I think this is the way it should be also and suggest that the threshold for entering a tournament at 1d should be 2050.

Historically perhaps our grades have been aligned with Japanese grades, whereas on the continent of late, the benchmark has been Korean or Chinese ?

A new field in the UK rating list could show the highest dan grade achieved - say red=shodan, orange=nidan, yellow=sandan ... and players could wear an appropriately coloured belt to tournaments or social events

Just to correct one error in the above: GoR only uses players grades when calculating ratings for players who are new to the system or who enter a tournament at 2 or more grades above their previous entry grade. (I don't know whether it applies to 2 or more less)

This latter case is intended to partially help overcome the anticipated deflationary effect of the system.

One important difference between FIDE ratings and GoR is that FIDE doesn't have to correlate the outcome of their system with an external system that measures the same essential thing, i.e. dan grades, especially from countries not part of the GoR system.

I agree. My personal preference with respect to my own entries has been:

3.3k-2.7k - enter as 3k.
2.6k-2.4k - enter as either 3k or 2k.
2.3k-1.7k - enter as 2k.

etc...



Hehe, that would be fun Of course, historically, I believe shodan is in line with "Black belt 1st dan" vis a vis martial arts belt systems. I once tried to get rough belt equivalents and decided it was too pointless an exercise to try. However, the colour on the rating list sounds fun, although it may simply encourage people hugely valuing their ratings above other aspects again, which I'm hoping we can avoid.

Jon, I remember in our conversations at one of the council meetings the idea of resetting our 1d rank to 2100 was discussed and rejected, but I can't remember the reasoning. With the advantages it would have mentioned in the above thread, are there reasons why we prefer drifting the 1d bar to lower and lower GoRs? (since joining in the council, it's already dropped another 5 GoR points)

If you do this in the UK, a UK dan will be much stronger than an average dan from another country. It only works if this reset is done on for all EGF members.
this will cause loads of people to lose (dan) grades, Which they probably do not like. It is also usually against the national policies (it is in the UK).
at kyu level it will limit your entry rank to much. people do not play enough tournaments to allow the rating to keep up with that.

Just had a quick though about it and.
keep things as is for 2kyu. Next, give a 1k a rating reset to 2000 at the first tournament they enter as 1k and fix rank and rating for dan players. so a 2099 is 1k and 2100 is 1d and when a 1d passes 2200 he automatically becomes 2d. This does not solve the problem of dan players joining tournament for the first time though.

_________________
Laurens
Go school: freegame's Teaching School
Author of: The Next Move a book for 15-3kyu players.

Is it against UK national policy? If so, where?

The rating won't need to "keep up", as it will be a one-off reset - yes, it will make a UK dan stronger, but I think our dans are already slightly weak anyway compared to places like Netherlands, Germany and Finland, so I don't see that as a bad thing. I'd much rather be a strong 1k than a weak 1d.



You can't do this. The resets are completely constrained by the EGF rating system. If it was possible to reset as GoR 2031 (for 1 dan UK) rather than GoR 2100 then we'd have less of a problem, but you can't, and the EGF have said they are not considering a change to this (in the past, anyway).

At the moment, the tail is rather wagging the dog. If there were static benchmarks such as 2100 as 1 dan, and you lost points, you drop rank. The 1 dan benchmark shouldn't be moving to keep some sort of parity with the players - the players should be maintaining 2100 to maintain 1 dan.

One idea would be to adjust UK 1 dan up a few points each month, maybe 4 or 5, until it returned to 2100. That way, no-one loses a big rank in a short time period, and it takes about 17 months to get 1 dan to 2100.

Not really. Three is just a very low number of games. Suppose this player had gone to the EGC as 3 dan, and scored 6/10 in the main and 3/5 in the weekend tournament, with a resulting rating of ~2300. Then you're suddenly talking of 15 games. Similarly, playing long weekend tournaments with 6-7 games (Paris, Amsterdam, etc) would already give a much clearer picture.



What I meant was: Your first instinct, when receiving the tournament result with an entry grade of 3d, is to change the player's entry grade to 2 dan in line with the regulations. You should consider not following that instinct only when there is a compelling reason for keeping the entry grade at 3 dan. A compelling reason would be if the 3d entry grade looks reliable.

Which surely is the catch 22? You will never get a "reliable" 3d from the small number of games, so the reset will always be considered appropriate. If you require a larger sample size of games, then in reality being able to reset above 2d will not happen. The EGC example is a very rare one, most tournament players have still never been to a tournament with longer than 5 or 6 rounds.

I don't think policy should ever have the default action as an interference. In my opinion, meddling post-tournament with entry ratings should be the exception, not the rule.

