Report: European Go Congress 2011

RobertJasiek · #1

Report: European Go Congress 2011

Venue and Date

The event was from July 23 to August 6. Advertised as a congress
amidst beautiful wine fields, the actual venue was a heap of a
university's awful creton blocks scattered over a by far too large area with great walking distances - pretty much the last place where you would want to make holiday. Rooms and furniture were acceptable, the too few rest rooms were not. Students type accommodation was within walking distance, partly a few tram stations apart. Those taking meals at the university also had to go there by tram. The alternative was a nearby supermarket until it was not nearby any longer: On Sundays and during the second week the rear gates of the campus were closed and people either had to climb over the fence or walk an extra kilometer. Not nice! Players arriving a day earlier had quiet some trouble to access their accommodation when keys were handed out elsewhere and did not work in the described building at 3 am and
you had to guess and then find building G when the used "alphabet" order was A G B C D E F H I J etc...

Organisation, Supergroup and French Kyus

For French standards with an expected 4 hours delay of round 1, the congress organisation was almost reasonable with exceptions. Rounds tended to start almost on time, although remaining clocks were started with ca. 18 minutes delay. Pairings were published only minutes before a round, except that round 8 pairings were available the evening before. Referees were easier to find than substitute clocks. According
to experience by weaker players, the dropping system was not held properly; I experienced quite some problems with it during the Rapid.

Generally the organisers were very willing to cooperate with the EGF's Tournament Supervisors and made our job rather easy. They even asked for my advice about how to distribute EUR 10.000 prize money to the top 12 players of the main tournament and KO; the French would rather have preferred the EGF to specify the prizes. With two major exceptions: 1) The organisers were unable to copy their player database file. Therefore the supergroup formation had to be done during night until 3 am (my second night in succession with too little
sleep; luckily I got an easy opponent in round 1). As a side effect, we were so tired to forget adjusting the fake 6d rank of the infamous Bahadur Tahirbayov, AZ, who got - for his standards - an exceptionally good result 2:8 by beating two 1 kyus. The supergroup of 30 players had an incredibly high bottom rating of 2525, which is about 100 above all earlier years. In other words, many strong Europeans participated.
Even Vladimir Danek's rating was too low to qualify for the
supergroup. 2) The organisers refused to apply the rule that
self-declared ranks are used and instead ratings were used to derive ranks. They got away with that because we supervisors were not supported by Martin Stiassny, EGF president, in this respect. While this made the French organisers happy, then same cannot be said about all French players. Many of them have such a low rank compared to most other European countries that they don't know what their French rank is worth internationally. Typical guesses were that French 10 - 15k is ca. European 5k, French 3 - 5k is ca. European 1k to 2d. E.g., of 20
of 22 (or 91%) players with kyu rank and 8 or 9 wins were French. It cannot be shown more clearly that using the EGF rating system for kyu ranks does not work at all, although some of those French kyus have told me that they play a lot of tournaments regularly.

The organisers and some fellow players were of very great help with medical problems and related translation.

European Championship

The new tournament applied: After round 7 of the McMahon, the top 16 Europeans are compared, some might have to play a relegation game and the top 8 enter the KO and possibly place 3 game. So after round 7, these were the standings:

Code:

#   Name      Rank   MMS   SOS   Rating

   Taranu    Catalin 7D    40    270    2730
   Silt    Ondrej    6D    39    272    2607
   Shikshin Ilja    7D    39    271    2744
   Mitic    Dusan    6D    39    271    2567
   Debarre Thomas    6D    39    269    2580
   Din. Alexandr    7D    39    268    2721
   Kacha.    Artem    7D    39    268    2663
-------------------------------------------------------
   Burzo    Cornel    6D    39    267    2624
   Surin    Dmitriy   5D    39    266    2536
-------------------------------------------------------
   Pop Cristian    7D    38    267    2667
   Balogh    Pal    6D    38    267    2629
   Simara    Jan    5D    38    267    2548
   Mero    Csaba    6D    38    266    2630
   Sankin    Timur    5D    38    265    2492
   Bohat. Dmytro    6D    38    264    2551
   Jabarin   Ali    6D    38    264    2550

