Life In 19x19

http://lyon-shinogi.jeudego.org/calendr ... ampionship

What do you think about the tournament system?

In my opinion it was really stupid to play in one group in 5 rounds.
It was better to make 2 groups or add one extra round.

Now there are 5 girls with 4 points and it's hard to understand why Zhao Pei is first.

yes, i agree the system wasn't chosen very fitting to the situation and results prove it. there is not much of a point in playing swiss system with even games and players in range from 11k to 6d.

how did you meant two groups? two separated groups or just two macmahon groups? i would vote for MM, with top group consisting of 8 (= dans) or maybe 12 players (= dans + 1-3k) and starting MM 1, others with MM 0.

2 MacMahon groups usually. It's the best decision

I would prefer accelerated Pairings
http://senseis.xmp.net/?SwissTournament%2FacceleratedPairings

is a bit fairer,

Everybody has (at least a theoretical) chance to become the winner. (and it was a championship, not a ordinary tournament)

I guess that they are forced to play this tournament in World Amateur Style though.
Quite rare to see a 5 player tie at the top of the tournament. Maybe next year the EGF will give some direction for the rules?

With such a high distribution of ratings, tournament system was effectively 4 round swiss tournament. And therefore as there was not someone, who was clearly above others and would win all her games, it is _very likely_ that there are several players who will share top place.

This would support the idea of more than one McMahon groups. However, it is important to keep in mind that top group size is rather irrelevant, because distribution of players' skill is the thing that matters. Idea is that first round should not be trivial for all those top group players who are seeded to top half of the top group in first round. If it is trivial (like in this case), tournament loses one effective round.

Accelerated pairing feels like a dirty trick, but since I cannot find obvious flaws in the logic, I would ask for the real world examples. Preferrably from EGD, because I would like to study pairing in detail, with my Asparagus.

Liisa wrote:Accelerated pairing feels like a dirty trick, but since I cannot find obvious flaws in the logic, I would ask for the real world examples. Preferrably from EGD, because I would like to study pairing in detail, with my Asparagus.

It is sometimes used in chess tournaments.

In go it is not often used because the topgroup is small and has a similar strength and in such cases Accelerated pairing can give more players with 2 wins than normal pairings. (but i did suggest it for future EGC topgroup tournaments)

I don't think that there is a Logical flaw in Accelerated pairing.

But you do have the risk (if the origiunal seeding is wrong) that instead of less there are more people with the 2 wins after 2 rounds.


The result of Accelerated pairing is that the players in the lower half need to win 2 games before they are paired against the players with 2 wins in the higher half. (this is based on a 4 group setup, see the sensei's article)
The prognosis is that they will not win these 2 games, but if they do they are still able to win the tournament.

Accelerated Pairings and the McMahon system are both attempts to solve the same problems: Too many players, or too wide a skill difference, to make Swiss attractive.

In general, McMahon is better at it, but in situations where it is required that all players have a (theoretical) chance to win the tournament, Accelerated Pairings allow you to start everyone in the same group, while still avoiding most wide skill gaps.

If it is required that Swiss be used for the EWGC, then AP would be a good idea. If it is not, then McMahon is a better choice, IMO.

I added tournament to my Asparagus, so that pairing can be studied in detail and make appropriate speculations. Indeed it looks like Vanessa would perform as well as Pei, and Rita had bad luck with pairing.

http://valkonen.kapsi.fi/parsa.php?touid=Q101015A

willemien wrote:I don't think that there is a Logical flaw in Accelerated pairing.

I think that ethical problem is that accelerated pairing gives lower rated players by definition disadvantageous SOS, because they need to play weaker opponents in first two rounds. On the other hand they will get stronger opponents in the latter rounds, because players are folded according sos, but I do not think that this will be enough to even out sosses within 5 rounds.

Other issue is that there might be other hazards in latter rounds if pairing does not go well even. Especially round three is difficult. Like in this case Rita was paired out of any SOS fights in round three. This is not good thing!

