Feature request: bot simulating "Mirror Go" strategy

Hello,

I would like to suggest a bot that would play Mirror Go.
I suppose it might be very useful for training oneself against such strategy.

Best regards,
DonKebabescu

1 Like

Would you want it to always only mimic your moves or should it choose at some point to deviate from copying moves? I’m not saying I could program it, I wouldn’t have the time to try right now anyway.

Just to help the request along for anyone that might be interested in it.

I imagine if there’s a situation where a capture happens that breaks symmetry the bot would have to change strategy anyway or resign.

If you want something to mirror your moves, play against yourself (duh).

If you want to see what the bot does when you play mirror Go, just play mirror Go vs a bot (duh).

3 Likes

I think “a bot Playing Mirror Go Stratey” means a bot doing what a human does with the “Mirror Go” strategy: copying the opponent’s moves until an appropriate time is reached, then taking advantage of having done that".

It doesn’t mean “have someone copy all your moves”.

I does seem that playing Mirror Go against a bot is a useful available substitute, though “far from the same”.

2 Likes

The whole point of mirror Go is to not lose to someone else’s superior fuseki.

A stronger bot would have superior fuseki regardless and play elsewhere immediately, i.e. not play mirror Go at all.

A weaker bot could mirror the moves, but it would be useless to play a weaker bot because it’s weaker anyway.

Ergo: It doesn’t make any sense.

2 Likes

I think it should deviate from copying moves at the right point, but also not to soon.
Frankly - I do not know how it’s done but I imagine that there are some weights assigned to the possible algorithm choices. So the bot would prefer mirroring to some extent, and would break the mirroring depending on the weights set, that would also be dependent on preferable kyu level of a bot.
IMHO, the bot kyu-level should be adjusted to the expected skill of human player, that would be capable of studying countering this tactics.

Also, the moment of breaking the mirroring could be dependent on the color the bot is playing (for example, more/less consequent mirroring after Blacks tengen in the first move)

I think it makes some sense though. It is known that mirror-go is beatable, but for 2 digit kyu-ers it appears to be very difficult, and a tool for practice would be very useful.

2 Likes

It seems you don’t quite understand.

  • If you’re playing black against a weaker player, first of all - why would you do that?? But also - you shouldn’t play mirror Go.
  • If you’re playing black against a stronger player, you might want to play mirror Go to make up for your ostensibly worse fuseki. Your opponent is stronger, so you will also see what he does to ‘counter’ mirror Go. Playing a regular LZ will do the job here. Just play a bot.
  • If you’re white against a weaker player mirroring your moves, you’re ahead by komi, just finish the game. If you don’t like manego, play (near) tengen. You’re stronger anyway, it doesn’t matter.
  • If you’re white against a stronger player, first of all, what are you doing? But if your opponent is stronger, you’ll probably lose anyway and mirror Go won’t have had anything to do with that. However, you can try to apply what you’ve learned playing mirror Go against bots, or just play near tengen…

https://senseis.xmp.net/?MirrorGo#toc2

2 Likes

Great points by @smurph.

I actually tried mirror go against AI and that was a nice experience. It handles the situation in a more subtle way than human players would.

That’s not how mirror go usually works. It’s the one being copied that needs to find an appropriate way and time to break the mirror.

actually gnugo apparently has a mirror go mode according to the documentation

https://www.gnu.org/software/gnugo/gnugo_3.html