MMS 39 means 5 wins. The first seven players qualified automatically because they have greater MMS than the last 7, who continue the McMahon. Only one relegation game between Burzo and Surin had to be played; Burzo won it by 0.5 points! This pair was a repetition, which could not be avoided. During the KO and place 3 game, no repetition occurred. Including this year, the expected average number of relegation game players drops below 5.3. It is the first studied year that relegation players had 5 wins; in earlier years, they always would have had 4 wins in the first 7 rounds. Burzo would not have wanted to use the second Wednesday as a free day; he would have
participated in a side tournament instead; Surin's weak English foiled my attempt to interview him about having to play the relegation game.

The quarter finals were bubble sorting by rank: The 7d players won. The top four Europeans turned out to be:

1 Shikshin Ilja 7D
2 Taranu Catalin 7D
3 Kachanovskyi Artem 7D
4 Dinershteyn Alexandr 7D

You need to ignore them for themselves in the main tournament result table
http://egc2011.eu/index.php/en/results- ... main-round
It might still be wrong; having left early, I do not know yet if there might have been reason to state 0.5 fractions in MMS and SOS; usually the rules round down. I count 6 ghosts ("players" missing all rounds) in the table, so the there might have been 760 main tournament players; in 1997 Marseille, there had been 51 ghosts and 539 genuine players. 760 and among them lots of high dans is much, especially when considering the doubtful venue.

The rules of play were AGA / French style: Simplified Ing Rules with pass stones and White passes last. Their announcement and explanation of strategy on the walls and in the congress tournament information booklet was very good (using also my comments texts). Nevertheless there will always be players who read nothing, then start wondering why the hell the last white pass stone changes strategy, as if they had never played under area scoring rules before (they had many times!), confuse the counting means "last pass stone" with the scoring definition and then and blame the rules for being more difficult than
Japanese rules instead of recognizing their laziness to read the rules and strategy information in time. There were also other players who read and understood the rules and were very happy with them. For me they were a joy, of course. They were also used in side tournaments and there free handicap placement without komi recompensation was used.

Side Tournaments and Other Events

Here are some side tournament winners:
http://egc2011.eu/index.php/en/results- ... oi-annexes
http://www.eurogofed.org/results/congress/egc2011.htm
The author entered 13x13 and 9x9 as 5 dan (refusing the stupid French style rating rank) and made it to the quarter-finals in each. Watching the later KO rounds was great fun, as always. 13x13 was announced as group qualifications but Swiss was used. Sort of. Some Gotha pairings paired players with 2 wins difference! Lighting qualifications were boringly slow: a "group system" without group table sheets. In between
every two games, result forms were evaluated and only then the next game could start. 20 minutes playing interrupted by 20 minutes waiting. Gulp! I did not qualify, being thrown out by two of those French kyus; it did not help that the handicap was reduced by 4 stones.

This year's referee workshop had a record number of 30 participants! And a record of almost 6 hours lecture for rules of play because also some WMSG Rules had to be prepared.

Psychee · #2

RobertJasiek wrote:

E.g., of 20 of 22 (or 91%) players with kyu rank and 8 or 9 wins were French.

How many GoRs will a kyu player gain on average by winning 9 games out of 10?

gaius · #3

So what happens if you once played a tournament for fun as a 20-kyu, and a year later you're 5 kyu and want to play EGC? Are they really going to force you to play 15 stones weaker than your actual strength?

If not, where do they draw the line?

Forcing kyu players to play according to rating is stupid indeed. In the EGC in Frascati, I entered as a 10 kyu, but I discovered quickly that it was probably 4-5 stones too weak. Fortunately they allowed me to reset my rank to 6 kyu after three rounds, and I played 5 out 7 in the rest of the tournament. Thát is the right spirit, if you ask me!

HermanHiddema · #4

When some people proposed to use EGF ratings to regulate Dutch kyu ranks, the Dutch Go Association had some research done into the viability of that system. It turned out that EGF ratings are almost completely worthless as compared to self-chosen kyu ranks. A player's self-chosen rank was highly predictive for their performance, while their rating had almost no predictive power at all.

karaklis · #5

Psychee wrote:

How many GoRs will a kyu player gain on average by winning 9 games out of 10?

