I would start new accounts with even greater RD. As in double or triple.
3 Likes
Just (re) read the section that Iâm highlighting
It took me ~2 crushing looking games to reasonably accurately figure out someoneâs level - I mean theyâre crushing because I want to see how they answer different moves, how do they fight ko, how do they spot double Ataris or shortage of liberties, how do they live with a group when theyâre surrounded. The point wasnât really for them to win it was to get accurately ranked. I could guess they were above a certain level and they won when they played a person at that level, success.
Glicko can do it in like 3 or so games in the best case, probably more on average I guess, and assuming someone doesnât resign when theyâre winning.
2 Likes
Thatâs only true if @anoek agrees.
1 Like
Unfortunately that seems to come with some other problems
âOh you lost to a 3d, you must be like 13kyuâ.
Iâm assuming this is what would happen above. I think ogs caps the rating deviation at 500 apparently
the front end rating calculator is just the python code above converted to javascript so people can interact with it
I canât remember all the details - I have to look these things up as I go along.
2 Likes
If glicko is so quick to rectify then why oblige a beginner to go through a crushing experience instead of giving him some chance by registering at its declared level? I mean why being so obsessed by sandbagging when glicko is offering a quick adjustment? Shoudnât we be prioritizing fun and welcoming ?
4 Likes
I just donât understand what youâre saying.
Nobody has to play ranked games. Play unranked, have 20 stone handicaps, give yourself 400 komi, do whatever you want to have fun.
If you want to figure out where you fit in in the system you have to lose games - if you never lose it doesnât know where to put you, maybe youâre 9d+
1 Like
Itâs part of the fun to be 22k instead of 25k.
Itâs fun to play rated fair games.
Itâs not fun to be crushed by a 12k.
3 Likes
I still donât understand what youâre suggesting.
If you want to be 22kyu, youâll have to lose to someone strong than 22kyu, and beat someone weaker or around that.
If you want to be ranked in the system you have to be willing to lose games.
I was told recently that this is the kind of handicap they start off with in Korea and China when starting out in Go, and youâre like 30kyu when you can beat your teacher with these 25 stones.
Itâs probably also not fun to be crushed by a 12kyu pretending to be a beginner when youâre 25kyu and have to constantly play new accounts thinking or pretending to be beginners.
We can just keep saying the same things over and over again, or we can try to devise a system?
1 Like
@Groin we can continue here if thatâs ok.
You can copy and paste your response over or post it in the other thread and I can move it here anyway, all the same.
1 Like
Most other systems donât let you register at some upper point (like 1k for example) which i fully understand. OGS is even more straight by automatic registration at 6k or 12k, i dunno in fact exactly which one. Seems not much worry as i guess those stronger players may accept a bunch of easy games for getting their rank updated.
Now my concern is more about those people starting to play go. Most servers donât put a constraint to them to enter the system as a 12 (or 6?)k and even if glicko will reajust it in like 7 games (with some good luck or manual reajustement by a mod) and OGS is by far the most unfair placr to go for them.
I really hope this could change by offering them an entry point at their level (25k?)
This entry could be discrete, not advertised too much by simply asking âare you a beginner ?â when registering a new account. No mention of the rank that will be attributed. Sandbagging may be even easier to detect and those beginners wonât have the hassle to go through that crushing experience they have yet.
4 Likes
I understand. Thatâs good for new players to the site that are 25kyu or below.
But itâs not necessarily good for actual 25kyu players currently on the site.
In a perfect world they all get even matchups, but in reality, someone thinks theyâre 25kyu but theyâre much stronger, or someone declares theyâre 25kyu (sandbagging) but theyâre not, or someone just wants to start at the bottom and rank up, or⌠and so on.
The actual 25kyus suffer, maybe they even consistently lose rating to new players and they donât get to experience
I donât think we necessarily need to change the entry point, but Iâll quote it again and again
You could make a separate unranked game pool for new players to automatch in for the first N games.
You have people challenge any level in unranked, maybe just give them a quicker way to find opponents.
I think we could come up with any number of matchmaking changes and then we donât have to worry about how it messes up the ranking system at all.
2 Likes
I have mitigate feeling on separated system but thatâs still interesting idea.
2 Likes
Allowing people to choose their own rank could distort the rank pool. That is a risk that the fixed starting rating avoids. Beginners claiming to be dans and dans claiming to be beginners could result in distortions and it would be worse if there were more of one type than the other.
A fixed starting rating is predictable and that brings stability. The total number of rating points in the system will be less likely to drift up or down over time out of proportion to the number of players.
It doesnât take many games for a new player to get a rating representative of their level and beginners can normally expect to lose most of their early games - I certainly did. People can also learn things by playing stronger players. I donât really see the problem with the existing assumption.
Impossible, as i propose only beginners to claim to be beginners.
From 7 if lucky to 15 around, thatâs not so few for someone discovering a game.
Thatâs not true if you are matched with beginners of similar level
Not everyone wants to learn and not everyone will learn from being crushed. Even some may just get discouraged and give up the game.
There are beginners coming here or in the main chat, asking politely for playing fun fair games and thatâs not what OGS is providing. Just see the facts.
Every time i introduce go to someone, my first wish is that he can quickly play with other beginners instead of me. I wonât recommend OGS for that very simple reason.
4 Likes
Not a proof, but I think a good argument: Chess.com shows this before your first game and they use Glicko as well:
I think this means one of the following is true:
- Initial rank doesnât matter much in Glicko
- Chess.com has found another way to anchor the pool (likely)
- They donât consider drift an issue (unlikely)
8 Likes
I see that you are still disseminating your misunderstanding of my position on this issue, which I thought I had laid to rest long ago. Now I must recap.
Some years ago, a player I will call Devourer (I donât remember his full name) spearheaded a discussion of this issue in another thread. At the time, my attitude was either undecided or go-along-with-the-status-quo (I donât remember the exact nuance). Since the discussion was rather technical, I didnât contribute much to it. I did point out that the original justification for ending self-designation of starting ranks was that it caused too much work the mods due to constant requests for rank adjustment. It seems like that would still be a major concern with any self-designation procedure.
At the time, I also pointed out that the entry point had an advantageous ancillary effect of limiting the range of sandbagging. That is a simple fact. You mistakenly thought I was arguing for the entry point because it limited sandbaggersâ scope. I never believed the limiting effect on sandbaggers was a decisive factor or even a major reason, and I said so explicitly in my response to you.
Subsequently, Devourer seemed to refute the idea that the mid-point entry was necessary to Glicko 2. Also, the defects of the entry point have become increasingly clear since then. Consequently, I am now inclined to think that the entry point is a bad idea. However, I do not have the technical knowledge to discuss the matter further, except to say that a reversion to self-designation would encounter the same problem as before.
So, please stop misrepresenting my position on this issue.
3 Likes
As for your doubts about my moderation of alt sandbagging, expressed here:
âŚlet me give you some clarity. My notebooks have the names and IDs of something like 1,000 probable alt sandbaggers spanning 2.5 years. Many of these players have hundreds of accounts. The two worst had over 1,200 accounts (no, I was not confused by schools, which are generally easily identifiable). The investigation of these accounts was a slow, tedious process. Consequently, I was able to deal with only a small percentage of them, and spent about half my moderating time doing so (because no one else was addressing the issue).
Why would I waste my time with someone who had only a handful of 19x19 accounts, let alone someone who had an extra account because he forgot his password? Obviously, I wouldnât and didnât.
4 Likes
In essence, drift is not really an issue for active players anyways. Itâs problematic for players who donât play ranked games for a long time and then have a completely inaccurate rank because the rest of the server drifted away. Perhaps the issue can be solved by (gradually) bringing volatility and rating deviation back to initial values for players who donât finish a ranked game for a long time, treating them more and more as new accounts, so to say?
3 Likes
There is already an ongoing discussion of that issue in another thread.
1 Like
Iâm aware, but this is related to the topic at hand: that letting players enter at custom ratings leads to drift, and that correcting for the drift will lead to more drift. (Unless the new players on average guess their rank correctly). One way to combat the effect of that drift, is by making the ranks of those most affected by it less certain.
3 Likes