Entry Grade

[snorri]

The change is done because a system exists. This quietly publicised system basically forbids entry at a grade above 2dan - unless you happen to be foreign or a foreigner living in the UK.

Than the tournament derector who signed up the guy as a 3d missed this rule. had the TD (or player) known about it, he would be entered as a 2d, and there would have been no problem at all

So the rule in question is this one?

For dan players BGA grade is also server grade - 2, but the highest entry grade is 2d unless agreed ahead of time with the BGA.

I assume this rule was introduced to solve some sort of problem. What was the problem? Was the problem with what was being submitted to the EGF or was the problem related to something happening in internal BGA tournaments / ratings?

[Javaness]

The EGD is a database, and as such doesn't have its own rating system, the one it implements is called GoR (Go Rating). By choosing to reduce the rank of the player who jointly won the tournament, the BGA chose to collectively reduce the rank of everyone he played ( a whole 3 people!) and indirectly everyone they played. So the decision can be seen as a positive move to correct the inflated ratings of BGA
players.

It is somewhat incorrect to say that GoR is an inflationary or a deflationary system, because it has no prescribed implementation.

I wouldn't normally comment in such a specific vein as below, but the original posting and subsequent discussion has been sufficiently detailed to have identified a specific case which was discussed by the BGA Council recently. The facts are:

- the player referred to was known to the EGD rating system, but at a lowish kyu rating due to there being no known games for many years
- in the EGD rating system if the entry grade is two or more grades above the previously entered grade then it does a reset and takes the entered grade as the base point for calculating future ratings, so what the entry grade is makes a significant difference to the exit rating (and other people's ratings)
- the player entered the Maidenhead tournament at an entry grade of 3 dan which the Tournament Director accepted
- the exit rating for this player would have exceeded 2300 only if the entry grade was 3 dan or above
- the BGA Rating Officer noticed this and queried the basis for the entry grade, referring the issue to the BGA Council
- the BGA has a policy statement, at http://www.britgo.org/policy/policies7 relating to rating resets.
This was approved some considerable while ago, to encourage kyu players who are rapidly improving to do so, but to restrict resets at dan levels because of the possible implications for the rating system as a whole
- the maximum dan grade for entry to a tournament for effectively an unknown person in this policy was set to 2 dan (and subject to pre-approval), except for people with grades assigned by foreign Go associations
- the BGA Council considered this issue and has a) submitted this tournament to the EGD using an entry grade of 2 dan and b) written appropriately to the person concerned, who will be entitled to enter his next tournament at 3 dan because of these results

----------------------------------------------------------------------------------

There is a continuing debate about the EGD rating system and whether it is deflationary or not - further statistical analysis is definitely required.

However, it should be noted that UK grades, as per our published list, are pretty well correlated to the European average (UK 2031.7 v 2039.5 for an average shodan). The EGD rating system uses at its basis a theoretical rating for an average shodan of 2100, which is only achieved by Slovakia and the Czech Republic (who originated the system).

It should also be noted that strengths published in the UK rating list are based on the person's rating and the European average correlations, as described in the FAQ, not on the theoretical basis for the EGD system.

Regards
J Diamond
President, British Go Association

[Javaness]

My own theory: Established players didn't like people jumping up above them in the ratings without having ever gone to any tournaments. As I intimated earlier, resubmission is probably a better way to deal with any such malarky. To my knowledge this is all very uncommon.

The change is done because a system exists. This quietly publicised system basically forbids entry at a grade above 2dan - unless you happen to be foreign or a foreigner living in the UK.

Than the tournament derector who signed up the guy as a 3d missed this rule. had the TD (or player) known about it, he would be entered as a 2d, and there would have been no problem at all

So the rule in question is this one?

For dan players BGA grade is also server grade - 2, but the highest entry grade is 2d unless agreed ahead of time with the BGA.

I assume this rule was introduced to solve some sort of problem. What was the problem? Was the problem with what was being submitted to the EGF or was the problem related to something happening in internal BGA tournaments / ratings?

[mumps]

The EGD is a database, and as such doesn't have its own rating system, the one it implements is called GoR (Go Rating). By choosing to reduce the rank of the player who jointly won the tournament, the BGA chose to collectively reduce the rank of everyone he played ( a whole 3 people!) and indirectly everyone they played. So the decision can be seen as a positive move to correct the inflated ratings of BGA
players.

We didn't REDUCE the rank of that player, just chose an entry grade that we felt to be more appropriate given his lack of prior history. His actual results could have been achieved by a 2 dan or a 3 dan (slightly more probable for a 3 dan). Given the current mapping from rating numbers to UK strength entering him at 2 dan was effectively at a nominal 2.7 dan for the rating system purposes.

It is somewhat incorrect to say that GoR is an inflationary or a deflationary system, because it has no prescribed implementation.

But there is an actual implementation, which includes at least one fudge factor (ε below) which has changed during the lifetime of the system:

"The rating system is derived from ELO rating system used by International Chess Federation (FIDE). It is based on the idea that one can define a probability of winning a game (so called winning expectancy SE) depending on the difference of opponents ratings D=RB-RA. For the player with lower rating (let us call him "player A") the quantity is given by

(1) SE(A) = 1 / [eD/a + 1] - ε/2

The winning expectancy of his higher (or evenly) ranked opponent ("player B") is obtained from the equation

(2) SE(A) + SE(B) = 1 - ε

If ε=0, Eq.(2) simply states that the sum of both winning expectancies should be normalized to one. However, such setting suffers from long term deflation as the new improving players take points from already established players. This is countered by various instruments like an existence of rating bottom, winning expectancy setting, rating resets in some specific cases and finally by introduction of a small parameter ε > 0.
At the moment, we use ε=0.016, a value fitted to balance rating variations in dan region. Although such small value has negligible effect on variation of player's rating at one tournament, the parameter ε allows to tune the long term system behaviour in a desired way."

[Javaness]

GoR relies on player ranks, which are set in different ways in different countries for both existing and new players. This is quite a key difference between GoR and FIDE ratings, which do not have any rank dependency. Aalthough obviously there is one formula present in GoR, the modus operandi varies significantly across Europe. In some countries, if a player is promoted at the end of a tournament, they will be entered at that new rank when the results are submitted. This will then quite often give better ratings for all concerned. These differences in implementation, I warrant, are why you will find differences in strength across Europe.

We didn't REDUCE the rank of that player, just chose an entry grade that we felt to be more appropriate given his lack of prior history. His actual results could have been achieved by a 2 dan or a 3 dan (slightly more probable for a 3 dan). Given the current mapping from rating numbers to UK strength entering him at 2 dan was effectively at a nominal 2.7 dan for the rating system purposes.

I don't think that writing reduce in capslock helped that argument much.

[topazg]

We didn't REDUCE the rank of that player, just chose an entry grade that we felt to be more appropriate given his lack of prior history. His actual results could have been achieved by a 2 dan or a 3 dan (slightly more probable for a 3 dan). Given the current mapping from rating numbers to UK strength entering him at 2 dan was effectively at a nominal 2.7 dan for the rating system purposes.

That's an understatement. His rating performance, in BGA rank equivalents, was around 3.8 dan.

[HermanHiddema]

We didn't REDUCE the rank of that player, just chose an entry grade that we felt to be more appropriate given his lack of prior history. His actual results could have been achieved by a 2 dan or a 3 dan (slightly more probable for a 3 dan). Given the current mapping from rating numbers to UK strength entering him at 2 dan was effectively at a nominal 2.7 dan for the rating system purposes.

That's an understatement. His rating performance, in BGA rank equivalents, was around 3.8 dan.

I think that that number is meaningless without an error bar. I think it is reasonable to say, for example, that the player's performance is somewhere between 2000 and 2500 in GoR terms

But really, what is the issue here?

It seems that a reasonable summary of the situation is:

There are regulations about player ranks. The tournament director did not follow the regulations. The BGA corrected the result to be in line with the regulations. The tournament result makes no compelling case against the BGA's decision.