Depends on the rank. Stronger ranks do not gain as much as lower ranks. As 8k you would gain about 220 points, provided your opponents have exactly the same ratings as yours.

HermanHiddema wrote:

When some people proposed to use EGF ratings to regulate Dutch kyu ranks, the Dutch Go Association had some research done into the viability of that system. It turned out that EGF ratings are almost completely worthless as compared to self-chosen kyu ranks. A player's self-chosen rank was highly predictive for their performance, while their rating had almost no predictive power at all.

Sounds interesting. Could you post some details about this?

HermanHiddema · #6

karaklis wrote:

HermanHiddema wrote:

When some people proposed to use EGF ratings to regulate Dutch kyu ranks, the Dutch Go Association had some research done into the viability of that system. It turned out that EGF ratings are almost completely worthless as compared to self-chosen kyu ranks. A player's self-chosen rank was highly predictive for their performance, while their rating had almost no predictive power at all.

Sounds interesting. Could you post some details about this?

The following research was originally done by Andre Engels (PhD in Mathemetics and 3 dan go player) in 2007, at the request of the Dutch Go Association. I have made a quick, rough translation here. Because it is very long, due to tables of percentages, I've hidden it.

Laman · #7

this is not about EGC anymore, but while the Dutch statistics are impressive, i can say that in Czechia we base ranks almost exclusively on rating and it can also work pretty well (ie. to most players' satisfaction). i am not an expert, but don't most of these inconsistencies arise because of different ranking (and therefore rating) practise in different countries?

RobertJasiek · #8

Yes, and that is also the question of Czech (and presumably also Polish) ratings for kyus: Do they hold as good descriptions when compared to all other European countries?!

tapir · #9

add to the statistic the participants who took part only for one week. there was quite a number of french participants who went 5:0 in either 1st or 2nd week. instead of being proud of this result this should be taken as evidence that french practice is incompatible with the rest of europe. the main purpose of all the fancy rating systems is to produce even pairings after all. this seems impossible to achieve if rating resets are arbitrarily discouraged / forbidden in some countries. the high amount of 5:0 / 9:1 results in a mcmahon tournament is thus not a national achievement (toughest ranks) but a system failure. mismatched games with such long time limits are just frustrating for both sides. there might be a problem with inflation in some countries in the 1 dan region, but it isn't effectively countered by artificially deflating the whole kyu range somewhere else.

the dutch study can't be quoted enough if you ask me. it can help in perceiving ranks / ratings as a tool to achieve equal pairings instead of a measurement of achievement.

ps if you don't allow self ranking in czechia you can't study it. satisfaction of the playing population is not a criterion. but a good measure could be the willingness of the higher rated player to play tournaments with full handicap. i predict it would be considerably lower in the countries with artificially deflated kyu ranks.

Laman · **#10**

RobertJasiek wrote:

Yes, and that is also the question of Czech (and presumably also Polish) ratings for kyus: Do they hold as good descriptions when compared to all other European countries?!

you have a point, our (Czech, Polish, French, ...) ratings apparently doesn't give good comparison to other countries. however, i am not yet convinced whose practice is better and therefore should be used universally. considering your passion for exact rules i am sort of surprised that you are against clear rules for rank assignment

tapir wrote:

... this seems impossible to achieve if rating resets are arbitrarily discouraged / forbidden in some countries...

in Czechia, rating resets are not arbitrarily discouraged, just not encouraged without a reason. it is unavoidable that even perfect ratings will be one tournament behind perfect self-assigned ranks, as ratings are based on player's previous performance. if someone attends tournaments infrequently or improves really fast, reseting rating is of course perfectly appropriate and for this reason it is implemented in GoR

Quote:

ps if you don't allow self ranking in czechia you can't study it. satisfaction of the playing population is not a criterion. but a good measure could be the willingness of the higher rated player to play tournaments with full handicap. i predict it would be considerably lower in the countries with artificially deflated kyu ranks.

