Thanks, @anoek
Hm, is it me, or ranks are slower to update than usual? Typically right after the game you can go to your profile, refresh the page and itâll be a new rank. Now it can still show you old rank, and update a little later.
If you have kept playing since then, wouldnât it make sense that you have improved?
Well, that was a somewhat anticlimactic way to retroactively hit 1d.
Now to get back there.
I think there is âsomething going on thereâ - weâre looking at it.
So on the one hand, Iâm sad that I got âpromotedâ to SDK âfor nothingâ.
But on the other hand, I think it makes so much more sense that the actually existing ranks represent a quasi-linear spread of skill of OGSâ playerbase from 25k for complete beginners (compared to âinvisibleâ 40k with the previous formula) to strong dans on the other end (compared to low-dans before).
(Of course, I would still prefer that 9d+ was represented as 10d etc!)
Thanks for doing so much hard work to tinker with the system. I like being 2 dan, btw, keep that up for sure!!
Itâs amazing how much one little bug can cause in terms of headaches and âtangential productivityâ.
Again, thanks a lot, Iâm sure the new ranking system tweaks will be great.
tbh I still donât understand why ranks under 25k was âhiddenâ. When I started learning go, the books tell me 30k is âjust read the rulebookâ (several books, not one), so I kind of thought itâs a globally-agreed concept.
Is there a good reason for the 25k limit? Could someone elaborate?
Iâm reminded of that scene from Searching for Bobby Fischer in which young Josh Waitzkin wants a certificate of achievement from his coach, but his coach tells him they donât matter and [redacted spoiler].
Maybe we all want a certificate because our inner 9-year-old wants approval from our parents, our inner adolescent wants the approval of society, or we just want the approval of ourselves.
I play almost exclusively correspondence games, so I donât play that many.
If I did improve (which I donât think I did), it would have been from 15k to 14k (13k max). Improving to 9k is unrealistic, specially because this rank jumped overnight without any new games, and so did the rank of my recent opponents (with whom I was pretty closely matched).
So I think this is due to rank scaling, rather than me getting better.
so, strictly speaking, the rank floor is arbitrary. 30k just happens to be one of the more well-known ones. The issue is that any floor depends ultimately on the ceiling, which for amateurs is set at 9d, or 7d (in accord with the nihon kiin), and that in theory the difference in rank works alongside handicap (this is why we get so many âif you want ranks to be stable make ppl do handicap moreâ comments). The problem is if thatâs true, then if the ceiling rank of 9d (or 7d) gets stronger, then all the ranks under it must get stronger, leaving beginners at an even lower ranking. This would leave the floor effectively undefined (but ofc most ranking systems donât do this).
The reasoning in particular for 25k being our floor (rather than 30k like is well known, 20k for EGFâaltho they are talking about changing to 30k, AGA doesnât seem to have a rating floor but the lowest on record is 49k, interactive way to go believed in a 50k floor, and Iâm sure other associations have different conclusions) is that upon researching the data, anoek concluded that stone handicap did not make a meaningful difference between ranks below 25k (although I do not know the methodology by which he reached that conclusion).
Seems like a falsy statement to me. Skill-based rating isnât based on âceilingâ, because we donât know the âskill capâ. We either have an arbitrary 0 point or set 0 as âknows nothingâ.
So basically the ceiling ultimately depends on the floor.
My take on the reason is that below a certain point every game pretty much has a 50/50 outcome probability. In that below a certain rank games are as likely to be decided by chance blunders than the application of skill. Therefore, all players below this point can be matched to each other and have even games. I guess the view was that that point is/was 25k.
Youâre discounting the idea that some players might be so bad at Go that their strategy gives worse result than random play.
With current rank conversion, yes. With OGSâ previous rank conversion, the point was around 40k (but displayed as 25k).
I feel like that about my games. Sometimes I throw the game with a mistake after working hard to keep the game even. I donât think you have to below a certain level for that to happen.
Anyone can make game-deciding blunders, thatâs not the same as the total confusion and pseudorandom moves made by some beginners; the kind of players that would get crushed by a true 25k
But that is against evenly matched opponents where you should have a 50/50 chance. If you played a 25k youâd not have this problem (unless you really did work hard to keep the game even!)
Is it strictly necessary to have a ceiling or a floor? Arenât they just cut-off points based on the distribution of ratings, beyond which points there are very few observations? Probably some arbitrary point needs to be defined as zero to anchor the system, but it isnât necessarily the ceiling or the floor.