With the nerfed one I’m not as sure. Even when katago is giving fractional territory to some places, one could look and think “I could live there, or take away those points” but I imagine one does so at a cost and it might be detrimental elsewhere.
I have read some suggestions in the other threads for this too and it could be quite useful. A simple toggle for live/dead stones and marking intersections, and maybe a faster way like a rectangle tool for marking whole areas.
One can still object to this for live games, but in correspondence at least it’s just about it not taking as much time to do it, even though you do have the time to do it.
This. I used to use Fox score estimator all the time (which is quite a good score estimator), until I realized that it was making me objectively better at late middlegame and endgame - but only in games where I was actively using it. My own intuition or ability wasn’t improving whatsoever, and I was just stunting my own growth.
Now, I pretty much only use it when I’m thinking of resigning and just want extra confirmation that I should resign, or when I’m ridiculously far ahead and want to see the absurd score difference that my opponent continues to play against. Maybe it’s not the best practice, because I still do regularly have moments where I’m surprised by the number and continue to play when I otherwise might resign, but regardless, without the visual representation of strength and big endgame points, I highly doubt it would unnaturally augment anybody’s yose or other aspects of play.
@Gia Even just the numerical information, like W+50.5 is already a lot. Especially because if this number comes from KataGo you can assume it is absolutely correct.
I don’t like the idea, I think I want to play without AI during the game. AI is great for analyzing the game after it is played, but I don’t want to use it during the game.
I mostly agree with you. Giving the score estimator a brain like katago’s gives me a bad feeling. Like you say, it may be misused to check for L&D situations, invasion points etc. It should not be as smart as that during any game, rated or unrated, nerfed or not nerfed.
I am also lazy and bad at counting so I often use the estimator, but really I think it would benefit everybody much more if there is no estimator available by default, at least in rated games, and only something like the dumb one during rated games if both players agree to use it.
After the game of course it is great to get the real katago score.
Maybe dumb idea, but anyway: how about a third game category: unrated, rated, serious. Serious is rated but with more restrictions: no analysis, no estimator, … anything else?
If I were allowed to ask a question of help from a much stronger player or pro in a game, how am I doing would be the most frequent one. Then I can determine the game tactics accordingly.
As I watch lots of pro games now, it is quite often the pro commentators don’t really know who is winning or losing. It’s that difficult.
I guess it’s not possible to change my vote, is it? This has never happened in the history of the internet but the comments section actually made me reevaluate my opinion I had previously felt like score estimator being allowed when both parties agree to its use would be fine, but I suppose this advantage could cause a small rank discrepancy between players of similar strength when one primarily plays games with SE and one primarily plays without?
Actually, now I’m curious. How much of an advantage would using SE give a player? Two stones? One stone? Less than that?
Anyway. I guess right now my opinion ultimately hinges on that question, and I’m not sure whether anybody has anything close to a concrete answer. If the advantage is negligible enough (less than a stone, I guess?) then I can’t imagine that it will disrupt the rating system in a meaningful way. Serious tournament games being hosted on OGS can require that SE be disabled. In the event that the players make a mistake and enable SE for a serious game, I imagine the tournament directors can decide - if no involved parties want to dispute the game’s outcome, then it probably wouldn’t matter, or if there is a dispute, the game can be nullified and a new one started. Regardless, I don’t think that scenario should be OGS’s problem.
If the advantage is significant, then I suppose I’d lean more towards “I’d prefer if it were never allowed for rated games, but also my preference is not a strong one because I don’t feel that I’m a strong enough player for it to matter one way or another, in my games at least, and I mostly just play 9x9 on this server anyway so uh does my opinion even matter here”
I use it basically to know if I should resign and not torture my opponent any longer (of course I don’t check all the time, just at the end)*, and on the rare occasions I was winning by a lot and I was wondering if I’m seeing things or my opponent doesn’t want to resign (their right btw, if they want to play to the end, we play to the end I’m saying this because it was an intense discussion about it elsewhere )
*I do check during a couple of teaching corr games I have going, but to discuss the moves/ think about the moves, not to trick my opponents. It’s teaching games.
