McMahon Bar

[Javaness]

Should there be clearer rules for deciding the top McMahon bar (or supergroup in weird cases) in European turnaments? Even some clearer guidelines. Recently I read comments about how tournaments have too big a top group, making the results too random at the top.

[Stable]

Random? You mean too many people had a chance to win? Oh noes!

/cynicism
(I'm sorry, it's been a long day...)

[TMark]

Whatever guidance on how to produce a unique winner from a large group of players in a large tournament falls down over the history. Many years ago, the European Championship was played by having one player per country over the rank of 4dan, the rest of the players playing in the main tournament. Then the numbers of players 4dan and over grew, but the bar was retained at 4dan. The BGA has argued consistently for a long time that the bar be set at a level to produce a unique winner, but sponsorship also got involved, where part of the tournament would be sponsored by the Ing Foundation and part by a Japanese sponsor. Unfortunately, I doubt if there will be a quick or easy resolution. It would be nice if some organisers said one week before the event that there would be a proper McMahon bar and no supergroup c**p, and dare the EGF to take the event away, but I don't expect it to happen.

Best wishes.

[amnal]

Random? You mean too many people had a chance to win? Oh noes!

/cynicism
(I'm sorry, it's been a long day...)

I think it probably refers to the fact that too many people above the bar means you don't get a unique winner, so the probability of needing a tie break of some kind increases massively. The normal tie break is SOS, which I assume is what's being considered 'random'.

[HermanHiddema]

Should there be clearer rules for deciding the top McMahon bar (or supergroup in weird cases) in European turnaments? Even some clearer guidelines. Recently I read comments about how tournaments have too big a top group, making the results too random at the top.

I'd be all for this. Something like this:

The top bar for a tournament with N rounds should be set in such a way that the number of players in the top group is

1. strictly larger than N
2. smaller than or equal to 2^N (two to the power N)
3. ideally smaller than 3xN

The first two items are mandatory. If no setting of the bar achieves it, then the TD must use a Supergroup to achieve the desired group size.

The last item is a guideline, and must be followed if any setting of the bar allows it without breaking the other two criteria. If no setting of the bar achieves the third criterion, then the TD is allowed use a Supergroup to achieve the desired group size, at his discretion. If he does not, then he should use a setting for the bar that results in the smallest possible number of players in the top group larger than 3xN.

[Stable]

Ah, thanks Amnal.

[Gresil]

Sounds good. What's on tap?

[Javaness]

These guidelines seem pretty good to me Herman, I wonder if the Pandanet Tour could be persuaded to adopt this? I don't care what happens in the European Congress

[HermanHiddema]

These guidelines seem pretty good to me Herman, I wonder if the Pandanet Tour could be persuaded to adopt this?

It might be a good guideline to adopt, though I think I would probably lighten them a bit more, change the third criterion to

3. ideally between 2xN and 3xN

to give TD's a bit more leeway. Every tournament is unique, and the distribution of players can be quite different between tournaments, so giving the man in the spot a little more leeway is probably a good thing.

I've once had to decide on a McMahon bar for a 5 round tournament that featured a 6 dan, three 5 dans, two 4 dans and no fewer than twelve 3 dan players

I don't care what happens in the European Congress

Me neither, there is just too much politics going on around that for me to care anymore.

[Phelan]

Sounds good. What's on tap?

I thought the exact same thing when I saw the title. If someone I know ever opens a bar, I'll recommend the name.
Edit: "Probably the only place you can get a decent cup of joe-seki."

[Koffein]

I've once had to decide on a McMahon bar for a 5 round tournament that featured a 6 dan, three 5 dans, two 4 dans and no fewer than twelve 3 dan players

then how about using gor to decide the topgroup? just put the x strongest player according to gor into the top group

[HermanHiddema]

I've once had to decide on a McMahon bar for a 5 round tournament that featured a 6 dan, three 5 dans, two 4 dans and no fewer than twelve 3 dan players

then how about using gor to decide the topgroup? just put the x strongest player according to gor into the top group

That's what I did. I put the bar at 3 dan, then put the 12 strongest players (by rating) in a super group, and paired the six 4/5/6 dans against the six 3 dans in the super group for the first round.

In the end victory was shared between a 5 dan with 4/5 who lost in the first round and a 3 dan with 5/5 who didn't start in the super group. And they hadn't played each other.

A good lesson that even the most careful choice of top bar and super group is no guarantee of getting a reasonable result

[willemien]

personally i think you made a wrong decision.

Players witha reasionable chance towin the tournament should all be in the topgroup. (players outside the topgroup should be informed that they cannot win any of the tournament prizes)

An alternative that i think would be better is to use accelerated swiss pairings for the first two rounds.

In practice this would mean:

group A the highest ranked 4 players (P1 - P4)
group B the next highest ranked 4 players (P5 - P8)
group c the next highest ranked 5 players (P9 - P13)
group B the next highest ranked 5 players (P14 - P18)


Round 1:

A vs B
board 1 P1 - P5 W1 = winner L1 = loser
board 2 P2 - P6 W2 = winner L2 = loser
board 3 P3 - P7
board 4 P4 - P8

C vs D
board 5 P9 - P14
board 6 P10 - P15
board 7 P11 - P16
board 8 P12 - P17
board 9 P13 - P18

Round 2:

- winners from A vs B
board 1 W1 - W3
board 2 W2 - W4

- Losers from AvsB are paired against winners CvsD
board 3 W5 - L1
board 4 W6 - L2
board 5 W7 - L3
board 6 W8 - L4
board 7 W9 - L5 (L5 is added to this group)

- losers CvsD play each other

board 8 L6 - L8
board 9 L7 - L9

From round 3 onwards the normal pairings can be used.

[HermanHiddema]

That really doesn't help. The second round boards 3-7 (winners vs losers) would all be extremely unpredictable games, as that group would consist almost completely of 3 dans. Which means you'll have anywhere between 2 and 7 players on 2 points after two rounds. Accelerated pairings were designed to make to sort out the top players faster, but in such a level field that doesn't work, especially if you make the initial top group larger.

I think there was very little wrong with my choice, the distribution of players over groups was quite similar to any other tournament. The strange result was due to the fact that one of the 4 dans, who had 2 points, fell ill and dropped out in the third round, and due to the 6 dan making a huge blunder against the 3 dan that got 5/5 in the fourth round (blunder lost him a large group, which lost him the game by 1.5 pt).

McMahon/Swiss allows ties, and if coincidence stacks on coincidence, strange things can happen

[willemien]

You are right.

It doesn't really help if the top players start losing to lower ranked players.
It makes strange things possible.

Still i think that of the criteria for setting the bar http://senseis.xmp.net/?McMahonPairing%2FBarTheory

the criteria that every player who is deemed to have a reasonable chance of winning the tournament should be in the top group, is the most important.

I have asked an expert in tournament pairing to do some experiments i am awaiting his reply.

BTW i think your avatar is highly amusing