

.... All Cellblock is is "Connect Five"?

[jcrdude]

Maybe I'll take a crack at it if I have time soon. Algorithms are fun to devise sometimes, lol.

.... All Cellblock is is "Connect Five"?

Yeah... connect five... but without silly "drop to the bottom" limitations.

I'm thinking of the logic involved... and basically the key to victory is to get a row like this

xoooox

Without getting blocked or letting the other player get it first. I'm trying to think of the best way to approach this...

[ShadowLink64]

I'm reading a masters thesis on how to win connect 4 right now. It's fricking long, and I somehow think Cellblock might be easier since the pieces don't drop.

[jcrdude]

I'm reading a masters thesis on how to win connect 4 right now. It's fricking long, and I somehow think Cellblock might be easier since the pieces don't drop.

Easier/trickier... you gotta keep your eyes in more places...

Instead of looking for win scenarios in 8 squares (only 1 per row)... you're looking all over. I think it comes down to who has the better logic... and I would hope that our computers could beat their computers

[ShadowLink64]

Yeah. I've abandoned looking at connect four strategies and am looking more at these kind now:

http://en.wikipedia....wiki/M,n,k-game

Cellblock has the parameters: m=10,n=10,k=5.

The best place to find algorithms are probably in computing science journals written by researchers.

--------
Alright, so theoretically, according to this m,n,k theory stuff, since we play first we should be guaranteed to win if we adopt a "perfect" strategy.

More shit to read: http://en.wikipedia...._in_a_Row_(game)

[jcrdude]

Interesting read... I'll go ahead and research this while I'm doing nothing as well

[ShadowLink64]

Ok, well, it looks like I'm officially building this program.

[therato]

Ok, well, it looks like I'm officially building this program.

YAY we can use it:)

Edited by therato, 30 August 2009 - 03:34 PM.

[jcrdude]

Here's the guide most people use for cellblock and from personal experience it works almost all the time:
www.jellyneo.net/?go=cellblock
They aren't kidding when they say it works 98% of the time.
This one also works, but I've never used it personally:
neopets.com/~Kiquin

A feature you might want to include is an option to try a win/lose strategy. You could suggest simply: Win once, lose once, repeat. Or Win twice, lose once, repeat (which I used). Or (tell me if this doesn't make sense) lose the minority of rounds, then win the majority - so if there are 7 rounds, instead of just winning the first 4, you'd first lose 3, then win 4. Or if there's only 3 rounds, you lose once and then win twice.
I know from personal experience and from what avatarlog says about it, that consistent winners do not get the avatar. Rare loses aren't enough. This is why it'd be a good idea to have the option for those who just want the avatar, to win some and purposely lose some.

I'd prefer, for the safety of round continuance to go win the amount you need to win -1, lose the same quantity, then go for the round win... If you're trying to win all the rounds straight... there's always a chance that the opponent will make a random move that will screw up your strategy. I'm currently testing out the DAILY strategy. Theirs seems ALMOST water tight... and I'm logging results. If I get too many losses, I'll switch to the JELLY strategy.

[souske88]

Nice to see this topic running
I used the daily strategy for my wins, I find it easier to understand (or maybe it's because I've been using it too much!). Of course the less moves, the better, but the odd time that your opponent will make a move that will screw up that strategy all together...then you'll really have to use your brains (unless you want to just lose and start over again).
Could a program think on it's own once the original plan is blocked off? I think that's the big question...
Also, tournaments 11 and onwards (from what I heard) deals with more complex strategies... I've never been able to fully test them out because tournaments 1-10 take so long to complete!

[jcrdude]

The problem with tournaments 11+ is that there's no stones on the board and so the starting point is completely different. That petpage up there lists a strategy for those tournaments.

No starting stones... plus rocks... I never bothered learning actual strategy until now...

But I think I'm getting better?

[souske88]

Oh wow, where you start out in tournaments 11+ is random? Never knew that...
The program could just do up to tournament 10... but I guess if you want to make the program utterly perfect...

[jcrdude]

I just got a weird ass glitch... it let me go two turns in a row... It was pretty bizarre...

I counted the stones and I definitely had 1 more than he did... not complaining because it gave me the win... but damn that's bizarre

[jcrdude]

By the way, here's my set of games for the day, SLink... hope this helps towards program completion... shows the method's strengths and weaknesses

All moves are coded as xy;xy. First set is my move, second is opponent's move. In parentheses, I note various situations.

