Why does "restrict rank" allow ? players?

I agree completely, but when I talk abut it I don’t want to stretch people’s credulity. I had many experiences in which I found 5 or 6 unreported, cheated games in a player’s history.

2 Likes

yep same. probably true for most mods :man_shrugging:

1 Like

OK, fair enough :+1: But the other benefits:

and lack of previously cited disadvantages:

still stand though :grin:

Under the system of declared initial rank I am proposing, this would not be an issue :slightly_smiling_face:

Current system:

  • Beginner (true strength = 30k) is assigned initial rank 6k, gets auto-matched against ~6k, promptly loses, complains to mods / forums, answer is “play more games and it will adjust automatically”, beginner doesn’t want to have to lose 5-6 times before getting fair matches so leaves OGS :slightly_frowning_face:

New system:

  • Beginner (true strength = 30k) correctly declares initial rank 30k, gets auto-matched against ~30k, fair matches follow, everyone is happy, no complaints. It is fair to assume that most new accounts will manage to declare the correct initial rank provided the accompanying guidance information is clear and simple (and right there) :sunglasses:

  • Beginner (true strength = 30k) incorrectly declares initial rank 9d, gets auto-matched against ~9d, promptly loses, complains to mods / forums asking to change declared rank :roll_eyes: answer is “abandon account and make a new one with correct declared rank” (or just “play more games and it will adjust automatically” as before), no need for manual intervention. This will be a minority of new accounts anyway

Indeed, mods should never manually change anyone’s rating - just let the system do that automatically. Important point: Declared initial rank is just a short-cut to bypass the current [?] stage where the rating behind the [?] is way off. If someone puts in the wrong value for the short-cut, that’s just tough - now you made it a long-cut! :roll_eyes: You should have read the guidance notes that were right there next to the bit where you choose rank (we did remember to add some guidance information right there next to the bit where you choose rank, right? :wink:).

Mods should not, will not (maybe even cannot) manually change rating. The only way is to play more games or make a new account.

3 Likes

Beginners can just say i am a beginner, and the system assign 30k, not themselves. No need to explain that 30k is not 1k (or 1d, or 9d … )

3 Likes

Exactly! To quote myself (from Glicko-2 does not require accounts to start at 12 kyu. So why do we do that? - #62 by dragon-devourer):

So chances of mistakes with declared rank are small. And even if there are mistakes, there’s no need for mods to manually change rating. Therefore, no burden on mods due to declared rank :grin:

1 Like

I don’t know how you come from increased rating drift to stability.
We are probably in this scenario:

since data from old OGS suggests most players chose the lowest rank

while the average rating of new accounts seems to be closer to the 1500 mark

If we just adjust the rating->rank formula to account for the drift, inactive accounts will increase in rank over time. In this case my question would be: for how long do I’ve to stop playing go to become 9d.

1 Like

No, we’ve never been and hopefully never will be in that scenario. If people will on average choose the lowest rank possible more often, we will be in the following scenario as long as we fix this lowest rank at a certain rating:

The important thing to note, is that if sandbagging is indeed extreme, then the lowest possible rank to choose from will lie somewhere in the SDK rank. If on the other hand airbagging is extreme, then the highest possible rank to choose from will lie somewhere in the SDK ranks.

I don’t think sandbagging nor airbagging will be that extreme, but I can see the scenario where sandbagging is happening relatively more often than people choosing the correct rank or airbagging, resulting in the lowest choosable rating being around 15k or so, instead of 25k.

This shouldn’t be adjusted for, since that adjustment will be what causes drift. Allowing people to choose accounts at fixed points will initially cause some drift when there is more sandbagging on average or more airbagging on average, and this initial drift will be worse or better depending on how extreme the sandbagging / airbagging on average is. But, what my data shows, is that this initial drift will disappear over time.

(Of course we can adjust for how the rating correlates to kyu / dan appropriately: that’s just a label we put on top of it and has no effect on Glicko)

1 Like

I don’t think it’s just a label though because the kyu/dan correlates to the ratings, when handicap games are played.

Isn’t there some adjusted calculation of ratings done when handicap is played, so that the game can be treated as even-ish (from a rating point of view) when the ranks are different and there’s handicaps.

(Not to disregard the rest of the post - but just highlight something I think I can comment on :stuck_out_tongue: )

Edit:

What I mean is the kyu/dan labelling system directly impacts the rating calculation for handicap games. Or maybe one could translate the the kyu/dan into rating windows/anchors and ignore the labelling. I guess I’m not really sure.

