So on stream I have the SE bar visible for viewers it I think the estimation is like 5kyu or something. Doesn’t really matter, but I think the SE is misleading as it doesn’t really reflect the game at all for amateurs.
I think having 3 bars would be better. One for actual territory (points+stones), one for expected territory, and one for AI’s estimation on who is winning. The tricky part would be making the visually obvious without a bunch of graphic noise.
This is probably an optimistic feature but I would like the DDK and beginners who are watching me play have an idea of what is happening, The SE is very misleading if there is a L&D problem as well.
Do you meant he score estimator which is provided for the 2 players while they are playing the game? That thing is really bad, but its bad on purpose… Too good SE for the players playing the game could be considered as AI assistance, since they could just put down 1-stone variation (their next potential move) and click SE to find which move is the best ://
If I and my opponent are reading a L&D but the SE says it is or isn’t dead, it doesn’t matter unless the players can solve it. On the flip side, the players know a group is dead but the SE doesn’t so it shows one player losing by 50+ for half of the middle game.
Well the policy is to not have external help for any games (including teaching) globally on OGS. Some still liked this old SE even full of errors so it was kept.
There is one exception, the alan turing tournament, so i guess you could negociate but as a whole i don’t expect OGS to change its policy.
One other way to avoid problems with a score estimator during the game is also just not to use it.
It’s “throttled” because if you don’t throttle it, then you’re just cheating with Katago, even if it’s in a more subtle way than just asking it for the best move. In this case though it’s just an old non Katago estimator, so it’s not really throttled even.
I don’t know exactly why we have score estimators during a game, but it’s not a unique thing, fox has one, KGS I think too. I don’t think pandanet does unless you’re spectating.
Oh i think, with adequate caution, it’s an interesting teaching tool sometimes like for weak beginners to show what is a territory, i mean not precisely but just an idea of sharing the cake.
Because of its failures, it still needs you to put a good check of coherence so it can be considered as really a score estimator but not a strategic helper. Some are just SO lazy to count…
But for me yes that’s a no way to use it myself
You know, even as spectators, the use of AI changed a lot the Kibbitzing. You can’t predict who will win as we did with funny kibbitzing. You can’t even argue on some moves all this because AI is there behind us. So all turned a bit “short” and less lively. You can still do it on pandanet but that’s not the place for joking
Instead of crazy Go estimator we need tool that has no Go knowledge, but answers clearly. We just need ability to draw line with a mouse very fast like in Paint. And then tool will instantly answer which % of total area painted shape has.
Yeah, this has been proposed before and I would support it, as it automates a routine task, while providing no outside help otherwise. I would still prefer to play analysis disabled and not have it, though