"Open 3" indicates that I have an open 3 on my turn and that victory is guaranteed, thus no further move logging is necessary

Any other notes indicate what the enemy is doing in case the strategy needs to be adjusted

Spoiler

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;7,2
6,7;3,4
7,6;4,3 (open 3 achieved)

7,4;4,7
6,3;6,4
8,5;9,6
5,4;5,2
4,5;3,6
6,7;7,8
7,6;5,8 (open 3 achieved)

7,4;8,3
6,3;4,7
8,5;9,6
5,4;5,2
4,5;3,6
6,7;3,4
7,6;4,3 (open 3 achieved)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;7,2
6,7;6,1 (open 3 achieved)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;3,6
6,7;3,4
7,6;4,3 (open 3)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;10,6 (weird placement noticed, plan changed)
7,6;7,2 (open 3)

7,4;8,3
6,3;4,7
8,5;9,6
5,4;5,2
4,5;4,5 (open 3)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;3,7
6,7;3,4
7,6;4,3 (open 3)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;7,2
6,7;3,4
7,6;4,3 (open 3)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;4,4 (he has an open 3)
3,3;4,3 (blocked; effective blocking is to continue plan, but plan is ruined; rethinking plan slightly)
4,5;3,6
3,4;6,7
2,3;1,2
3,2;7,7 (might as well go for the 4)
3,1;8,8 (I missed his 4 in a row; loss)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;7,2
6,7;7,8
7,6;8,7 (open 3)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;7,2
6,7;3,4
7,6;4,3 (open 3)

7,4;8,3
6,3;4,7
8,5;9,6
5,4;5,2
4,5;3,6
6,7;3,4
7,6;4,3 (open 3)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;7,2
6,7;3,4
7,6;8,2 (open 3)

7,4;4,7
6,3;5,4 (plan screwed)
8,5;5,2
7,5;9,5
7,6;7,7 (opponent has open 3)
8.8;4,4 (opponent has 4)
3,3;4,3
8,6;8,7
6,7;4,5 (opponent has chance at 5)
4,6;4,8 (open 3)

7,4;4,7
6,3;8,3
8,5;9,6 (weird move, continuing)
5,4;10,5
4,5;7,2
6,7;9,9 (open 3)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;7,2
6,7;3,4
7,6;4,3 (open 3)

7,4;4,7
6,3;8,3
8,5;9,6
5,4;5,2
4,5;7,2
6,7;3,4
7,6;4,3 (open 3)

7,4;4,7
6,3;8,6 (odd move, continuing plan)
8,5;9,6
5,4;5,2
4,5;7,6 (4 in a row)
10,6;(no move detected... taking second turn, open 3)

7,4;8,3
6,3;3,8 (odd move, continuing)
8,5;9,6
5,4;5,2
4,5;3,6
6,7;3,4
7,6;4,3 (open 3)

7,4;4,7
6,3;7,7 (open 3 for opponent)
8,8;4,4 (4 in a row)
3,3;4,3 (resuming plan)
8,5;4,5 (Chance for 5, open 3)
4,7;4,1 (Chance for 5 again)
4,2;6,8 (open 3)

7,4;4,7
6,3;8,3
8,5;1,2 (odd move, open 3)

Edit: Lucky you, indeed, Ange

Edited by jcrboy, 30 August 2009 - 05:59 PM.

[Cory]

Cam makes me so excited when he talks about algorithms.

[jcrdude]

Whoo! Finally got around to it. Been putting that off for a while now

Me too... and bout time

[souske88]

Congratz on the trophy! Amazing luck there lol
I don't think I'm oging to play that game any time soon... too repetitive xD

Oh, you are going to make this a reality, Shadow? Appreciate your (and also anyone who helps with this) hard work!

Edited by souske88, 30 August 2009 - 09:00 PM.

[iargue]

Shadowlink is (Unless he says fuck it) using Alpha-Beta pruning.

Which basically he looks at a series of moves, and discards them based upon their level of evaluation. Its similar to normal move watching, but when it reaches a point where the move is worse then previous ones, it stops there.

This way we can be fast....and unbeatable.