then does this imply that if they (you?) don't base ranks on ratings in Netherlands, they can't study it?

i entirely agree that handicap games should give a very good measure of ranks quality. unfortunately i failed to find many such tournaments in EGD, it seems they are pretty rare. i can give an examples of two Czech handi tournaments, one from 2009 and one from 2010 (both with over 20 players, so there are enough various ranks included). i haven't analysed the results deeply but they look quite ok for me. i will greatly welcome links to full-handicap tournaments from other countries

PS: it sounds really surprising to me that you want to unify European rankings by abandoning ( / not using) a rating system. i was under impression that ratings were invented for this very purpose and that it was the reason why EGF adopted GoR. it only failed in the following step - to enforce use of the unified rank/rating policy in member countries. current situation of inconsistent policies in different countries further hurts good comparison of both ranks and ratings

i am not saying that GoR and current Czech ranking rules are the best possible, but i find it hard to believe that there is not a rating system better than everyone playing with the rank he wants

PPS: do chess or other games and sports also face the same problem? i would expect that someone had to already find a satisfying solution over the years

John Fairbairn · **#11**

Quote:

i am not saying that GoR and current Czech ranking rules are the best possible, but i find it hard to believe that there is not a rating system better than everyone playing with the rank he wants

It sounds to me as if at last awareness is creeping in that trying to keep track of kyu grades is go's equivalent of the Uncertainty Principle.

In fact, it is well known empirically that even at dan level playing strength in tournaments can vary up to two stones downwards depending on whether an early-morning start was required, how stressful driving to the event was, how cold the room was, how much the wife nagged, etc. etc. Pros, and a few amateurs, have learned how to blot out distractions but most of us are just puppets of the environment, so that even dan ranks are unreliable from week to week. The latest results among the top Europeans are proof enough of that. Ratings, at least in the short term, are just an opium for the go masses. As with sperm, coming first is all that matters. Anything beyond that is irrelevant in reality.

HermanHiddema · **#12**

Chess, I believe, has mostly moved towards the Glicko system, away from Elo. The EGF Ratings are basically just Elo ratings with an extra factor thrown in to make handicap based ranks line up with 100 rating points.

Glicko ratings add a confidence factor, and allow ratings with lower confidence (few data points) to change significantly faster than ratings with high confidence (many data points)

Regardless, any rating system can only perform well if it has sufficient data. The current state of affairs for Go in Europe does not, IMO, provide that. Most people, especially weaker players, play one, maybe two, tournaments per year. That is not enough data, especially in a strength range where quick progress is likely.

As you get to the range of stronger players, the strong kyu (1-3 kyu) players and dan players, you find that they also play more tournaments. So for those players, the rating system becomes progressively more meaningful. In my opinion, the current rating system is reasonable (i.e. generally accurate to within about 1 rank or 100 rating points) from about 3 dan forward. It is also, IMO, inflated by about 50 points.

One thing that is, IMO, a very strong conclusion from the research above is that self-chosen ranks work very well. Sure, we all know cases of players significantly under or overrating themselves, but those are the exception. Most players I know are quite aware of their own playing strength, and quite conscientious about registering at the "correct" rank. Here, social interactions and peer pressure are at play to keep people honest.

And of course, self-chosen ranks are not random. People have a reason for choosing those ranks. Very often, those ranks are based on their games with players at their club, and the handicap they need among themselves, and what the stronger players tell them their rank is. Or, similarly, on their online games, and what rank they have there, and again what stronger players tell them.

And it works. It works, in fact, very well.

So, for me, the main reason that I do not support using ratings for kyu ranks is simple: If it ain't broken, don't fix it.

RobertJasiek · **#13**

Laman wrote:

i am sort of surprised that you are against clear rules for rank assignment

Clear is insufficient. Rules must also be good. The implied self-ranking rule "A player chooses the rank for which his winning rate best approaches 50% against opponents of the same rank, is clearly less than 50% against stronger ranked opponents and is clearly more than 50% against weaker ranked opponents." works better and is much more easily justified than so far any rating system.

Quote:

then does this imply that if they (you?) don't base ranks on ratings in Netherlands, they can't study it?