So as there are some problems with accelerated pairing, and benefits are little, normal McMahon is just plain better IMO. It might be better than regular Swiss, but Swiss in general suits very badly in go tournaments, imo. This is I think the reason why McMahon was invented! (Although, we should import McMahon into Chess tournaments, because it would suit also Chess tournaments better than Swiss =)

Liisa wrote:I added tournament to my Asparagus, so that pairing can be studied in detail and make appropriate speculations. Indeed it looks like Vanessa would perform as well as Pei, and Rita had bad luck with pairing.

http://valkonen.kapsi.fi/parsa.php?touid=Q101015A

willemien wrote:I don't think that there is a Logical flaw in Accelerated pairing.

I think that ethical problem is that accelerated pairing gives lower rated players by definition disadvantageous SOS, because they need to play weaker opponents in first two rounds. On the other hand they will get stronger opponents in the latter rounds, because players are folded according sos, but I do not think that this will be enough to even out sosses within 5 rounds.

Other issue is that there might be other hazards in latter rounds if pairing does not go well even. Especially round three is difficult. Like in this case Rita was paired out of any SOS fights in round three. This is not good thing!

So as there are some problems with accelerated pairing, and benefits are little, normal McMahon is just plain better IMO. It might be better than regular Swiss, but Swiss in general suits very badly in go tournaments, imo. This is I think the reason why McMahon was invented! (Although, we should import McMahon into Chess tournaments, because it would suit also Chess tournaments better than Swiss =)

I haven't thought about the SOS problem with accelerated pairing, I thought there were enough rounds left to sort that reasonably out.
Maybe the only solution here is to use SOS-2 (Don't count the 2 lowest SOS scores) instead of normal SOS


The problem with (normal) McMahon is that you allready before the start of the tournament the organisation denies some people the possibility to win the title. (everybody under the bar is no contender for the title)

willemien wrote:The problem with (normal) McMahon is that you allready before the start of the tournament the organisation denies some people the possibility to win the title. (everybody under the bar is no contender for the title)

I do not see a problem here. If someone thinks that their actual skill should be enough to compete from victory (e.g. not playing in tournaments recently), they can adjust their rank so that it is high enough to include into top group.

SOS-1 or SOS-2 of course would solve that issue, but this would also reduce the accuracy of SOS because there is not too much data to waste in 5 rounds. Problem is that it is just impossible to find justified winner from 24 players in five rounds in all cases. More rounds is needed or smaller top group. My opinion is that 12-14 is upper limit for top group in 5 round tournament. But more important than size, is that top group's skill distribution is narrow enough.

Liisa wrote:

willemien wrote:The problem with (normal) McMahon is that you allready before the start of the tournament the organisation denies some people the possibility to win the title. (everybody under the bar is no contender for the title)

I do not see a problem here. If someone thinks that their actual skill should be enough to compete from victory (e.g. not playing in tournaments recently), they can adjust their rank so that it is high enough to include into top group.

But if you allready only have such a small tournament as here splitting it up in two groups (a topgroup and the rest) is hardly a option either

Liisa wrote:SOS-1 or SOS-2 of course would solve that issue, but this would also reduce the accuracy of SOS because there is not too much data to waste in 5 rounds. Problem is that it is just impossible to find justified winner from 24 players in five rounds in all cases. More rounds is needed or smaller top group. My opinion is that 12-14 is upper limit for top group in 5 round tournament. But more important than size, is that top group's skill distribution is narrow enough.

Some people say that SOS looks more accurate than it actually is. so if it gives more with the same SOS score can be seen as more truthful.

