Glicko entry points and new players

This discussion has been going on since before I joined OGS. Given that it comes up time and time again, I think that shows that there is something to the point that beginners get a rough welcome here.

I don’t know how to show proof, mathematical or otherwise, that the ranking system will not be adversely affected by any particular change to the system.

In the interest of moving things forward, can any of these actions be taken to improve the situation or provide more data on what any further changes might do?

• Adding a prompt on sign-up asking “Are you new to Go?” with Yes/No answers resulting in different starting rank and/or volatility and/or deviation .

• Adding a prompt on sign-up asking “Are you new to Go?” with Yes/No answers having no user impact but resulting in a data gathering process.

• Removing the 9 stone rank restriction on ranked games so new accounts can challenge 20kyu+ accounts

9 Likes

Like I linked way back

Where you can find some info like

Where there’s a lot of comments on what the rating system needs to do

and so on, have a read.

This was 2018 bear in mind.

I think some simulations were done when humble rank was introduced originally

I’m not sure any of these directly answer @benjito but just trying to help find things somewhat relating to the way the current system is now.

3 Likes

Thanks @shinuito.

I think these quotes are helpful to paint a picture, but don’t support some of the key claims made in this thread. Namely these claims:

  • That sandbagging is a distraction and a second order argument (comment)
  • That anoek has “run the numbers” on the pick-your-own-rank proposal and determined there will be rank drift (comment)

In contrast, the core argument against pick-your-own-rank seems to be:

6 Likes

Suite du sujet I think the 13k default rank is doing harm [Closed] :

I took time to read the following answers and what I noticed is that for a certain number of them it doesn’t start from offering a new system to attract and keep beginners but instead to mitigate the effect of crushing strike for the very few full beginners we have.
iMHO that’s a wrong perspective, that’s not the way to keep newcomers by not considering their fun because you know there are so few around here.

There are already confusion from some who think that losing a lot is a fatality inherent to the game. Although they survived and kept interest, they didn’t realize by their own experience that if they were matched to same level players, they could have enjoy early wins along their discovery of go.

I may go further in this elitist test where only the bravest may earn the right to be a go player. That could be linked to a western culture value in which you have to suffer to get some positive progress in life. Not sure that is what new to the game are searching on internet.

There are some nice proposition in this 2018 thread already which appear here again, but then the debate turned mostly about the humble rank as the magic answer.

The “are you a beginner?” button was already suggested but disappeared pretty quickly out of the debate.

A beginner found a solution. Lost all his games with a bot to get his ranking adjusted. Can’t we propose better experience? Or if not, let’s bring a pop up with this elegant solution

Small note: 13k beginner with high or low confidence just doesn’t exist. A 13k is politely said not that strong but still much stronger as a beginner.

4 Likes

Thanks for finding that - I stand corrected! I guess this makes it a lot easier :woman_shrugging:

3 Likes

This issue of how to initialise new players came up at my Go club yesterday, because we started using a rating system recently based on the OGS system. In the Go club the initial rating does not matter much because we ask new people how strong they are and give them handicap (or not) depending on what we know about them. The system rating starts at 1500 independently of their stated level and soon converges on a level representative of their strength depending on their results.

This got me thinking that maybe something similar could improve the experience of new players at OGS: ask them how strong they are and adopt a shadow rating used for match-making based on what they say, without changing the underlying rating calculation. The shadow rating and the true rating should converge after a few games. Revert to the internal rating after X matches, discard the shadow rating. Not sure what X is but probably not more than 10 games.

Hope this is helpful, comments welcome, looking forward to the point-by-point rebuttal!

Alex

9 Likes

This is kind of a side point, and maybe doesn’t belong in this thread, but it feels to me as if it does.

I prefer handicap games in general, and would like to see more handicap tournaments and so forth. All the offers I put up for anyone to accept are auto-handicap. Not only do I not want to get slaughtered by a much stronger player, but even more I don’t want to spend weeks in an unhandicapped correspondence game with a 20k who isn’t even looking for a teaching game.

A few times now (I started playing on OGS a few months ago), I’ve had to write a reminder to myself (which so far I haven’t remembered) to politely explain to any ? player who accepts one of my handicap-game offers, that the auto-handicap can’t work with an unranked player, and I prefer not to add another non-handicapped game to the tournament and ladder games I’m already playing, and cancel the game.

On one occasion I think this got my first and so-far-only game with a sandbagger, but on the other occasions I had to kind of slog through a game where I didn’t want to resign and artificially inflate a newbie’s rank (leading ultimately to a bad experience for them, as extensively discussed in this and other threads), nor possibly offend them by offering a teaching game when they hadn’t asked for one.

Why are auto-handicap offers even visible to ? players?

They ought not be eligible to accept, since the requested handicap can’t be set properly. In the meantime, I’m going to try to remember the polite explanation and game cancellation in the future whenever another ? player takes up one of my offers.

2 Likes

Love it! This is exactly the kind of thing I was thinking about.

On an unrelated note, what are you using to track ranks of your club members?

3 Likes

Thanks.

We ask people to register the results of their matches in a Google Forms questionnaire. The data is accumulated in an online spreadsheet. I wrote a simple R Shiny app that reads the data from the spreadsheet, calculates Glicko-2 ratings using the PlayerRatings R package, and publishes the results.

Further details here (in Spanish):

1 Like

One could liken this to a “provisional rating system” sorta like how humble rank kinda works, but with a bit more freedom, where there is a real and provisional rating calculated, but the provisional is discarded once there is a certain level of convergence

4 Likes

I found a formula for humble rank but I’m not sure how, why or when it is used. Is it used for new players until their estimated rating is considered reliable?

  • Why: to alleviate the pain felt by newer players who joined and got matched in even games against the median rank (6k)
  • Who: everybody (but the effect is smaller for long-time players).
  • When: iiuc, it is only used to determine handicap, not matchmaking (source). Arguably, it should be used for matchmaking.
2 Likes

I’m feel fairly certain it’s being used in automatch, but it might not be in use or being checked properly for rank restrictions on custom games.

Otherwise I don’t know how you could automatch with players at 25kyu after like two or three games.

3 Likes

the quote sourced is a little confusing, namely

which makes me wonder if this is meant to somehow imply it is not applied for automatic matchmaking, which has a more direct relationship to rating,

That being said, I’m not certain why it’s not used for custom

1 Like

Yeah I’m not sure whether that comment was just about custom or all matchmaking. Might be something that someone would just have to test out.

2 Likes

… just encountered one more fail of OGS system
the first opponent [5k] resigned because bored
new user still actually plays like 25k, but got stable 13k


14 loses in a row only to drop from 8k to 13k


Lose Your First 50 Games As Quickly As Possible

lol, that’s literally what some new OGS users have to do

8 Likes

actual beginners don’t understand rank system
OGS should ask new users “are you beginner?”
and if they choose “yes”, they should not be allowed to play ranked game with anyone with higher rank than than their current rank until [?] disappears. Because that’s what Go beginner who for some reason knows OGS system and Go rank system would do now. System should help beginners.

4 Likes

The player had 2 ranked games against guest players. That shouldn’t be allowed.

2 Likes

tbf, this player seems to be seeking out higher level players (such as that 2d) in custom matches, which will naturally create a slower decline, so calling it “stable” feels like a misnomer

2 Likes

The 14k-5k offer acceptance is an indicator of how this player is confused by the system, if @stone.defender is right in the estimation of the player level. i really wonder what kind of experience he’s getting when he simply sign in for hopeless offers he gets with an overrated level by the system.

4 Likes