Life In 19x19

http://eidogo.com/#jV9DyZoZ

5 dan setting.

At the position where it resigned, its own score estimator says it is ahead by 3 points. I guess just turn off the resign feature?

Turn off the estimator instead.

I loaded this position into Crazy Stone and it said that Black's win percentage is 93% with a score of +27.

I think the score estimator (if you mean View / View Options / Situation Status / Text) just bases its output on a very small number of playouts.

When it resigned, I have just a "Territory" button on the Android app. How could it be wrong by 20+ points though?

The game isn't over, so territory estimation is just estimation. Black has a bunch of interesting threats, like P13 and M6. Also Black is going to pick up some points around L19. I dunno if it all adds up to 20 but there is plenty of endgame left to be played and it looks like Black will have most of the fun.

Putting undue trust in crappy score estimaters seems a common beginner trait these days (particularly on OGS). Perhaps one advantage of the KGS one being so bad is it warns people off it, as does the many jokes at its expense in the kibitz. Perhaps they should be renamed "Score Randomizer" instead?

By the way, if you thought that White was ahead in the final position, it probably would be a good idea to turn off the resign feature, but for a different reason: Crazy Stone might be giving you too much credit in its playouts, and you would benefit from playing out the game to the conclusion and being forced to find good endgame moves.

Ophitoxaemia wrote:When it resigned, I have just a "Territory" button on the Android app. How could it be wrong by 20+ points though?

The score estimator does not use the Crazy Stone AI, it is an entirely independent bit of software, and is nowhere near as strong as Crazy Stone itself.

For a real life comparison, this is like a dan player resigning even though the 15 kyu spectator thinks he is ahead