A Game Against Chatterbot

tekesta · Post by **tekesta** » Thu Sep 12, 2013 3:52 pm

Hello again. I know that it's pointless to post games against gobots, but I could not get any human players on KGS and I had to settle for chatterbot. In this game I lost to chatterbot by 25.5 points. In the bottom right corner I got sucked into an unpleasant fight as a result of my invasion, created a big dead group (which, in hindsight, may have been saveable with the right combination of techniques) and had to go elsewhere to get profit. I was pleasantly surprised that I did not lose by a bigger margin. Much obliged in advance for your kind commentary.

Post by **Joaz Banbeck** » Thu Sep 12, 2013 7:11 pm

10: Joseki is C6 or D5. Either of those moves can kill C5 by black, which is the shape move to attack the corner. C7 is risky - especially when he has the ladder breaker at Q16

12: Not bad, but R10 is bigger.

18: R9 is huge. Remember, the longer you play in one location, the smaller the moves tend to get. ( IMHO, one of the biggest steps from DDK to SDK is learning to tenuki )

20: Right idea, just a bit risky. R9 is safer.

40: Good tesuji

46: Way too thin

Dinner is waiting...maybe more later.

EdLee · Post by **EdLee** » Thu Sep 12, 2013 9:37 pm

tekesta wrote:I know that it's pointless to post games against gobots,

It's not pointless, especially since you lost, by quite a margin.

Post by **Joaz Banbeck** » Fri Sep 13, 2013 6:30 am

With apologies to Ed, I disagree with one piece of advice that he gave.
Yes, as noted, 44 is a bad move. ( Known as the 'armpit hit', BTW. )
And, yes, the continuation with 46 at 'a' is way too thin. Move 46 at 'b' is better. Both moves, however, suffer from being too close to stronger black stones.

Click Here To Show Diagram Code: [go]$$Wc $$ . . . . . . . . . . | $$ . . . . . . . 1 . . | $$ . . . . a . b O c . | $$ . . . . . . X X . . | $$ . . . X . . . . . . | $$ . . . . . . . . . . | $$ . . . . . . . . . . | $$--------------|[/go]

If you want to continue playing around 44, then 46 at R6 looks best to me. ( Even 'c' may be playable )

-------------------------------

Going back to move 44: as an alternative, you might try this when approaching a 4-4 stone with an ogeima extension:

Click Here To Show Diagram Code: [go]$$Wc $$ . . . . . . . . . . | $$ . . . . . . . . . . | $$ . . . . . . . . . . | $$ . . . . . . X 1 . . | $$ . . . X . . . . . . | $$ . . . . . . . . . . | $$ . . . . . . . . . . | $$--------------|[/go]

It can be tricky to play, and it may depend on surrounding stones, but this is way better than 44 in the game.

Also, both of the following can also be reasonable, depending on the circumstances:

Click Here To Show Diagram Code: [go]$$Wc $$ . . . . . . . . . . | $$ . . . . . . . . . . | $$ . . . . . . . . . . | $$ . . . . . . X . . . | $$ . . . X . . . 1 . . | $$ . . . . . . . . . . | $$ . . . . . . . . . . | $$--------------|[/go]

Click Here To Show Diagram Code: [go]$$Wc $$ . . . . . . . . . . | $$ . . . . . . . 1 . . | $$ . . . . . . . . . . | $$ . . . . . . X . . . | $$ . . . X . . . . . . | $$ . . . . . . . . . . | $$ . . . . . . . . . . | $$--------------|[/go]

Mike Novack · Post by **Mike Novack** » Fri Sep 13, 2013 6:39 am

tekesta wrote:Hello again. I know that it's pointless to post games against gobots, but I could not get any human players on KGS and I had to settle for chatterbot.

Why do you think pointless? Let me ask you (and everybody else seeing this game) a question. If you didn't know and if you hadn't told us that "chatterbot" was a bot instead of being a human do you think it would have been obvious? Can you point to moves in this game made by your opponent that shout "clearly no human would play that way; must be a bot".

If a go playing bot can pass the Turing test (if we can't tell whether a human or a computer) then your reasons for saying "pointless against a bot" are mystical, not rational.

My personal answer to the question -- your opponent played like I would expect an opponent who was several stones stronger than you to play and nothing stood out to make me think "not a human".

My judgement of the situation -- at your current level of play don't worry about not being able to get a human opponent if this bot is available. You can learn from this bot until you begin winning some games against it (say 1/3 of the time).

Understand? There are bots and then there are bots (different behaviors). There are ways we tend to play and there are ways other humans tend to play. Thus there are interactions between ourselves and a bot against which we are playing which determine whether will be a good or poor learning experience. Because this game looked like two humans playing, one stronger than the other, a good fit for now. If the bot against which you are playing becomes too consistent (too predictable) time to move on to another that works for you.

Boidhre · Post by **Boidhre** » Fri Sep 13, 2013 9:03 am

Playing a bot you can beat easily over and over again is very bad because it'll just reinforce certain bad moves that you play that the bot doesn't "know" how to punish/counter/whatever. It's the same with only playing against people who you can beat very easily who again can't refute your bad moves.

Playing bots that kick your ass on the other hand is useful because the bot very obviously can find the holes in your play.

EdLee · Post by **EdLee** » Fri Sep 13, 2013 10:49 am

Hi Joaz, I didn't notice any disagreement.

Which part(s) are you disagreeing with?

Post by **Joaz Banbeck** » Fri Sep 13, 2013 1:29 pm

EdLee wrote:Hi Joaz, I didn't notice any disagreement. Which part(s) are you disagreeing with?

For move 46 you suggested Q5. I preferred R6.

EdLee · Post by **EdLee** » Fri Sep 13, 2013 7:53 pm

Hi Joaz, I checked with a pro; the opinion is either one -- (b)Q5 or (c)S5 -- is better than R6.

Life In 19x19

A Game Against Chatterbot

A Game Against Chatterbot

Re: A Game Against Chatterbot

Re: A Game Against Chatterbot

Re: A Game Against Chatterbot

Re: A Game Against Chatterbot

Re: