Actually there are three numbers in Glicko-2 – rating, deviation, and volatility. We just don’t get to see the volatility
To be honest, I remember what it’s like to play noobs while at ~13k and it was extremely boring, and on the other end was the frustration of experienced unrated players. I knew about this issue, and I make it a point not to cancel noob games so that things don’t get weirder (like the SDK Conrad mentioned).
I guess I’m up in my little SDK privilege that I was thinking the issue was when experts make new accounts, because those are the people sent my way.
I believe I have mentioned before elsewhere that it feels like enforced sandbagging and initiations by fire, and may not be worth the accuracy, and I still somewhat stand by that.
What about an option to exclude people which are ranked as “?” from the automatch system and from accepting games posted? I mean a similary system to the one that lets you exclude ranks from accepting your games.
I think asking people their experience level is the best way to handle this. Lots of new people get paired up with a ddk/sdk opponent, get smashed, and never come back. Just let people choose Beginner/Intermediate/Advanced, and then start them at 25k/13k/6k based on their response. 25k is low enough that most new players will at least find their games somewhat interesting, and 6k is low enough that I don’t think letting new players go to 6k will mess up the rankings.
I’ve said this before on the forums, but in my opinion the best solution is to ask new players for their rank in a simple drop-down menu. The list could either be absolute (30k - 7d) or slightly truncated (eg. 20k - 5d). The lowest rank could be explicitly marked “total beginner”. After choosing a rank from the menu, there would be no uncertainty period.
I think this method best reflects the reality of people who join OGS or make alternate accounts here. By and large they’re not simpletons or bumpkins who can’t understand what rank is, and they’re not trolls or sandbaggers either. New accounts are going to be either beginners, who’ll go for the lowest rank, or experienced players from other servers who’ll already have a rank to refer to.
I find it ironic that people are using “I can’t get even games” or even “some people theoretically won’t get even games” as a reason why they don’t want to use the proven improvement on overall site rank accuracy.
I think 13k puts OGS at a position that it’s neither beginner friendly nor it help the server attract strong players (which is what OGS lacks). I think the purpose of making it 13k is to make the initial rank cater to people from a variety of rank, but in reality, it just made OGS an ~13k- friendly server which is kinda awkward.
An (not so good) analogy would be like taking a poll of where people wanna go, and some people want to go to far west, so want to go to the west, some people wanna go to the east, some want to go the Far East some want to take at home, and in order to take everyone’s opinion into account, they decide to sum everything together and take the average, and boom, on average, we decided to not go anywhere.
Being ~13k friendly neither helping with expanding OGS’s user base by attracting new beginner nor expanding the strong player base which would help to attract strong players from other server.
And if you are a high dan new to the server trying to rank up, it’s gonna be a pain. Unless they got a friend here who’s willing to donate points from their alt to help them rank up quickly. After playing a few game making progress to higher sdk, you got match up with a sdk beginner(yes, they exist nowadays) who refuse to resign, you got back to square 1 and left utterly confused about the ranking system here and left the server.
I think KGS did something like that at one point, and a ton of non-dans registered as dans and inflated the ranks. Playing a game or two at SDK doesn’t seem like it’ll seriously drive dan-ranked players away from the server.
I could see letting people register as their rank, up to 5k or so, and then adding a 5k+ rank, which would start with higher uncertainty, in order to cater to dans. I don’t think people should be able to claim a dan rank, though, without playing any games. Even then, I think some banding would look better from a UX perspective. Having a drop-down with 20+ options can get somewhat miserable, especially on smaller devices. Beginner/30k-20k/15-20k/10k-15k/5-10k/5k+ gets people pretty close to the right rank without the UX issues.
Note that no-one is criticising Glicko for being a bad change.
It’s good that we have improved site rank accuracy.
So I don’t see the irony.
And it is really an inaccurate paraphrasing. “I can’t get even games” is a generalisation. The issue with the new system is much more specific than that.
Even “some people theoretically might not get even games” is at best a spin on the actual issue.
The actual issues are that:
Certainly (not just theoretically) beginners who join will get games that are not just un-even, but are out of their league
Along with that, demonstrably, their process of getting a first game will be marred by people cancelling out on them
and SDKs/experienced DDKs will be faced with either playing out a game against a noob that they weren’t expecting, or rudely cancelling out.
and experienced players will also have the experience of games being cancelled out on them by people who think they might be noobs (once again, demonstrably: it happened to me).
Beginners can’t even set up games with other beginners if they want to, because their rank is so far away from 25k that they aren’t even allowed to by the system!
These are the specific problems being discussed.
To me now we have a double irony: we have real problems caused by the current situation, and the only argument for the current situation is a theoretical one: theoretically maybe the rank system will be perturbed by people lying about their experience.
But even this gets balanced out. Some people will lie and claim they are dans, some people will lie and claim they are beginners. So these feared outcomes balance each other out. I don’t think a system should be designed around whether a few people lie, anyhow especially if that makes the normal case do harm.
But KGS didn’t have uncertainty to help them, did it?
Whatever starting rank people get, it should be accompanied by a high uncertainty. I’m not even clear why the 13k that people get now doesn’t have more uncertainty given to it.
And with high uncertainty, that means that the impact of these ranks is much less (right)? One person’s rank only starts materially affecting the other’s as its uncertainty reduces.
Real world experience: I taught my adult brother Go, he only played me a few times but enjoyed it. He knew the rules but little strategy due to limited experience. I encouraged him to sign up for OGS, play lots of games. He did, got 13 kyu “beginner” rating, played a couple games and got shellacked, tried to find games by creating his own, OGS limited his ability to play others too far from his ranking (he wanted 25 kyu players), came back to me and said “I like Go but I’d better play you some more because I’m not good enough for online Go yet.” Hasn’t signed into OGS since. End of story.
Huh - I forgot about that aspect of the problem: beginners can’t even set up games with other beginners if they want to! I added this to the list above.
From my understanding glicko needs a central starting point. lichess.org does it exactly the same way (they also start at 1500 elo) and I don’t for a second imagine that that is a conincidence.
You can’t simultaneously say glicko is great but i wanna pick what rank i start at because that stops being glicko.
If a new player plays 2-3 ranked games they can fall all the way to 25 kyu (let alone the 16 kyu required to play ranked games with 25 kyus) so I’m guessing the issue is more that new players are not playing ranked games for whatever reason…
Furthermore, new accounts DO NOT start at “rank 13 kyu”, they start at “rank ?” they may well have the equivalent elo score of a 13 kyu but that is not the same thing. The system is designed so that losing to an unranked player does not affect your ranking, meaning if people can refrain from being a dick and just let new players play a couple of games they will all get sorted to their appropriate area of the very large rating pool.
I have about 50% game cancellation and that has remained about the same both before and after glicko was implemented so don’t try and tell me cancellation has gone up because it was always a problem.
Back when you could pick your own rank, mods and admin were constantly flooded with requests to change their rank. Now we have a new system specifically designed to find an accurate rank in very little time so that the mods don’t have to worry about all of that.
If anyone plays 3 games and quits because they lost them all, they were going to quit anyway…
Also, from my own personal experience… players at roughly 13k are more likely to give new players a gentle game than ones at 25k…
almost every 25k game I see ends by resignation after a massive group dies. At least if an experienced player finds they are paired with a “rank ?” player they could possibly have the decency to ask if they’re a brand new beginner and perhaps offer them a teaching game (which can still be easily won without humiliation)
So, the way I see it… the only people that will leave this server because of these supposed “problems” are the people who were never going to play anyway, and the dicks that can’t spare 15 minutes of their day to help out a new player… sounds alright to me don’t let the door hit you on the way out if that’s the case
I really disagree with that assertion. Yes there are these people who have a deep interest in go and won’t quit no matter what, but there are also a percentage of player who may or may not quit go depending on their game experience. So we shall try to make a system that makes their life easier.
Also, There’s a big difference to losing 3 games and being completely rekt and shattered in 3 games. If I were to lost my first 3 games the same way as I lost to an 10k bot (where I have no idea how all my groups turned out dead in the end), I would’ve thought I have no talent in this game and quit. I was lucky that back then I got to play with other beginners, even though I lost, I can still kinda make sense of their moves and that was quite fun. I think a player is mostly likely to quit in the beginning phase, so I see no point in making the barrier of entrance higher than it should be.
