It still takes a bit of work to swap between tabs and interpret what the winrate or score changes mean assuming it’s the nerfed one. Figure out how you can use that info.
I highly doubt it’s difficult to just play your opponents move and your move into another tab with AiSensei/zbaduk open and then you barely even need to think, you just have the bot telling you everything, can even test if the move you want is going to lose you the game before you go for it.
Anyway, maybe you both are right that there should be some tournament mode that disabled everything for everyone, spectators, players etc.
of course we know that there is a universe of people who cheat.
the problem is that there is a larger universe of people with the ones that wouldn’t go as far as blatantly cheating with a katrain/ai-sensei, but would use the L&D tools provided by OGS itself to get an advantage.
and don’t forget that the only thing being requested is that the game owner can choose if AI-assisted SE is available or not. Why shouldn’t he have the option?
/edit: I replied before seeing your last comment. I think we agree on the essential then. 
Clarification: Tournament games are always ranked games.
I think they mean official tournaments, previously IRL games.
Compulsive checking of SE is a terrible disease anyhow. Let’s get rid of it, yay.
Fast and accurate score estimation is a critical skill and component of go strength, and one which most amateurs are really bad at.
The current score is important in part because if you’re winning you should play safe and simple and if you’re losing you have to be more aggressive or make the situation more complex.
Providing any automatic score estimation in live games seems no different in principle than providing direct assistance with move selection.
Maybe you could offer it as a mutual opt-in feature for unranked games, but that’s the most I can imagine reasonable.
Indeed, words mean things…but sometimes they don’t.
Maybe the previous SE engine was not good, but this improvement is in the wrong direction
we need to give user more control during the estimating, not a smarter Engine that does it 10 times better than human
This has been discussed bit already, but I’ll clarify it again - opening the game in a private window, like Private Browsing on Firefox or Incognito Mode on Chrome, will not give a player access to the un-nerfed SE, as private browsing does not change your IP address. Private browsing is not as “private” as you might think; all it really does is not save cookies, browsing history, and other data to your local machine.
Setting up a VPN to gain access to SE in the way you describe would require quite a bit more effort than plugging the game into an AI on your computer like Lizzie or KataGo, or in the cloud like AI-Sensei or ZBaduk. As @shinuito mentioned, some of these services may try and cooperate with OGS to prevent abuse, but here’s the thing - if somebody has the technical know-how to abuse the new SE during a game where they’re not supposed to be using it, they would probably cheat in much more effective ways instead.
I don’t disagree with you here, but I think most, if not all, of those people, would be deterred by the effort that would go into this type of cheating. If they’re not, I doubt they would be the types of people to balk at the idea of blatantly using AI whenever they see fit. Cheating is cheating, and using clever workarounds to get access to SE is likely less severe than opening the game in a private window with a different account to view kibitzer’s comments, which is something you can actually do (or at least I think you can) on OGS right now.
I’m concerned if the existence of SE could provide more opportunities for cheaters but that’s not really what is worrying me.
I’m much more concerned by some tolerance from the server itself vs using information outside the players own efforts of their brain. That would be a (r)evolution in OGS policy (about the use of AI) which would officialize that this is no more such a problem.
In my opinion, for unranked games there should be the option to choose between the two score estimators!
I know that only the ‘nerfed’ SE would be available, but that one isn’t less of an assistance.
‘i’m thinking of tenuki, but have I missed something in attacking this group?’
check
‘oh great the group is marked as dead’
Removing fractional dots doesn’t really offer much less of a L&D help.
So I’ve not followed all the discussion but is there broad agreement that SE now needs to be unavailable for ranked games anyway?
If so, has there been a suggestion of a “counting mode” as an equivalent to “analysis mode”. If analysis mode is acceptable because it’s the player’s brain being used to mark sequences on the board then the aim should be to have a similarly user dependent “counting mode”
The way of envisage it would be that in counting mode, the user makes intersections as black or white and a count of black and white is kept as a running total. After a bit more play, you could go back to counting mode and your previous count is still there but you can update it.
The benefit is that the adding up and remembering the count for each player is done for you but whether or not a point is black, white or undecided is entirely up to the player.
I am unable to count but only because I can’t keep track of various numbers in my head so this system would be just like having a pen and paper to keep track of the numbers.
There are various UI options of course. I imagine some kind of marquee dragging to select rectangles in one go and buttons to choose black and white marking. And clicks on stones mark as dead but stones in a marquee would be ignored. Stuff like that.
I think OGS score estimator definitely needed some help. All score estimators will be memed about but OGS was pretty special. Making it use KataGo is a risky direction though and the possibility of playing out variations then asking for Kata’s opinion is very worrisome.
However there aren’t that many details on implementation here aside from it being “significantly stronger”. A few of the life and death examples are worrisome but how cherrypicked are they? Even the old score estimator would be “helpful” if in 1/10,000 situations if you interpret it generously.
Personally I want a certain amount of human input. Maybe do a liberty check and anything with more than three liberties is always alive and must be clicked to mark as dead(and vice versa) or some other simple solution to the life and death concerns and then we can move to further observation about how helpful of a score estimating tool is too helpful.
Edit:
I also get that this is meant to be some kind of referee to auto-score lower level games. Fox has a button for something like that and I think it can still be used in that case without the liberty checks - just give the players warning that auto-scoring will be final and end the game? Then you can use max power Katago to get your accurate result.
I think, apart from the remembering the count between closing the game window, all of this should actually be doable front-end, so someone from the community might be able to implement such a counting tool.
That would almost certainly be outside assistance and not something agreed upon in game setting. I guess they could make it and see if it gets merged 
Could it be linked to the “disable analysis” setting so as to be disabled if analysis can’t be used?
I see it as very much in the same vein so would imagine that they sound be usable/not usable together.
How would it be more outside assistance than the SE currently is or previously was?
You’d be the one deciding what is territory, the only thing the system does is keep the tally for you and mark the intersections. Seems a lot less controversial than a score estimator.
I mean really, the minimal implementation would be just to add white numbers to the analysis tool