The problem in this tournament was more that nobody won all her games (that would give a clear winner, and there could only be one of them)

But also under that situation (one player winning all her games) the problem would be that it likely is that more players would have 4 wins.

i wanted to see what your progam would make as initial wallchart

but if on

http://valkonen.kapsi.fi/parsa.php?touid=Q101015A

I click on one of the lins after
Results after round:
I get directed to the Finnish Champhionship Qualifications (for my unknown reasons)

Handpicked anti-randomized pairings (old-school with pen and paper) would not have shown this behaviour. There were six players 3d and above and 5 rounds - does not sound too hard to pick suitable opponents for all who are still contending the title.

I am amazed by quotes like: "SOS-1 or SOS-2 of course would solve that issue, but this would also reduce the accuracy of SOS because there is not too much data to waste in 5 rounds." A glance at the result table shows that there is no accuracy to talk about. There is a two point difference (SOS) between first and second placed just by the truly random first game and later on this difference did increase, even though the second did not lose a game (until the last round). So yes, the first one played a harder tournament (SOS rationale), but 2nd to 5th did not get a chance to play it. Which is in a way a feature, if you lose in the second round, but obviously it is a flaw if you have straight wins. So we got the opportunity to see a 3 point SOS-difference which still does not look convincing.

I wait for the day when a WAGC is decided by whether the winner was paired against a 5k or a 2d in the first round, maybe some people will then realize that there is something wrong with the system.

tapir wrote:A glance at the result table shows that there is no accuracy to talk about. There is a two point difference (SOS) between first and second placed just by the truly random first game and later on this difference did increase, even though the second did not lose a game (until the last round). So yes, the first one played a harder tournament (SOS rationale), but 2nd to 5th did not get a chance to play it. Which is in a way a feature, if you lose in the second round, but obviously it is a flaw if you have straight wins. So we got the opportunity to see a 3 point SOS-difference which still does not look convincing.

Indeed three point difference is here far less convincing than it normally should be. If you examine closer the pairing first round was arranged by GoR as it should arrange in regular Swiss. Random is always bad option in first round because it can give huge benefit or handicap just by good or bad luck. Therefore naturally Pei got strongest opponent in the first round of top half players.

Second issue was that Rita was paired downwards on third round. This gave her intolerable disadvantage.

Third issue was that Pei was paired manually constantly against other top players. This gave Pei huge advantage in winning competition. Three point difference in SOS should not be possible in regularly paired tournament.

And fourth Natalja lost on second round, but this should not exclude her away from winning competition. This is the reason why 20 players is over sized group size with this distribution of ratings.

I wait for the day when a WAGC is decided by whether the winner was paired against a 5k or a 2d in the first round, maybe some people will then realize that there is something wrong with the system.

In swiss pairing it is just plain wrong to use random in the first round, because of this reason.

Willem wrote:But if you allready only have such a small tournament as here splitting it up in two groups (a topgroup and the rest) is hardly a option either

I do not see point here. I stated that maximum size is 12-14 imo, but minimum is 6-8 for top group. But this depends heavily on rating distribution. In this particular tournament I would choose 10 players into top group (1k+) and for the rest regular McMahon tournament with MMS-1 handicaps.

Idea is that in this case first round was wasted, because smallest GoR difference was more than 800 points between pairs. I would say that minimum difference between pairs should not be much bigger than 300 points in the first round.

i wanted to see what your progam would make as initial wallchart

but if on

http://valkonen.kapsi.fi/parsa.php?touid=Q101015A

I click on one of the lins after
Results after round:
I get directed to the Finnish Champhionship Qualifications (for my unknown reasons)

I think that your session was just expired. You need to click again link. If this is not the case then there is a serious bug! That Finnish championship qualification is default tournament that is shown if no other tournament is selected with "?touid=", or session is expired.

tapir wrote: I wait for the day when a WAGC is decided by whether the winner was paired against a 5k or a 2d in the first round, maybe some people will then realize that there is something wrong with the system.

Usually the draw is "fixed" for the first round, to avoid the top contenders playing someone with no chance of competing. The organisers are sufficiently aware of the possibility to try to avoid the repercussions.

Best wishes.