[Community Puzzle] Atari Go 9x9

Yes here Coding Games and Programming Challenges to Code Better
Since i m rewriting mine aswell at the moment, the upcoming battles will be interesting my friend ^^

Ouch! That opening in the lower left is horrible. And your bot had a number of chances to connect your weak group to your strong one but didnā€™t take the chance in more places than the upper left, so I guess that I have to be grateful for that. :slight_smile:

Iā€™m looking forward to our next level of battles. I will definitely implementing some special code to handle cuts at the edge since that seems to be the best way of picking up stones.

Actually connecting groups i havent even considered until now ^^ If i find the time, i will try implementing that aswell. My current Programm takes the following steps:

  1. Get all my Clusters and its neighbors
  2. Look for a way to maximise the free space of the cluster with the least free space
  3. Get all enemy Clusters and its neighbors
  4. Look for a way to minimize/kill the free space of the cluster with the least free space
  5. Compare those two options and take the better one
  6. If neither is available, either play:
    6.1 Nothing if you are leading in score and all your clusters have at least two eyes and are safe
    6.2 A random move that chooses a relatively safe space

But i feel like that wont be enough ^^

Now I have my new infrastructure in place after fixing a lot of bugs. It seems that either itā€™s playing a liiiittle worse than before or you have improved because now Iā€™m number 2 instead of number 1 and youā€™re number 1 instead. It seems that your strategy is solid.

The good thing is that I see lots of things I can improve pretty easily so letā€™s see if we can both move on to league 2 and continue our battles there.

I m done for tonight, implemented a lot of stuff, what i m missing now is a good starting strategy if the enemy doesnt decide to play a ā€œbad moveā€ first. ^^ Learned a lot about my Bot through playing against you and the funny thing is even though I managed to get to League 1 now, my Code still lost versus you so you should be on your way soon! See you in League 1 my friend :slight_smile:

I will be outside today (Iā€™m currently in the Swedish mountains). Hopefully I will be able to do some programming this evening.

Seems I got promoted without having to do anything more. So here we areā€¦ :slight_smile:

Dont mind checking the profil of the guy who is in first place ā€¦ :smiley: that s a real challange man. So take it as a goal to get my second place, good luck :wink:

Yeah, that was quite impressive. But I had a top-3 go program in theaigames.com site before it folded. Go, mind you, not atari-go, which is a lot simpler. So, Iā€™m going to give it a shot. :slight_smile: I am also quite impressed with your second place in such a short time. You probably have captured the essence of Atari-Go quite well.

Actually, I think I can make a program that can never lose as black, since black will always play one more move if you just make a safe enough group. The trick is creating a program that can win as white.

I must say that I am a little disappointed. I was almost sure that in league 1, the board size would increase to at least 13x13 but hopefully 19x19.

Hm. The fact that the game plays for a fixed number of turns throws an odd wrench into it. For example, groups that should be dead may not be if they can survive until turn 80 without being captured. So, say one currently has a lead in score with 10 turns left in the game. One could theoretically setup a dead ladder that would take more than 10 turns to capture but that the opponent canā€™t ignore (because ignoring might allow you to capture some other large group.) Thus you lock the opponent into having to attack the ladder. And so you get to run out the clock by just extending the ladder.