Ratings can be run independently of their usage but ratings can measure only reported results. A self-ranking can always rely on more results: also the not reported ones.

Quote:

i entirely agree that handicap games should give a very good measure of ranks quality.

Agree with whom? Not me!

Quote:

it sounds really surprising to me that you want to unify European rankings by abandoning ( / not using) a rating system.

Since self-ranking (when abiding by the implied rule) works better. (It is a different question whether ranks should be only increasing or dynamic.)

Quote:

i was under impression that ratings were invented for this very purpose

So the EGF rating systems inventors might have claimed. I don't.

Quote:

and that it was the reason why EGF adopted GoR. it only failed in the following step - to enforce use of the unified rank/rating policy in member countries.

It fails in many respects, see earlier discussions elsewhere.

Quote:

i find it hard to believe that there is not a rating system better than everyone playing with the rank he wants

Maybe because social control is missing in Czechia?

Quote:

PPS: do chess or other games and sports also face the same problem? i would expect that someone had to already find a satisfying solution over the years

The "solution" is that the media believe the food served (that rating numbers are meaningful and as precise as the appear to be).

tapir · **#14**

some other answers arrived in the meantime. i still leave my post as it is... just skip the outdated parts

Quote:

then does this imply that if they (you?) don't base ranks on ratings in Netherlands, they can't study it?

obviously there are ratings in the Netherlands. that is why andre could compare the predictive performance of self assigned kyu ranks with ratings w/ the known results. w/o self assigned ranks you obviously can't compare but just show the predictive performance of rating based kyu ranks. (i at least have not seen such a study, that the players who keep playing are 'happy' with the system is no evidence.)

Quote:

it sounds really surprising to me that you want to unify European rankings by abandoning ( / not using) a rating system. i was under impression that ratings were invented for this very purpose and that it was the reason why EGF adopted GoR. it only failed in the following step - to enforce use of the unified rank/rating policy in member countries. current situation of inconsistent policies in different countries further hurts good comparison of both ranks and ratings

enforcing an unified rigidly rating based system in the kyu ranks when you have some growing, some shrinking national populations without much interactions but an even input of new rating points per game (epsilon) (with no or very low additional input by resets) will lead to disaster. you will have an unified system. but finnish, english, french players will still not be evenly matched. but more importantly:

Quote:

i am not saying that GoR and current Czech ranking rules are the best possible, but i find it hard to believe that there is not a rating system better than everyone playing with the rank he wants

there is nothing wrong with the system if you allow some flexibilty. but especially in the kyu ranks rating resets = entering tournaments at self assigned ranks should be encouraged rather than discouraged. but of course if you have many underrated players already in a subpopulation you won't dare reset 3 ranks higher as obviously would be appropriate for some french players because there are other underrated players as strong as you at your current level.

in general, i believe the rating obsession does no good to the game. it is well known that rating considerations play an important role in tournament choice already now. it would be worse in an even stricter enforced system, because a stricter system will fuel the obsession even further. even now players talk for hours about their ratings at tournaments. i am really fed up with it and would not enjoy what a rigid system would do to the game / tournament scene on a social level. and if i look at tournament participation in one of the stricter countries = poland, it looks as if they are reduced to pre-hikaru levels in 2010 after some big growth in between. i am not really aware of the underlying causes but have a gut feeling that rating obsession and rank pride and the frustration of a part of the population played a role in this.

BobC · **#15**

Ranking is one of my bete noires in go. In about six weeks I may go to a BGA tournament (for fun god forbid) and I have no clue about what rank to declare. It seems this is a universal problem which is leading to dysfunctional tournaments.

The internet hasn't helped. In the olden days my limited experience was that playing a game a week against a strong player was almost unheard of and access to good learning material was limited and so deriving a rank on a timescale of months was appropriate. Now, when any zealot can play 200 games against some of the best players out there in a few months and have ready one-click ordering of good learning material. The obvious happened and ranks became volatile.

I have long since held that Kyu ranks should be tied to a formal learning structure as well as game performance. In that way you assure minimum standards and guide the learning experience for novices.