One reason for rejecting the idea was that it would imply reducing the grade for more than 50% of our existing dan players and insufficient statistical justification was provided for doing so.

Citing the theoretical basis for the GoR system as 2100 for 1 dan wasn't judged a statistical justification.

My personal view is that GoR has a deflationary impact of about 4-5 points per year, so what we should actually be doing is adding about 60 points to people between 1990 and 2090, gradually tapering above and below that level, and changing the GoR system so that it's no longer deflationary.

If someone beat 3 3-dans then prima-facie the strength is likely to be >= 3-dan. (If there are 8 players at 3-dan in a 3 round tournament then there will always be one 3 dan player with 3 wins. We don't know whether the 3 wins are because he/she's stronger or just a random result.)

It is clearly the exception, as it's the only one that I remember coming before Council in my time (but that's only a few years).

I agree that post-tournament meddling should be an exception. In this case, the tournament organizer made an error in allowing the player to play as 3 dan, because regulations say that 2 dan is the maximum. This whole case should not have happened.

The regulations also allow exceptions to be made, if they happen with prior approval of the BGA. The player could have requested for an exception to be made, if there was a good reason to enter as 3 dan, rather than 2 dan.

Furthermore, lets not forget that EGF ratings are heavily inflated. Entering a British tournament as 3 dan is effectively saying: I think I can hold my own against other British 3 dan, which is equivalent to saying: I think my rating is probably within 50 points of 2240. The reset rating of 2300 is outside that range. In the case in question, resetting to 2 dan and processing the result puts the player's rating quite close to 2240, and therefore effectively has the result of making the player a British 3 dan.

There were three that I was aware of in 2009 alone, and you were president at the time. Ask Geoff.



Thank you, that's useful. What is the theoretical basis / statistical justification for the current system of moving the shodan bar? How exactly is the 1 dan mark defined (IIRC, it's the "average GoR of all players entering themselves as 1d", but I'm not sure the specifics). If this is true, then is it also true that if all 1k players entered as 1d this shodan bar could drop by 40-50 GoR points?

What would constitute sufficient statistical justification for the reset? I would be willing to work at producing a fair bit of data and reports if I had a benchmark of what evidence would be required for this.



I agree with the deflationary impact. There has been some (but apparently insufficient) effort to mitigate that in the GoR system. However, your proposal seems at odds with recent decisions and BGA policy. If we wish to add 60 points "across the board" to players by 20xx, then we should encourage resets into dan ranks as this will achieve precisely this over time. Regardless of how these 60 points are injected, it will still make our players "weaker" than their equivalent international peers, which also seems to be something we would wish to avoid. Unless we can convince other countries to do the same (which I think it would be naive to believe there's a hope of doing), we'll end up with some of the weakest 2100 players in Europe.

Changing the GoR system would be great, but AFAIK Geoff and others have raised this to Alex directly and the EGF AGMs indirectly to no avail. I understand that no country has much impact on GoR system policy. In honesty this is probably a good thing - European-wide rating systems should not be easily influenced by individual nations - however, without a system in place that seems to be addressing the efficacy of the system, I can't see GoR realigning itself any time soon either.

I think, looking at past performances and having looked into about 12 years of tournament history when I was in council, that European shodan now is genuinely stronger than it was 10 years ago, and perhaps 70 GoR points is a reasonable estimate of how much stronger. It seems either we are happy with GoR for shodan drifting, so that shodan is equivalently the same strength, or we would rather 2100 represent the value of shodan and need to either mass-adjust players up 60-70 points (as in your proposal, this seems theoretically ideal but practically I can't see how it is feasible) or to accept that it's harder to hit shodan now than it used to be, and let people's grades drop.

Issues as I see them:
* Current GoR system has ratings dropping for people maintaining their strength at dan levels
* Current GoR system can only handle resets to 2x00 ratings, which are not in line with the vast majority of European country dan grades
* GoR values for dan ranks vary by large amounts across different countries
* Actual playing strength represented by the same GoR varies by medium+ amounts across different countries
* The GoR value for 1 dan is not defined by playing strength, but by the unregulated professed rank within the UK, which may or may not be deflationary / inflationary in and of itself

Issues with resolving the issues
* We have no way of assessing actual strength of individuals, as both the player in question and an assessor will have their strength changing over time - there's no static anchor
* There is no way of easily regulating the professed rank when it's +/- 1 stone, nor is that particularly helpful to do
* Without regulating the professed rank, there is no other way of evaluating an "appropriate" shodan bar
* Without more international participation by each country, it is hard to gain a view of how closely correlated the dan ranks are of different countries
* Without more international participation by each country, it is not easy to control for or adjust ratings to correlate different countries
* We need co-operation with whoever is in charge of the GoR system to address the feasibility of resetting to GoRs other than 2x00

This makes the assumption that EGF ratings are inherently reliable. But EGF ratings are as much a free floating system as BGA grades are. The mere existence (and necessity) of the anti-deflationary epsilon factor in the GoR formula proves that the system is not inherently stable.

