I forgot where i read it but i remember that there was some discussion to have seperate tournaments for the different world tournaments (Toyota, ING, Pandanet,WMSG, i don't have the complete list) where europeans are invited to.

This particulary can result in that the same player, can be the european representative at many tournaments and other top players will not be sent to world tournaments.

I myself am against this idea as long as europeans are not really contenders for the world tournament title.

But i would like to open the discussion about it.

_________________
Promotor and Librarian of Sensei's Library

Okay, so let us collect some facts first.

• Which are those "world tournaments"?
• What are the respective details?
• What kind of opponents does the selected player get to play?
• What is the current selection process?

_________________
A good system naturally covers all corner cases without further effort.

He is talking about events like the Chunlan cup

Currently the winner of whatever tournament gets picked that year (Masters/Ing/European) gets to go to one of those events.

O some enlightenment.

The proposel is in from the Pussian proposals for the AGM 2010
(see webpage http://www.eurogofed.org/egf/index.htm under AGM 2010)



As said before i am against it for as long as the European representative does not have a serious change to win the tournament where he is as european representative.

_________________
Promotor and Librarian of Sensei's Library

Yes, it was my own proposal. Russian Go Federation agreed with it.

We had European Masters cup before for deciding World Cup representatives. It's cancelled now.

This year we had Fujitsu preliminary online (Taranu, Pop, Mero, me). I won it and played in Fujitsu cup in Japan.
I also wanted to play in Chunlan (invitation arrived few months after Fujitsu preliminary) and I suggested to play another preliminary online

EGF decided to send Csaba Mero to Chunlan - without any preliminary. Just because he was 2nd in Fujitsu.

The topic must not sound like "Winner takes all". I just want to have fair preliminaries for World Cups. That's all

It's very strange that EGF decides the candidate now. We need some rules here!

_________________
Go lessons: http://breakfast.go4go.net/
Russian Go news: https://vk.com/go_secrets/

What was your result in the Fujitsu cup?

_________________
Promotor and Librarian of Sensei's Library

I recall earlier years where the EC placements were used.



Right! I will see whether something can be done here to replace semi-arbitrarily wild cilds by fair qualification rules. I can only make suggestions though; the AGM, else the EGF Committee decides. To get fair qualification rules, we need such rules indeed.



Good questions.

* Which are those "world tournaments"?
Fujitsu every year, Ing cup (1X4 years), Chunlan cup (every second year)
* What are the respective details?
Free airticket, 5star hotel, Near 2000 usd as first round fee or 5000 if the person loses in round 2 (best result in European Go history)
* What kind of opponents does the selected player get to play?
From 6-7d (Fernando Aguilar) to top 9-dans like Lee Sedol
* What is the current selection process?[/quote]
No idea. The old system with European Masters is broken and new one is not introduced yet

It's a question number 16f in EGF meeting agenda: http://eurogofed.org/egf/agm2010agenda.txt

_________________
Go lessons: http://breakfast.go4go.net/
Russian Go news: https://vk.com/go_secrets/

Right now the only proposal is from Russia. Too late to propose other systems, so can only vote for Russian system or against any systems (EGF will decide players they want to send, like they did in 2009)

Will repeat Russian proposal:

European champion has the first choice of World cup. For any new events top 4 players by European rating play the online preliminary (on KGS). European champion has the right to take part and win his second or even 3rd event straight. Of course, if the European champion is inside best 4 in EGF rating list. If he is weaker, he has only first event

_________________
Go lessons: http://breakfast.go4go.net/
Russian Go news: https://vk.com/go_secrets/

I have not reflected the real contents of the proposal in comparison to alternative proposals one might create. I would like to remark on the ratings though:

Instead of using the rating at a particular moment, I would like to see used an average rating and a minimal number of rated games during the last, say, 12 months.

Yes, I agree. Minimal number of rated games is good idea. But how many games? 10 rated games per year is probably enough. What do you think?

What is average rating and how to calculate it? It's more complicated, I guess.

_________________
Go lessons: http://breakfast.go4go.net/
Russian Go news: https://vk.com/go_secrets/

I think the most modest useful minimum is 25. This about means attendance of, e.g., the previous EC, a national championship plus 2 weekend tournaments. Everybody interested in being seeded should be able to play that few rated games!



I suppose the Rating Commission would be able to produce figures easily (also for EC qualification ratings). Therefore a good average can be used: Consider each day's rating of a player, add it for all days and divide by the number of days in those 12 months.

(For each tournament, one also needs a deadline when to form the average ratings. Maybe 4 or 6 months in advance of a world tournament (or 1 day before the EC) so that the qualified player can confirm and buy his flight tickets in time.)

In Korea they have similar preliminary systems for Fujitsu, Chunlan and other events. Hankuk Kiwon invites best 4 or 8 pros by rating and they decide the candidates in special mini tournament with 2-3 games.
They don't use such thing as average rating.
It's important to decide who is in best shape right now.

25 games - too much, I think. 10-15 will be enough

_________________
Go lessons: http://breakfast.go4go.net/
Russian Go news: https://vk.com/go_secrets/

How many months before a tournament would you think is the best? Recall the time necessary to clarify whether some eligible player cancels and for buying tickets.

Is average rating of the 3 or 6 preceding months (before the rating deadline) closer to determining "best shape right now"? I prefer average so that a player cannot artifically push his rating upwards in the last moment by playing trivial tournaments or do you think that is impossible for the top, say, 4 Europeans under the current rating system?

It should not be possible to change one's rating upwards intentionally. That follows from the ELO mechanism. Statistically, the expected rating change from a single tournament with correctly rated opponents is 0, no matter what the players' absolute or relative ratings are. In order to fabricate an expected gain, a player would need to identify a population of overrated players (more overrated than himself) and attend a tournament in which he would be likely to get them as opponents.

I think that the discussion around how to use the rating is an indication for the current system's shortcomings, by the way. No one believes that the rating is a reliable indicator of strength.

_________________
A good system naturally covers all corner cases without further effort.

Below the top, it is straightforward to increase one's rating: Play a "local" tournament where one is the strongest player.

I'm not sure this is strictly true in the long term, unless the ratings method is poor. I don't think it is a reliable strategy in the EGF system.

What matters is that it works shortly before a relevant tournament, for which one needs rating points.

The current standing:
Dinerstein -2742
Taranu -2739
Guo Juan -2733
Shikshin Ilya -2706
Shikshina Svetlana -2695
Pop - 2676

Svetlana has no reason to fight with her brother and Pop is 30 points behind of Ilya.

Do you know how to get 30 points in local Romanian event?
He has to beat Catalin 5 games straight, for example. I don't think it's possible

_________________
Go lessons: http://breakfast.go4go.net/
Russian Go news: https://vk.com/go_secrets/

No, it does not. Even if you play 5 games against players you have a 90% win probability against, your expected rating change is exactly 0. The probability that the one loss per 10 games you would statistically suffer happens in just these 5 games (colloquially speaking) is 50%, and the rating system works on the principle that your loss in the 4:1 case is exactly as big as the gain in the 5:0 case.

Of course, this works on the assumption that none of the players is misrated (or even, that the errors cancel out). However, do you think that someone can not only identify overrated players, but also find a tournament where he would be paired against these players sufficiently often?

_________________
A good system naturally covers all corner cases without further effort.