It strikes me that if the EGF put in place an online "satellite" tournament to every major tournament, then a number of things would happen:

1] A ranking could be determined for each player in the lead up the the event.
2] A token fee could be charged for online participation in a lead up event which would result in a credible rank.
3] Rank fees could be used to setup better assessment processes and provide prizes/defray cost of tournaments.
4] The satellite tournaments might serve to advertise the event and get higher engagement.

Ranking does seem to be a shambles as it seems neither to aid the learning process or inform healthy tournaments.

hyperpape · **#16**

Self-promotion is a good thing. In the AGA, you can rate club games, so long as participants agree and you either do not use time limits or use appropriate time limits (much like the Congress self-paired games). I've always thought that more people should make use of that option, rather than just playing two tournaments a year.

Laman · **#17**

HermanHiddema: thank you, your answer sounds reasonable and not like a flamewar i was afraid of starting

Robert Jasiek wrote:

Agree with whom? Not me!
...

So the EGF rating systems inventors might have claimed. I don't.
...

It fails in many respects, see earlier discussions elsewhere.

i meant to agree with tapir, who wrote "but a good measure could be the willingness of the higher rated player to play tournaments with full handicap. i predict it would be considerably lower in the countries with artificially deflated kyu ranks." in post #9. i haven't seen a proof, but i assumed that handicap system changes strength difference between players as it should, which would allow us to measure distance between ranks. handicap system might not work, so my argument my argument might not be correct

i meant all ratings altogether, that they should make possible a reliable comparison of different rated subjects

i didn't mean that GoR failed, but that EGF failed

tapir wrote:

obviously there are ratings in the Netherlands. that is why andre could compare the predictive performance of self assigned kyu ranks with ratings w/ the known results. w/o self assigned ranks you obviously can't compare but just show the predictive performance of rating based kyu ranks

although there are ratings in Netherlands, due to more frequent rating resets there are more points injected into the local system, so i think the ratings has different dynamics. for this reason i doubted the study results' aplicability elsewhere. i could be wrong

tapir wrote:

enforcing an unified rigidly rating based system in the kyu ranks when you have some growing, some shrinking national populations without much interactions but an even input of new rating points per game (epsilon) (with no or very low additional input by resets) will lead to disaster. you will have an unified system. but finnish, english, french players will still not be evenly matched.

i agree that different populations may diverge with unified rating. that is certainly a system flaw. but what bothers me is that with self-assigned ranks there is also absolutely nothing preventing ranks from diverging, so while it may work fine locally, globally it solves nothing.

i actually had an idea that the epsilon constant may not be same for all, but different for different countries and (automatically) adjusted according to results of international matches

PS: i am not against self promotion, if it is reasoned

LR24 · **#18**

hyperpape wrote:

Self-promotion is a good thing. In the AGA, you can rate club games, so long as participants agree and you either do not use time limits or use appropriate time limits (much like the Congress self-paired games). I've always thought that more people should make use of that option, rather than just playing two tournaments a year.

This is actually a pretty cool thing, I didn't know it exists. This would make for some more-serious games at your local club, too. The only problem I can find with it is too many games against the same person. The ratings wouldn't be too much accurate when compared from one club to another then. Not idea how it works out in practice though.

gaius · **#19**

I don't think I would like regular rated games at my club. I like to play club games casually (I go there for fun), and I don’t care too much about lost games. I do play tournament games at maximum focus though, so I like that EGF rating is determined by those. For me, the situation in Utrecht perfect: only the yearly club championship games and inter-club-league games are EGF rated. Other games at the club are not EGF rated but count only for the "club rating". At the same time, there are very many tournaments every year within 2-3 hours traveling distance where you can play for EGF rating. Excellent!

The situation will be different in more isolated regions though...

xed_over · **#20**

hyperpape wrote:

In the AGA, you can rate club games, so long as participants agree and you either do not use time limits or use appropriate time limits (much like the Congress self-paired games). I've always thought that more people should make use of that option, rather than just playing two tournaments a year.

have you seen the AGA's top ten list of rated games? http://www.usgo.org/ratings/
I know someone in our local club who takes full advantage of this

Report: European Go Congress 2011

Who is online