You'll have to ask Geoff about the justification for the current system, but the mechanics are defined in the FAQ.

If all 22 British 1 kyu players entered as 1 dan (and this was accepted by the TDs) then there would be absolutely no difference to the calculated ratings as if they had entered at 1 kyu after the event. Since the 1 dan mark is calculated across Europe I assume that the 1 dan correlation number would change by about 3 points. [I haven't done the exact calculation, but there are about 470 European 1 dan players.

I don't know, but I'd probably know it when I see it... please work with Geoff on producing suitable stats.

I said it was my personal opinion.

That's why UK dan resets aren't the correct answer.

GoR provides a basis for comparison of people across Europe using a rating number. I'm not sure it should profess much else, since each country still uses a different system for assessing the grade of a person (some of which are mainly independent of the GoR rating). If everybody used the same system then I think we MIGHT have the same grade and rating have the same meaning across Europe...

OTOH I don't know how that might fit in with Oriental grades.

here:




So maybe putting in anchors is a step in the right direction. There probably are active players who are at a stable level for 5 years or more. picking out a few in each country, determining their "real"rank, and reset and fix their rating to this desired rank.
KGS uses anchors as well, but I don't know how they implemented this exactly. klick

_________________
Laurens
Go school: freegame's Teaching School
Author of: The Next Move a book for 15-3kyu players.

I'm not particularly interested in what any one country wants to do to award dan or kyu diploma's. If somebody can think of a way of persuading all countries to adopt the same approach I might listen. What I am sceptical about is the merit of altering a rank to produce a less accurate rating. Perhaps in the context of European non-allignment it won't matter but I still don't like to see it.

If it were possible it would be nice to add some more featurs to GoR, like allowing second iterations of the algorithm for huge performance improvements at certain ranks. However, this would require altering the rating algorithm.
PS: These rating threads seem to be popular

from what i've heard, there is some movement towards assigning initial MMS at tournaments (at least at big ones) according to player's rating, not considering his declared rank (GoR 2050-2150 one MM group, 2150-2250 another, etc.).

if it is so, it would severely weaken importance of ranks in favour of importance of rating, make tournaments more fair and effectively solve problems of unevenly strong ranks in various countries.

still i can't say i really like it, because with such a change, ranks would become effectively meaningless and i like the nice, old-fashioned ranks more than only modern, cold, numerical rating

_________________
Spilling gasoline feels good.

I might be wrong, but probably not.

Yes, several tournaments implement such a policy.



No and no. It is not more fair, and it does not solve uneven ranks.

The EGF rating system is not inherently more reliable than the old ranking system. In fact, research in the Netherlands showed that for kyu ranks, the old ranking system was more reliable than the EGF rating system, by a very large margin.

Furthermore, GoR as a system tends to drift over time. It tries to compensate for this by the epsilon parameter in the formula, but the correct value depends on the number of tournaments that go into the system. And since that number of tournaments is different per country and per player, the epsilon currently is at best an average. Countries with fewer tournaments entered per year will drift differently from countries with more tournaments entered, unless there is a very strong tendency by players to play abroad. Which there isn't.

You misunderstand this slightly. Dan certificates are awarded in the UK, and once awarded are never revoked. However, there is an expectation that if your current strength is nearest to, for example 1k, then even if you have previously received a 2d certificate, you are still expected to enter as 1k. It is possible the wording in the policy is not clear enough.

This "mass shift" that would pull people's nearest grade down is not against any existing UK policy.

Perhaps this will be the case for Barcelona (19-20 Feb). One can navigate from http://www.aego.biz/ to find the information - and it says



There is also the option for simultaneous games with Lee Young-sin 8p on the Friday evening.

Maybe this should be another topic, but payment in advance, often with a Paypal option, seems frequently to be becoming a condition for receiving early reduced entrance fees.

This is interesting.
In what direction would the drifting go for a player if
S/he plays a lot of tournaments vs very few (and what is the average number?)
Her/his country organizes many tournaments vs very few (and again, what is the average?)

And lastly, couldn't the formula be changed to compensate for this?

/Mats

_________________
mohsart - games & books
http://spel.mohsart.se/