Wouldn’t the rating pool drift down until it’s centered around this fixed lower bound? In my understanding it would develop to a point where the now decided lowest entry rating takes over the position the 1500 has now. (-> nothing changed in the long run)

We are already in the driftless state. Since the rating difference between players is fixed, any driftless state for the same pool should be indistinguishable from another, if we adjust for the overall rating offset.

So the question is, what are the choices for initial rating, which don’t introduce drift.


right now the entry point is 6k, well in the middle of the SDK ranks :wink:


Besides sandbagging, there is the tendency of humans to not like to loose, so users will more often than not chose a rank lower than their true rank. (We can account for this to some extent, but word of mouth will lead new accounts to declare lower and lower ranks. Besides many would complain OGS set their rank not to the selected one)
I’m quite sure I choose the lower bound of my self estimation myself.

My major concern is most users just sticking with the default. “I just want to check this out.”, “I don’t care right now”, …

Fair question :slightly_smiling_face: The answer is this: Assuming new accounts choose an initial rank closer to their true strength than 1500, then this will have reduced impact on their opponents’ ratings compared to starting them at 1500. Less impact on opponents’ ratings = improved rating system stability.

But how does that compare to their true strength? Maybe most new accounts are genuine beginners.

That plot does not show that the average true strength of new accounts is close to 1500. Of course the average rating of new accounts is close to 1500, since we start them at 1500. Those plots show the peak true strength is at more like 17-18 kyu (although it’s a bit hard to see - would be nice to see the histograms after X number of games each on separate plots so the scale is adjusted and the peaks can be seen more clearly). This is not really comparable to the present day as the system has changed significantly.

Also, the plots further down in that topic:

show the 50th percentile is around 16 kyu for September and 17-18 kyu for last year. So not 1500 = 6 kyu.

As @Vsotvep says, do not do this.

While this may be possible in theory in an extreme example, in practice it won’t work like this. While there may be a small downward drift due to an overall average of under-declared rank (e.g. true strength = 16 kyu, declares are 18 kyu), the effect will be small due to some cancellation with over-declared initial ranks (so the net is only slightly in favour of under-declared), and the fact that the under-declaration is only slight (a couple of ranks). Plus, there is still anchoring against other servers / associations, just with an offset. So we may drift to a point where X rank OGS ~ (X + 2) rank from other servers / associations. But that is OK. They’re all slightly different anyway (Rank - worldwide comparison at Sensei's Library) so these differences will also cancel out to some extent.

In the extreme scenario of every new account choosing 30 kyu, then in theory this would happen. But that’s just never going to happen in reality. We need to trust that most people will try to choose the correct rank. We can encourage this with suitable guidance information. Worst case scenario is that there is a benefit for those who choose the correct initial rank in the short term, and in the long term we end up no worse than we are now. But like I say, that is the extreme worst case. Reality will be much better than that.

This is covered by the options suggested previously (emphasis added):

and this would just revert to current behaviour, i.e. stick them in at 1500.

Would it? Where’s the evidence of that?

Would they? We cannot know either way. That’s just speculation.

That’s just anecdotal evidence. You would choose lower bound, I would chose best guess of my true rank, next person might choose upper bound. Everyone’s different so these effects will approximately cancel out on average.

As @Vsotvep has said:

1 Like

What I meant to say is that we can call 1500 10k or we can call it 1k or 15k, it doesn’t really matter. Once you fix one kyu / dan rank to rating, the rest does indeed follow from handicap requirements.

As long as the starting rating stays fixed, and beginners will be able to select, say, 1000 as lowest rating and 2000 as highest rating, we can always shift the label of what kyu / dan ranks these are to keep the comparison with external ranks as it is currently.

I’m not sure why it would centre around the lower bound, but I can’t immediately say it wouldn’t, either. In my example, it seems to centre around 900 rating points. I should test this by simulating a combination of sand- and airbagging at different playing strengths, but I don’t have time today.

4 Likes

To be precise, I expect the persistent sandbagger to enforce his entry-level to be equal to his true strength. The whole pool gets pushed down each time he enters with a rating lower than his should be rating.
Only if he enters the pool at the position he ought to be, he doesn’t remove point from the rating pool anymore.

