there is a difference between "strong" rankings and a broken system. by disallowing rank resets (and this is as i understand the main feature of the czech implementation), the improving players play each other and compete for the same points. leading to deflation (over all) or outdated ratings (for some players). i guess there are a lot of czech players whose GoR looks like this one http://www.europeangodatabase.eu/EGD/Pl ... y=10201510 or that one http://www.europeangodatabase.eu/EGD/Pl ... y=10233509 - do they really lose their strength despite continuously playing? (even ales cieply himself has trouble in maintaining his rank.)

well, there is deflation, but the system should be able to counter it with proper parameters, i think. the problem arises from that the system is implemented differently in different countries and parameters are set to the average...

rank resets are not really disallowed in Czechia, i myself jumped from 17k directly to 10k without any complaint, they are just not usual. how do you implement the system in your country? example: till last tournament i was 4k. i am young, improving player (i hope), someone could say i have already strength for being a 2k. so should i reset my rank and enter the tournament as 2k instead of 4k? i don't think so - i will be 2k when my GoR says i am 2k, what would be otherwise its purpose?

for larger gaps between strength and rank, resets are natural, but there are not many players improving by 3 or more stones between following tournaments

_________________
Spilling gasoline feels good.

I might be wrong, but probably not.

Firstly I don't think the above statement is correct.
Secondly I think that you and a couple of other commenters in this thread are underestimating the need for statistical significance a lot.

What you should do is to calculate "performancee ranks" according to the suggestion in the 2007 Nordic Go Journal article, and compare them to actual ranks. Or you should look at the winning statistics between equally ranked players from different countries in a systematic way (it is not enough to look at some single tournament, except perhaps the European go congress. There are also possible selection problems as discussed in the same article and one needs to be confident that they are not too big). For sure some individual players are "overranked" or "underranked", in every country, but this does not allow you to draw any conclusions about an entire country. For the record I won't believe any statements in the style of "my feeling is that " this country is over ranked compared to that country, they are not objective and not based on sufficient amounts of data.

The country-wise comparison of GoR can be carried over into rank comparison using the EGD table that was referred to early in this thread. Judging from such an argument I don't believe that there are differences between average ranks in different european countries that are as big as several stones, although 1-2 stones average differences might occur. Differences between the countries with many players are smaller I think. Clearly I'm talking about ranks above 10k, or above 5k, I don't think there is any point at all in discussing the accuracy of double digit kyu ranks.

I would advise people who are interested in matters like this to use the EGF rating database, which contains the relevant data, from every year since 1996. The EGD source code should be open for this purpose. The best significance is obtained of course if statistics are made from the results game by game. Due to ignorance in database handling I never made any serious attempt with that myself, I mostly just tried to get along by using the average tournament-wise GoR increments of players. I'm sure many of you people who read and write here can easily do better than that, if you just care to spend a little time on it.

cheers,
Henric

So you mean to say that you're sure that Swedish players are (in general) underranked compared to average Europeans?

/Mats

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

i played my first tournament as 3k, after that there was never a need for a rapid rank reset.

if someone says i might be 2 stones stronger, but i am not sure of it then i wouldn't. but if you hit 1 dan at kgs or igs i would strongly advise you to reset your rank to 2k or 1k EGF.

That's not what I'm saying.
It is probably true, but I already said that I don't think the differences are big anyway.

I was saying that your list is wrong, in my opinion.

Here is an example of an argument based on reasonable statistics
(http://goforbundet.se/ng/200901.pdf pp. 12-13):

In EGC 2008, for the players with GoR>2000, players from
SE gained on average 4 GoR points per game (based on 171 games)
CZ gained 2 Gor points per game (68 games)
PL gained 1 GoR point per game (70 games)
RU lost 1 GoR point per game (109 games).

Approximately one can argue that (Fig. 2, p. 12) this means that
SE players were underrated by 80 GoR points wrt average,
CZ players underrated by 40,
PL players underrated by 20
RU players overrated by 20.

As far as GoR is concerned and in this rating interval, your ordered list is wrong.
What about ranks then?
We can correct for the rank-rating differences listed in the EGD table that was mentioned,
which gives AVERAGE differences (in GoR points, http://www.europeangodatabase.eu/EGD/Stats_Country.php):
SE:-73
CZ: +25
PL: -24
RU: -5
These numbers are not for the rating segment >2000, I believe that those numbers
must be lower at the high end of the rating list, so they overestimate the correction.

Still, the result would be expected rank differences
SE: 80-73=+7 wrt european average (so the difference is less than 1/10 stone).
CZ: 40+25=65 wrt european average (i.e. a bit more than half a stone)
PL: 20-24 = -4 wrt average,
RU: -20-5 = -25 wrt average.

So the ordered rank strength list becomes:
CZ, SE, PL, RU.

It should be clear from this discussion that the uncertainties are pretty big,
even though hundreds of games are taken into account. But IMO it is enough to conclude that
there is no empirical ground for your ordered list.
Another conclusion that should be obvious is that the rank "strengths" are not
VERY different, the differences are SMALL.

In the rating interval 1000-2000 FR, PL, CZ all gain more GoR points per game than SE, but FI
gains less.

It is easy to improve on these estimates of rank strengths, using the results
in individual games more directly, but I don't believe that the results will be
MUCH different from the above.

It is a worth while effort to monitor the differences, by studying winning
percentages:
* If the ranks and ratings are drifting apart, we could have a big problem.
* It is unfair if big discrepancies develop. They could have an effect both
on the results lists in tournaments, and on admission to top groups,
sponsored events etc.
* It is interesting in itself to understand how well the rating- and
ranking systems perform.

But I think it's wise to investigate the matter empirically, rather than
guessing.

cheers,
Henric

Can you explain little bit more, how they play for money in the club? Sounds interesting..

Probably each player wagers the same amount, and winner takes the pot.

_________________
Offending ad removed

Yes but, i would like to know more, for example are there any time limit or any other restrictions...If you play for money, you can make any dirty cheats then..

IIRC, you can wager on points margin (per point) or just on result, and they each have their distinct name.

magicwand, do per point wagers have an upper limit? What about resignations?

this is a bit off topic, but i believe Magicwand talked about bang neki. in principle, you can gamble by betting just on the result of the game, or - more famous - you can bet also on the point difference, where resignation equals 100 point loss. there is also Hahn tournament system derived from this idea. (it is used for example at the Berliner tournament Go to Innovation.)

_________________
Spilling gasoline feels good.

I might be wrong, but probably not.