[Javaness]

Obviously, according to national policy it's ok. I'd also bet that the player in question doesn't care 1. about his BGA rank, or 2. about his EGF rating. My query (still) is, if you have a player with a reset rank, which looks reliable, should you rush to change it?

I think that that number is meaningless without an error bar. I think it is reasonable to say, for example, that the player's performance is somewhere between 2000 and 2500 in GoR terms

But really, what is the issue here?

It seems that a reasonable summary of the situation is:

There are regulations about player ranks. The tournament director did not follow the regulations. The BGA corrected the result to be in line with the regulations. The tournament result makes no compelling case against the BGA's decision.

[HermanHiddema]

Obviously, according to national policy it's ok. I'd also bet that the player in question doesn't care 1. about his BGA rank, or 2. about his EGF rating. My query (still) is, if you have a player with a reset rank, which looks reliable, should you rush to change it?

If the rank looks reliable, then you should consider not changing it. But for a rank to be reliable, you definitely need more than three games.

[topazg]

Obviously, according to national policy it's ok. I'd also bet that the player in question doesn't care 1. about his BGA rank, or 2. about his EGF rating. My query (still) is, if you have a player with a reset rank, which looks reliable, should you rush to change it?

If the rank looks reliable, then you should consider not changing it. But for a rank to be reliable, you definitely need more than three games.

That's a catch 22

Do you mean you should not consider changing it if the rank looks reliable, or that you should consider changing it if the rank looks unreliable? They are two very different things...

[richardamullens]

I think it is all very simple.

We want our highly graded players to have grades that are manifestly justified.

We prefer to err on the side of caution. If a player really is stronger, then that will become apparent in later tournaments and also give them an incentive to enter.

Personally my feeling is that UK grades overestimate our ability. The evidence for that is the results of our play against foreign opposition. Therefore if improving players "suck" some points from established players it is no bad thing. If established players want to maintain their GoR they'll have to work to do so. Probably in Go one has to run just to stand still.

I am very happy with the European rating system and the existing UK policy. It is good, in my opinion, that it has been firmed up and that steps are being taken to ensure that people enter at grades which have a basis in reality rather than wishful thinking. The result I feel is that players get better games and tournaments become more fun.

[tapir]

I think it is all very simple. We want our highly graded players to have grades that are manifestly justified. Therefore if improving players "suck" some points from established players it is no bad thing.

All this sounded somewhat different in a neighbouring thread... I mean why does the BGA keeps track of the non-improving 1d rating at all, if it considers sucking away some points from established players as __desirable__?

[topazg]

I would actually be in favour of the BGA just saying "2100 GoR is 1 dan, as it is supposed to be", and if players aren't keeping above 2100, their rank is going down, rather than adjusting the GoR for 1d to be ~ the mean of entering 1d players.

Sure, we'd all lose two thirds of a stone in rank, but it would fix over half of the rating reset controversy (at the moment, the reset to 2d actually = reset to 2.7d) and it would feel like the tail is no longer wagging the dog. We'd also get to see how deflationary or not the rating system actually is in practice.

I personally and happily would volunteer to lose 0.65 stones for this to happen across the board.

[amnal]

I think it is all very simple. We want our highly graded players to have grades that are manifestly justified. Therefore if improving players "suck" some points from established players it is no bad thing.

All this sounded somewhat different in a neighbouring thread... I mean why does the BGA keeps track of the non-improving 1d rating at all, if it considers sucking away some points from established players as __desirable__?

There is a strong (but subjective and arguable) case for it being '__desirable__'. As topazg says, we have a significant offset from the GoR set system, such that our players who consider themselves 1 dan are close to a stone weaker (give or take errors etc. etc.) than those in a country where those who call themselves shodan cluster around 2100 (as is supposed to happen).

I would also be all for resetting British ratings to '2100 GoR is 1 dan'. Getting to shodan is no good if it's 'congratulations, you're the weakest dan player in Europe'

[topazg]

I would also be all for resetting British ratings to '2100 GoR is 1 dan'. Getting to shodan is no good if it's 'congratulations, you're the weakest dan player in Europe'

In our defense, that makes us "about in the middle", GoR wise

Not that that answers the "whose GoR 2100 is stronger than whose" either...