search

pasky / pachi

A fairly strong Go/Baduk/Weiqi playing program
http://pachi.or.cz/
GNU General Public License v2.0
514 stars 117 forks source link

Can countercap #50

Closed lemonsqueeze closed 8 years ago

lemonsqueeze commented 8 years ago

Has can_countercapture() only check for snapback instead of full-fledged is_bad_selfatari(). Faster, and fixes situations where the snapback also looks like a good nakade (doesn't seem to make it any stronger though =)

pasky commented 8 years ago

Depends on PR#49 - but you mentioned it doesn't need to?

lemonsqueeze commented 8 years ago

Yes, this one can be rebased earlier. Let me check ...

lemonsqueeze commented 8 years ago

Could be rebased earlier, but there are merge conflicts because tactics_undo also changes it. Might be better to leave it where it is.

lemonsqueeze commented 8 years ago

I wish i knew, all my attempts at using new can_countercapture() here failed, winrate dropped. One thing that would make sense is under the stones situations, maybe they're not so uncommon in moggy. Or maybe it upsets moggy's balance somehow (i guess there's always a chance opponent doesn't capture after taking snapback)

lemonsqueeze commented 8 years ago

Rebased on current master.

pasky commented 8 years ago

I see. That happens, but could point out to a possible issue with the code. No need to solve that now, but I think it'd be worthy an XXX comment at that place of the code with what you just wrote in your comment here, do you agree?

lemonsqueeze commented 8 years ago

Definitely, here it is. It's all the more puzzling that before the is_bad_selfatari() changes that allowed nakade without atariing surrounding group, situation was strict already if i'm not mistaken, pretty much all snapbacks were blocked. Might be worth testing again at some point, i think i was still play-testing multithreaded at that time, could have been hit by the mt bug.

pasky commented 8 years ago

Awesome, thanks!