It’s difficult for a player to figure out what’s going on with the system if he doesn’t speak English. And I’ve seen beginner settle at different non[?] rank for different reasons… such as opponent resigning early cuz they kept playing random moves. Which would cause more confusion to them on why their opponent quit and why they got completely shattered at their rank. There are non-English speakers in this world who can’t just pop into English chat and ask for help. If we want to keep the 13k I think we should at least have something that explains the ranking system to beginner. (ok that’s another proposal)
Yes but a brand new beginner will get equally smashed by a 25k as by a 10k… at least with the 10k one can hope there is some kind of chance they might have mercy and actually take the time to teach a lesson. Bots always play weird and are harder to understand than humans, especially lower ranked ones.
I agree starting out in the middle can be confronting when someone thinks they’re starting at the start / bottom… i had a similar experience when i first started playing chess. I started at 1500 elo with the pre-conceived notion that I was rather good at the game since I played in a few clubs IRL and got handily smashed in online play… but that could all very easily be mitigated with some kind of disclaimer at account creation that says very clearly
"Your new account starts at global average, new players should be prepared for a swift drop and advanced players should climb quickly. You should start getting paired with roughly even strength opponents after 2-5 ranked calibration games."
This has got me thinking. This and the assertion that Glicko must have new people “starting in the middle”.
I can’t debate that assertion, I just have to take it on face value.
So - the thing is in presentation terms they start at “?”, but technically, they are at 13k. Most importantly, in the ways that matter, they are treated as 13k: that’s how the “get a game” system sees them.
So if we have two indisputable things
Glicko needs your technical rank to start at 13k
This makes the initial period experience bad (there’s enough evidence in this thread to assert this as given)
then we need to look for solutions within that space.
How about this: start everyone’s technical rank at 13k, but give them a “qualifying rank”. This is the rank that they ask the “get a game” system to treat them as. I think it would also be shown. So a new beginner would look like 25k? and a new SDK would look like 6k?. They keep this qualifying rank until their uncertainty reduces to an given level.
Now, the get-a-game system lets beginners play other beginners. Glicko kicks in and computes the technical rank as it always would have. As soon as the 25k? (technical 13k+/-9) person loses the first time, their technical rank will plummet as will their uncertainty. Similarly, as soon as the 6k?(technical 13k+/9) person wins their first match against another SDK, their technical uncertainty will reduce and their technical rank will increase to the right level, and they lose the ? and start being treated as their glicko rank.
Sounds like an interesting proposal… but my gut says it would be a nightmare to code… overall though if what you suggest is possible, it sounds really good!
OGS is the first sever I played on so I know what being a beginner here feels like. After losing (actually not so badly) my first game.I managed to won my second game on OGS as a brand new player (with “knowledge” of go from hikari no go). Playing 10k is just a totally different experience. And regarding the teaching part, honestly speaking beginners tend to be friendly and chatty… I had some good conversation with a lot of my opponent . As you move up the rank, conversation became sparse, as stronger player tend to focus on the game itself more.
We definitely need a better explanation of the ranking system here… Like putting a help icon next to your rank in your profile page and explain what that complicated looking matrix mean, and how beginners are expected to drop to 25k etc .
Actually no… Glicko doesn’t NEED it, Glicko RECOMMENDS it. And the statistics when setting it up (as we are told) has more accuracy when it starts at 1500.
But I personally doubt that accounts for more than one percentage point and is probably not worth the initiation by fire.
That being said, I really like GreenasJade’s suggestion, but I have one question: when should the ? go away? When uncertainty drops? To what level? What if they don’t play often and so uncertainty doesn’t drop quickly? Like, if someone is a 20k? and they drop out for a month and come back a 13k in actual strength and are still 20k? What then?
I know this is even further removed from reality, but that indicates a move in the right direction. To my question I would suggest a time limit on the ? rank or a certain amount of rating drift away from the alleged rank. (As in if a 25k? raises in rating a certain amount (even after dropping) they will lose the ? rank)
don’t let the door hit you on the way out if that’s the case 