Proposal: New users choose beginner/intermediate/advanced and drop ranked game restrictions

kinda, but it might be even more confusing if your brain automatically associates k with “thousand”, and wonder why thousands go down with the higher strength options

4 Likes

I didn’t even think of that; I would only associate “k” with “thousand” in contexts where it made sense, like “Such and such ran a 5k last weekend”, but never when reading a Go rank

I haven’t chose these rating entry points myself but i can accept them as the fruit of statistics (see posts up). They still surprise me but i should see them as entry, so more to the bottom of a class, not as some middle point in a class.

Things would be more clear with a range (like Intermediate 14k to 6k )

1 Like

If this is the case, then it should rather be “22 Kyu” than “22k”, etc.

Oh, @meili_yinhua beat me to it.


Another thing: I understand that we want to keep it simple, but what about, say, 2d and 7d players? Shouldn’t we have such options also? (Not that I would need that :rofl:)

I don’t think that’s necessary. Since new players’ rating has a high standard deviation, dan players should rank up quickly enough. Moreover, on Fox it’s not possible to register at a higher rank than 3d, and 3d on Fox is pretty close to 2k on OGS so it should be fine like that.

3 Likes

FTR, for anyone arriving new to this thread, the consensus from the previous forum discussion (this same thread) is in the OP (edited after discussion).

  • Don’t show numbers by default. Instead, show a “?” for “more info”.
  • If users click on the “?”, show a broad rank range. The specific starting rank is not particularly relevant to new users and is subject to change. The deviation is still high enough the rank will change quickly in the first few games.

The ranges were / consensus was:

  • Beginner: (don’t show a range, but 25k+–12k)
  • Intermediate: 16k–1k
  • Advanced: 4k–9d+

These ranges are/were intentionally overlapping and broad.

In the interface design phase, we decided to break “Beginner” up into “Basic” and “New to Go”. So I think the right ranges are:

  • New to Go (don’t show a range, but 25k+)
  • Basic: 25k–12k
  • Intermediate: 16k–1k
  • Advanced: 4k–9d+

My opinions:

  • Strong opinion: do NOT show a specific rank, but instead a range. The initial rank itself is subject to change as we see the effect on the rating system. The range is the intent/goal. The starting rating/deviation is an implementation detail.
  • Strong opinion: do NOT have any visible range for the weakest level (“New to Go”). New players shouldn’t fuss about some weird thing they don’t understand.
  • Weak opinion: for intermediate/advanced, either have a “?” (forum consensus), or just show the ranges straight up. Either solution is better than not showing ranges.
    • There was some disagreement during interface design, and I was fine leaving the “?” off to start off and see what happened, but I think this will result in “reddit-as-documentation”. This is a problem, because “reddit” will document the starting rank (not relevant), but we would prefer players choose by the intended rank ranges.
  • No opinion: whether to have a range available for basic. Probably many players at this level won’t actually know the ranking system or where they sit in it. But maybe many do.
10 Likes

Yeah somehow i’m bit hesitant of handing out dan ranks right at the registration, it feels like it can really screw the match-ups for legit dans (used to be a real problem when people were allowed to register as 6d, ogs dan ranks turned into total mess and ogs got a bad reputation within stronger players). Also there isnt actually that many new dan-level users registering here, majority of all new users seem to be pretty new to the game.

Maybe it would be ok to have “?” next those buttons with a small table saying where each options puts you at first, i think more expereinced players would appreciate that ^^

6 Likes

I would rather drop away that question mark because of its bad reputation.

To put an explanatory table on the side is a reasonable idea for experienced players.

3 Likes

No way, up some level (2k imho) you have to validate your level through competitive play.

3 Likes
  • No idea
  • Almost know the basics (22k)
  • I know the basics! (12k)
  • Oh dear I need to learn the basics (2k)
  • Ok I think I finally know the basics (5d+)
8 Likes

Just noting: this conversation is now mixing up

“What do we show on the choosing page”

and

“What are the choices (what rank actually is assigned”

5 Likes

:grin:
Haha being around that strength, I often think this ^^

(And to this, maybe, maybe not :rofl:)

It reminds me of the story from Treasure Chest Enigma (if I recall correctly) of a top pro remarking, ‘I wish I knew how to play this game’ ^^

I like the ranges idea & displaying them , better also ^^

(Edit : or if that is too cluttered/complicated looking and may seem confusing to beginners, programming the buttons displaying something simpler to read

(for example the image in stone_defender’s post, with the closest single starting rank being implemented at the time, which can change/update automatically as the implementation changes)

And agree with this.