Personally, I’d rather be considered weak (which is corroborated by the facts of my rank, my rating and my games themselves) than a weakling, I don’t know how cheaters do it, my pride would never :-P.
A comment I learned from pro about AI is if you don’t know you are winning, you are not even if AI says you are winning big. Same thing goes if you are not sure if you are losing, you are not losing and should not resign. This is especially true when we play even games with opponents on the same rank.
Resigning is a respect to our opponents’ ability to hold the lead to the end. So if I don’t know how much I am losing, I can safely assume my opponent does not either, thus he/she might exercise unnecessary risky tactics etc. No need to be too courtesy. Though I am all for resigning when I don’t see any chance to win back the game.
I don’t disagree, but I have a question: newbies/ beginners are bound to make lots of mistakes in counting, but they also need ranked games if they are serious enough and want to advance. So, all those newbie-miscounted games, would they lead to much cancellation/ annulment? Is there a better or quicker way other than having automatic counting (at the end of the game, we agree on that, not during)?
This isn’t fully clear - does this mean that even if the games I create have this L&D-solver SE disabled, our opponent just needs to open a new window to have access to the nerfed version?
Or does the IP limitation enforce the ‘disabled’ setting as well?
Just opening a new window does not give them a new IP. It should be enough of a deterrent that an opponent would have to open a new window with a VPN or proxy connection to get access to the score estimator. However, at this point the opponent might as well be running their own AI offline.
Both will be considered a ban-able offense as well, is what I understand.
I think the question is not about how IPs work, but rather whether a spectator on the same IP as the player with the disabled settings would have access to the nerfed SE, in spite of SE being disabled for the players
Maybe spectators should have to be logged in to use anything AI-related in ongoing games? I guess it’s easier to track if accounts are suspiciously the same person, rather than guests/ spectators.
Not a fan of the current nerf. There should be a threshold for fractional nerfs below which it will not show up for either side. Otherwise, influence would be totally OP in the eyes of the nerftimator.
For live games, I think the default should be no analysis and no estimator. This is, I believe, the default on KGS. It seeks to emulate the playing situation of a real-life game as other members have explained. In a real game, you have nothing but yourself and the stones.
For correspondence games, I believe analysis should be enabled and score estimator should be available in its newly balanced form. This is again an appeal to tradition – in the past people played correspondence games with all tools available to them, with the effort to choose the ideal move instead of worrying about time pressure. Of course, this excludes other people (though I’m sure this has ben broken often), as well as AI.
I have changed my mind since the previous score estimator poll thread: AI is too overpowered. To use the score estimator in its full form should be an option available to no one, certainly not in ranked games. I think choosing to use the full SE vs the nerfed one may be allowed for unranked games, but the default should be nerfed. The nerfed estimator should be OK for both ranked and unranked games, though; I don’t really care about the default in this case.
tbh i kinda liked the old SE, like yeah sure it was the absolute worst score estimator of any go server, but it was unique to ogs ^^
For most cases, its enough to have an estimator which tells that the game is still close. I didnt mind SE giving me “black by 10” “white by 12” when i clicked it twice on in a row, it was good enough to tell me that the game is still close enough.
It doesnt need to be super accurate, its an estimation; just having it about the right ballpark was/is good enough.
I wanted to add that katago estimator should recognize japanese or chinese counting. it makes a difference in evaluation. in gtp its kata-set-rule scoring area/territorial
I had previously felt like score estimator being allowed when both parties agree to its use would be fine, but I suppose this advantage could cause a small rank discrepancy between players of similar strength when one primarily plays games with SE and one primarily plays without?
I’m saying this because it was an intense discussion about it elsewhere 
)