1 Like

You quoted my own plots. The ranks on the axes are calculated with the previous rating → rank conversation formula, with an entry point of 1500=12.3k. 16k there is just 4 ranks below entry-level, which is now more like 9k.
New accounts (staying on OGS long enough) have a higher rating that I’d expect then.

  1. You can compare that rank distribution with the current one and the current influx.

  2. A more fancy way would be to use a faster adjusting rating system, like the current glicko, and compare the both distributions.

In the case of 1) we see a somewhat bell like shape for the current player distribution and influx, which is rather different from the exponential distribution we see in the historic case.
Either
a) The rank distribution change drastically over the years, with the early players barely knowing how to play the game as a whole,
b) A 30k than equal a 1d now, which would mean over the half of the bell curve got cut off, accumulating all lower than average players at 30k
c) both plots show totally different things. As both claim to show the strength distribution of players on OGS, I would at least expect both to be about the same shape class,
d) a slow changing rating algorithm kept one of them closer to the initial ratings than the other
e) the difference are just a fluke of chance

who knows

1 Like

Two years of “my rating dropped by 5 points after I won a game” gave a preview of it. If it is an obvious bug like “I selected 12k, but my rank got initialized at 7k” is obvious and worth reporting.

1 Like

That was a bug, now fixed. :neutral_face:

That would be a bug. We would want people to report that. :slightly_smiling_face:

Assuming declared rank was implemented without bugs, OGS would set the rank to the selected one, so no one would complain :face_with_raised_eyebrow: I’m not sure what point is being made here :thinking:

There is no “removal of points from the rating pool” in Glicko. The deviation and volitality make sure of that :wink: For example:

  • Sandbagger, true strength 1d, creates new account at 30k. Deviation is high because new account.

  • They win against a bunch of established SDKs. Sandbagger rank increases a lot because SDK deviation is low. SDK rank only decreases slightly because sandbagger deviation is high.

  • Sandbagger reaches 1d so they start to lose on purpose. Deviation is still high because rank has been changing a lot. And now, because they have gone from winning all the time to losing all the time, volatility is also high.

  • Sandbagger loses to established SDKs on purpose. Sandbagger rank decreases very fast because SDK deviation is low and sandbagger volatility is high. SDK rank only increases slightly because sandbagger deviation high.

Net result is SDKs lost a tiny bit of rank and then some other SDKs gained a tiny bit of rank. Both so small that it gets lost in the noise of ratings changes from other opponents and dominated by changes due to opponents with low deviation (i.e. established and consistent accounts). Sandbagger rank is all over the place so deviation and volatility remain high. No net removal of points from the ratings pool. :sunglasses:

I don’t think it is meaningful or helpful to compare to the historic case. As noted, there are many possible explanations and finding the truth is difficult, if not impossible.

But I just don’t buy the argument that an overwhelming majority of new accounts would declare an initial rank of 30 kyu to the extent that the rating system breaks. Declared initial rank is used on DGS and it works great. If the above claim was true, DGS ranks would be broken but they are working nicely.

Simple guidance is all that is needed. From the DGS sign up page:

And when you go to your profile to set your initial rank, there is an association / server drop down and an optional rank converter, plus this guidance:

And there’s plenty more on the DGS FAQs (too long to quote).

I’m not saying their guidance is perfect, but it does the job.

Remember, the whole point of this is to avoid the frequently reported problems with [?] accounts, from both the new player and their opponent, in getting unfair match-ups. This is simply not a thing on DGS. So OGS should learn from them and do declared initial rank. (Incidentally, I always though DGS was using Elo, but now I look closer I think they are actually using Glicko as well :sweat_smile: Further evidence that IT’S FINE!)

3 Likes

Yeah I don’t think so. DGS said they use the EGF rating system, but I’m very certain they mean the old EGF system E.G.D. - European Go Database | EGF ratings system which is like a modified Elo system.

The Egf changed their rating system recently (like this year or something) and recalculated the ratings.

I doubt DGS also did the exact same recalculation of all the ratings to make that statement still true in the FAQ :slight_smile:

Ah, yeah, ok. That’s probably why I thought it was Elo then.

Although, Glicko is modified Elo…? :wink: :laughing:

2 Likes

The EGF system is still a modified Elo system. In fact it did not change all that much. The update only tweaked some parameters, such as the (implied) Elo to rank conversion function, the K factor function and the bonus function.

1 Like

Alt sandbaggers almost never do this. When an account reaches within about 2 stones of their real rank, they abandon it and simply create a new provisional account.

1 Like