I would probably have done the same at SDK level and not chosen ‘‘advanced’’ until at least 3, 4 or 5d myself (or at the least a dan rank) either. ^^

I think it’s probably not intuitive to everyone (or even not what is expected to many people) what the ranges refer to from the terms alone.

5 Likes

It’s worth mentioning why were the ranks left off.

My recollection of the discussion is that a strong goal of the design discussion was to achieve the highest possible “decision to click” for beginners.

The reason for this priority is that beginners are the worst problem if they are ranked wrongly.

It really really does not matter if a 2d thinks they are intermediate and starts at 12k. We might get the occasional “sandbagger” report, but the 2d will rapidly get the correct rank because the starting deviation is high, and there is no “quitting in exasperation” associated with this kind of problem. The correct person wins the initial ranking games.

In contrast, if a beginner clicks “skip” (or bypasses the page) then they are ranked at 15k for initial ranking games and can end up easily in games with SDKs. These are awful because SDKs abandon them in disgust, and the beginner get their rank increased.

This is the specific problem that the whole system is aiming to fix.

So the argument was “make this screen as easy to click FOR BEGINNERS as possible”.

And that led to “don’t show anything they don’t understand”.

Personally, I think stone defender’s suggestion is the correct one.

7 Likes

I also like @square.defender’s suggestion, as long as it’s a rank range on the buttons, not a specific starting rank. We want a 6k to click “intermediate” (16k-1k), not “advanced” (4k-9d+).

6 Likes

Great! Let’s do it then :slight_smile:

For specifics…

While stating in advance that I don’t care strongly, I’m not wild about the idea that 12k and 25k are in the same bucket.

I think this has its roots in the fact that we don’t really agreed on what 25k represents (noting for the others that we’ve had this conversation 1-1 :slight_smile: )

My experience is that 25k at OGS includes folk who play random stones, and don’t know how to end the game.

Someone who clicks “Basic” is going to get a 22k starting rank. It just doesn’t seem right that someone might look at a few games, see that they compare in skill to 14ks here, and end up with a 22k starting rank.

Of course, this is in the category of “it doesn’t matter” … they will rank up quickly, and they will not create the “too highly rated beginner” problem, but they will wonder why they are paired with folk who barely know the rules, and they will smack those “fragile beginner” folk around on their way up…

Do you think we might give ranges only to Intermediate and Advanced? In this way, we are acknowledging that Basic people probably don’t even know their rank anyhow - they don’t even know what a rank is. Wheres intermediate and advanced people probably are coming from elsewhere and do know what their rank is.

The goal of that suggestion is to encourage people who really do have prior experience to choose intermediate.

That doesn’t sound like a 25k, but rather someone who either has an inflated rank, or is correctly ranked lower, but OGS isn’t displaying their rank (I thought they lowered that, though?)

1 Like

That’s what @dexonsmith said too :slight_smile:

As far as I’m aware we have not lowered the displayed-rank-floor.

As soon as we do that, if we do, then we surely would move the “New” starting rank down with it, and this discussion might go away…

2 Likes

Generally the idea is to start players near the bottom of the range.

  • If they belong at the top of the range, then after 2-3 wins, they’ll be there. Fixed.
  • If they belong at the bottom, they’ll start with well-matched games. Great.
  • If they belong a whole range down, it just takes one loss to get them close. Fixed.

We can see how it works in practice, and adjust the precise starting rating. Maybe centre of the range is better. Maybe top of the range. The deviation is high so they’ll move quickly regardless.

But my inclination is bottom of the range, since starting with a few easy wins feels better than starting with a few terrible losses.

4 Likes

Correct. Not yet.

I’m looking into this with @anoek, and I think we can/should lower the floor as part of “rating system v6”. There’s some data analysis I need to do, looking at ratings over time, but the hypothesis is that with a better understanding of small board handicaps the rating system will successfully differentiate players weaker than 25k.

If v6 does lower the rating floor, we’ll probably also retroactively give a lower-than-25k starting rank to “new to go” players.

Progress on v6 has been really slow because I’ve been busy with non-OGS stuff the last couple of months.

6 Likes

Out of curiosity, why is “consideration of handicap” needed to determine if we can discriminate below 25k?

I can appreciate that if we have “wrong handicaps for small boards” then any rating considerations are simply messed up, so that needs to be fixed before the data is helpful.

However, although the rating system is expected to be calibrated by handicap, it’s ability to discriminate rating levels should be independent of handicap usage … it has to work in an environment where no-one even plays handicap games.

My simplistic (naive?) view is that there’s simply going to be a rating level where below this rating the player’s play is indistinguishable (in win rate) from random play, and that’s the rating floor.

2 Likes