My smartgo app does this thing at the end of the game; I'll finish playing and pass but the computer continues to fill in there own space, losing a point each time. It'll put stones in random places like out in the middle of an open space. Is this just the computer being stupid or is there a legit reason to do this?

_________________
Thinking like a go player during a game of chess is like bringing a knife to a gun-fight. Thinking like a chess player during a game of go feels like getting knifed while you're holding a gun...

Just the computer being stupid. There are cases where there is some weakness inside your territory that needs to be defended once the outside liberties get filled at the end of the game, but most bots do the stupid thing you observed as they only try to win, so if they are ahead by quite a few points they fill in their territory. This can be quite a problem when the bot is playing Yoda Norimoto 9 dan professional in a tournament as it's rather rude...

Sometimes I'll do it to my opponent if I'm ahead by a significant margin, the aji's run out, and the opponent won't resign.

Why not just pass in that situation?

1. It has more mockery.
2. If they then pass the game enters the scoring phase but it isn't yet finished.

Make sure that the program is set to play by territory rules. By area rules it loses nothing by filling in its own territory -- unless it kills itself.

_________________
The Adkins Principle:
At some point, doesn't thinking have to go on?
— Winona Adkins

Visualize whirled peas.

Everything with love. Stay safe.

I had no idea this was an option either. Glad I looked at this post!

Does this option actually exist? Most monte carlo bots only play area rules afaik (it's easier for them to count and establish group status, and doesn't punish their fill in). And as they only try to maximise win percentage not margin would it actually help?

From what I've heard from Mike Novack, Monte Carlo bots don't work well at the beginner level, so I don't think SmartGo is using a Monte Carlo algorithm. I also can't find any mention of Monte Carlo on SmartGo's front page or on the Sensei's Library article on the program.

_________________
We don't know who we are; we don't know where we are.
Each of us woke up one moment and here we were in the darkness.
We're nameless things with no memory; no knowledge of what went before,
No understanding of what is now, no knowledge of what will be.

I've heard that some go engines try to win by 0.5... I dunno if this is true or not.

And here I was thinking A.I. research was still millenia away from mastering human emotions!

This may also explain why computer go isn't particularly successful yet: the programmers are wasting their time making anthropomorphic sentient go playing entities, instead of focussing on algorithms to play better go.