The new SE still doesnt seem to work on boardsizes larger than 19x19, regardless if its ongoing/finished game, or if its my game or i’m spectating >____>
It just says this
Indeed, and it wont, I’ll add some code to detect that and fallback to the old SE for those games. (That’s a limitation of KataGo)
9 Likes
How about determining bot game outcomes (not the topic if this discussion, but rather “the next thing after it”)?
Scoring bot games with the new system is releasing tomorrow 
I wanted to make sure things were working ok and seemingly mostly sane before forcing those scores, but it looks like it should do pretty well.
5 Likes
SE still estimates wrongly when there’s unequal number of captures.
3 Likes
Been afk much these days, and find it hard to catch up, and not sure I read all of this thread, but …
This I found quite disturbing … if we’d follow this vote then we’d take away something that we now have (admittedly the 47% is current state and thus may change).
IMO the SE should definitely be allowed for correspondence games.
…
[a few minutes later]
Read a bit more … so … do I understand correctly that the SE settings we have now will stay available?
I’m bad at counting any further than I have digits on hands and feet, and I’d want the “dumb” SE to stay available in correspondence games … do I understand correctly that this will be the case?
5 Likes
I think so
1 Like
thx 
1 Like
Yep, basically it’s no change for the in game experience. Once the game is over, or as a spectator, the score estimator is significantly improved.
6 Likes
something still works wrongly on 13x13 for example
https://forums.online-go.com/t/new-score-estimator-give-incorrect-estimation/36883
Is it off by 10 points W+2.5 to B+7.5 because black has 10 more captures than white? (B 13, W 10 captures?) Specifically captures taken off the board.
1 Like
It’s the thing where se adds Black’s captures to the score in Chinese rules.
1 Like
Should be fixed now
6 Likes
Players in ranked games should have no access to any score estimator. In ranked games that are also tournament games, spectators should not have access to estimators either. In a case where an estimator is available, it might as well be the best one. What is the point of anybody using a crappy estimator?
1 Like
Maybe an idea for the folks that don’t like to have a counting tool at all: a manual counting tool. Something simple: mark areas on the board and just add up the combined sizes of all marked areas. That could also be useful in teaching games, or to estimate moyo sizes.
Edit: on second thought, if analysis means no computer help at all, even this counting tool should be disabled.
4 Likes
I guess even using a computer for transmitting moves should be forbidden.
Joking aside: In correspondence you basically have all the time you need. So more or less everyone can count, because counting quickly is hard, counting isn’t. So all that any purism forbidding any kind of tool does (as far as correspondence is concerned) is forcing people to do something boring.
I’d be entirely happy with a manual tool like you outlined. I’d probably leave a site that decides forcing me to do something boring is a good idea though.
7 Likes
This is an ambiguous poll. The question lumps together the players and the spectators. I would like spectators to have AI analysis results if they wish them, but never players themselves. I think the results are invalid if anyone got confused as I did.
2 Likes
Anoek, If you are revising the score estimator, I will put in this related request-- Instead of reporting only the blunt “white ahead by 7.5 points” I’d like the display to report an explicit breakdown of the elements of the total scores, e.g,
w territory + captures + komi = w total
b territory + captures = b total
That would make me very very happy!!
thank you
I intended that poll question to mean “Should the score estimator be available [to players] for ladder and automate site-wide tournament games?”
I believe most people would have interpreted question that way. However, even if some people interpreted the question more broadly as about spectators as well, the results still indicate a meaningful preference, since an overwhelming majority indicated a desire to keep SE away from ladder/tournament games (whether or not they interpreted as applying to just players or both players and spectators). However, one can only conclude that the vast majority do not want players to have access to the SE in ladders/tournaments, while it remains inconclusive as to the opinion with respect to spectators.
Another implied condition is that this question only applies to the situation while the game is still ongoing. However, I think it is clear to everyone that score estimation and AI analysis after a game is not controversial.
3 Likes
My friend and I were looking back at our game tonight:
At move 194, the KataGo score estimation (the bar, just below the clocks) says W+14.1, but if you select “Estimate score” from the rh menu it says W+0.1 (+0.2, +0.3 if you press again) and shows groups as dead which KataGo actually knows are not dead (B corner top left, W dragon at bottom). That’s to say, it seems to be the old score estimator still being used when analysing the game when it’s over.
Can you clarify, please, as you seemed to be saying that should not be the case.