The problem of newcomers to the site being matched with SDKs.
If the newcomer is in fact an SDK, their rank quickly gets there as they beat a few DDKs.
OTOH, if the newcomer is a DDK, there is less chance that they are paired with SDKs to the detriment of both.
That’s what was intended to happen.
Seems like a reason to just make the initial rank lower 
or let new users pick!
To recap: the initial rank is set at the midpoint for Glicko-maths-rank-pool-balance reasons.
Humble-rank was intended to make it so that the effective-rank-while-the-glicko-settles is more what we’d like to experience.
(“was intended to” because there was confusion about whether it was working in all matchmaking initially, confusion about whether that was fixed, and further impact of the rank-recalibration that happened afterwards)
https://www.reddit.com/r/baduk/comments/176lebi/how_can_you_change_your_rank_on_ogs/
Oh the irony that a non-naughty 6d player is forced to sandbag as a 12k, because allowing new accounts to choose their rank would allow naughty people weaker than 12k to sandbag.
I think that the substantive argument is that allowing people to choose their rank would distort the rank pool.
If you could refute that, then the “pick your rank” argument would have more legs.
Already happened in prior threads.
Links?
The forum search didn’t help me find it, but I recall @Groin and @shinuito referencing it.
Glicko doesn’t require new accounts to be 1500 / middle, that’s just the recommendation in the absence of any information about their strength. If you can get that information then use it. Even if it’s not perfect, I’m pretty sure things like the sum of squares difference to true rank will be less than starting everyone at 12k. As new accounts have high rating deviation they have less impact on existing players’ ratings that the changes in their own. In fact if their starting rank is closer to their true rank then they will take/give fewer rating points out of the player population to reach it, so an initial rank could lead to less distortion of the rank pool. Someone did some simulations that showed non-12k starting rank was fine.
But also the periodic reminder, my main driver for wanting this change is to make OGS a better experience for new players, particularly beginners who currently are subjected to a bizarre hazing ritual of getting beaten up by strong players until they can play other beginners. OGS’s purpose is surely to be a nice place to play Go (and encourage new players to join the community), not be a slave to the rank system (not that I accept it is even worse for the rank system) and sandbagger-phobia. The ranking system exists to serve the players, not the players existing to serve the rank system.
We have same memory. The original thread seems to having been merged, and still some parts, like some maths. simulations missing (at least from searching).
In my understanding of previous threads, that’s not exactly this but that it could lead to some global shift which would involve some periodic reajustement.
I remember for example someone saying that because of COVID times, we may expect many people trying a few games and leaving which would affect the global rating if we introduced an entry point for beginners.
That’s not the counter-argument that we need.
The argument that’s needed is to show that allowing newcomers to choose their rank will not skew the rank pool. IIUC.
The whole business about sandbaggers is a distraction - a second order argument, AIUI. It’s like "we have this decision (to start people at the middle rank) and we have this problem (sandbagging) and we look for solutions at the intersection.
OGS’s purpose is surely to be a nice place to play go. One of the considerations is having a stable rank pool. Once again, it’s only “as I understand it”, but this is my recollection of the primary reason for “the way things are”.
I don’t think that’s as true as we want it to be.
If some random person comes along and thinks “I’m 7d on fox I’ll pick 7d” and we give them max deviation of 350, and they lose to an established 3d
the 3d gains like 12 rating points, but that’s actually comparable to beating an established player at their own level
This probably will happen regularly even if we could provide an accurate rank comparison chart, because people probably won’t go to that effort to play games. It doesn’t affect them anyway, but it probably will cause some instability at the entry points.
@Vsotvep was playing around with glicko calculations and I believe they suggested it might not matter if we let people set ranks
I’d probably rather see it done with the actual OGS games database, which should be doable on python from here
At least then you might get some overall pictures of how it affects certain entry ranks or overall rating drifts and so on specifically for OGS with all the strange things real people do with their ranks and games.
Various quotes about letting users choose their own rank.
It’s not just an OGS thing that people can’t reliably choose their own ranks, I played a player at a recent tournament where they said they were strong like first rank (it wasn’t said to me directly so I don’t exactly know) in China or something so they were supposed to give me four stone handicap as a two kyu - I won by a good bit, then we played even and I won by a lot, and then I could estimate there were more like 8-10kyu than X dan, which we could test by having them play a 11 kyu where they won.
Only that we discounted the first round pairing (so it’s like I sat out), and played some additional games to figure out their rank did we give them a better chance at having some more even games in the tournament.
I see this as somewhat analogous to what glicko is doing anyway, someone isn’t sure how strong they are, they play some games lose some games and they quickly within like three games (if something silly doesn’t happen) start to pair with people their own strength approximately as in with this real life case.
For total beginners, who know they are total beginners, there is a point about having them play a 12kyu to whichever, which I think can be solved by fixing the pairing algorithm, rather than messing with the ratings entry points
If this could be… First the system won’t allow them to play an established beginner because of a too high difference of strength. Then Let’s imagine 2 beginners playing at a 12k similar entrance point, what will become of each rating, after the game? One getting to be 8k, while the other down? All seems weird.
I guess we can come full circle back to the original point of the thread here.
If there’s people that do this, and it’s fairly easy to make a new account, or let’s it’s not even malicious it’s just that people who think they’re beginners are actually stronger than who we would pair them against, is that a good experience for established beginners? For them to get stomped by people unintentionally or intentionally choosing to be paired with beginners.
Is it a better experience for the established beginner to have the new player play one or two games, confirm that yes they are actually a beginner, now play them and you should have an interesting game?
Who should the system prioritise? Is there a compromise that works for both new beginners and established beginners?
I mean I also find this a bit strange for this reason but the system apparently does try to pair new players [?] with each other, before just picking people at their level.
At some point though you have to just play established players in the system if you want to know how you fit into the system.
I’m warming up to the rank-choosing, even though I previously said that the idea needs answers to certain questions.
You assert that there is no substantial risk of rank pool distortion, let’s just believe it for the sake of argument.
I said that any features based on annoying popups and questionnaires has usability concerns. But there are other ways, i.e. if the choice was optional without prompt. The new user would have to deliberately get out of their way to set the starting rank – no disturbances.
I said that the starting rank determination would have to be accurate, to make sure clueless or malicious newcomers don’t start at 6d. But actually it would be sufficient to improve on the status quo in terms of accuracy.
I don’t think basically that answering sandbagging problems with a sandbagging created by the rating system itself is the right thing to do.
I don’t understand what this comment means
To fight sandbagging, the rating system ask you to play unfair crushing games for a while, with the similar effect as being sandbagged.
What i would call a failure in solving the sandbagger problem.
I’m glad you think sandbaggers are a second order argument, I agree. But my impression was the first order rank pool skewing issue was already demonstrated as not a problem (thanks @shinuito) so now sandbaggers was the primary objection. But maybe that’s just because @Conrad_Melville is very vocal about it (though I’m still not clear if he thinks dan players making a new account because they forgot their password are alt sandbaggers, in which case the inability to choose a starting rank is the cause of their forced sandbagging as I expect most of them wish they could say they were dans, not 12ks).
