Best play-off format?

[paulg]

As one of the Scottish selectors I am looking at play-off formats for the European Team Championships next year. The play-off will consist of two teams playing 128 boards, the winner being selected for the championship.

 

The format of all the European Team Championships are based a round-robin of 16- or 20- board matches scored by VPs.

 

My question is whether it is better to have a single 128 board match, or to have eight 16-board matches each scored by VPs.

 

Does anyone have strong views on which method will get me the team that will perform better in the Championships?

 

Thanks,

 

Paul

[mich-b]

I think a single 128 boards match is better.

The convertion from IMPs to VP is useful in a long round robin with a large variance of strength between the teams (like the European championship itself), and is needed to prevent large wins against the weaker teams deciding the winners.

This is not needed when 2 teams are competing - every IMP won there was won against the same quality of opps, and should count the same, independant of how many IMPs were won or lost in that segment.

For example if a team is winning a segment in the playoffs by 50 IMPs , and scores 10 more, I think they should count fully, unlike if this happens in a match against weak opps, which is part of a round robin.

[Codo]

I do not know if there is anybody out there who tried it before, so we all have just our opinions. Here is mine:

 

I would go for the 8 matches.

 

When a team is a little better on average but has one nightmare round, it may lose the 128 game but still win the competetion over 8 matches.

 

Example: You win 7 matches with 10 imps (17-13) and lose one cause of tireness, illness or whatever other reason with 100 imps (0-25).

 

In this case you had been more successful in the RR but lose the 128 game match.

 

I did not calculate this for many possible scenarios, but overall 8 matches should be closer to the needs of a RR then one single match.

[Lobowolf]

Another vote for one long one.

[Trinidad]

One long one is better to determine the best team. All 128 boards count equally.

 

If you take 8 16 board matches then the slam swing on board 1 of match 1 will not count equal to the slam swing on board 1 of match 2.

 

Example: In match 1, pair 1 of team A holds the NS hands and bids and makes 7♠. Pair 2 of team B bids 7♠ and goes down 1. The swing is 1510 + 50 = 1560 for 17 IMPs to team A.

 

In match 2, by an incredible coincidence, board 1 is exactly the same as board 1 of match 1. But now pair 1 of team B gets the very NS hands and bids and makes 7♠ whereas pair 2 of team A goes down. 17 IMPs to team B.

 

Of course, you want these results to cancel each other. In a 128 board match that is what happens: the score will be 17-17 in IMPs.

 

In 8 16 board matches this is not the case. Suppose that in match 1, team A blitzes team B. They win by a landslide, which converts to 25-0 in VPs. If they would have tied board 1, it still would have been 25-0. No difference in the VPs for A or B as a result of the slam swing on board 1.

 

In the second match team B would have lost by 8 IMPs, if they would have tied the board. This would convert to a 12-18 loss in VPs. But they did have the slam swing and win by 9, scoring and 18-12 win in VPs. Now, the very same slam swing suddenly gains team B 6 VPs and loses team A 6 VPs.

 

If you want to select a team, you want to use the most accurate way to predict the better team. One 128 board match is just more accurate than 8 16 board matches.

 

Rik

[TimG]

The format of all the European Team Championships are based a round-robin of 16- or 20- board matches scored by VPs.

 

My question is whether it is better to have a single 128 board match, or to have eight 16-board matches each scored by VPs.

Do you think team strategy should be different depending upon whether a team is playing a series of 16-board matches as opposed to a single 128 board match?

 

If you think there should be a difference in the way teams approach these two possibilities, then I think a good case can be made for your trials to mimic the championship format.

[paulg]

The format of all the European Team Championships are based a round-robin of 16- or 20- board matches scored by VPs.

 

My question is whether it is better to have a single 128 board match, or to have eight 16-board matches each scored by VPs.

Do you think team strategy should be different depending upon whether a team is playing a series of 16-board matches as opposed to a single 128 board match?

 

If you think there should be a difference in the way teams approach these two possibilities, then I think a good case can be made for your trials to mimic the championship format.

I knew that posting the question would help.

 

I think that the 16-board match rounds are pretty much irrelevant to the team strategy and, when I've been NPC, I've told the team that they need to consider it as a 320 board tournament. The conversion to VPs make little difference in the long run.

 

Thanks for reminding me!

[benlessard]

16 boards is a bit long to keep a running estimate. But if the caliber is very high and you hit a streak of dull boards than "state of match" strategies become importants. To play 1 long match slightly penalize the players who are very good at keeping a running estimate.

 

I also think that in teams with 3 pairs or with a playing sponsor or substitute than dividing into 16B make sense since in the tournament you often have the chance to send the weak pair against the weak team.

 

In both case the difference is pretty small so it shouldnt be a huge concern.

[Gerben42]

probably eight 16-boards matches. Even better would be 20-board matches, if you look at Italy's record on boards 17-20.

[Hanoi5]

The problem I see with the VP's format is that a team could win in VP's and not IMP's, they just need 'that' luck.

 

When only two teams are matching (like in a final or semifinal) only imp's are taking into account. If there are only two teams for the trial that should be the unit measuring the win: imp's.

 

However, I do think that there should be segments in this 128-board match (though I'm sure that'll be taken care of).

[hrothgar]

The format of all the European Team Championships are based a round-robin of 16- or 20- board matches scored by VPs.

 

My question is whether it is better to have a single 128 board match, or to have eight 16-board matches each scored by VPs.

Do you think team strategy should be different depending upon whether a team is playing a series of 16-board matches as opposed to a single 128 board match?

 

If you think there should be a difference in the way teams approach these two possibilities, then I think a good case can be made for your trials to mimic the championship format.

I knew that posting the question would help.

 

I think that the 16-board match rounds are pretty much irrelevant to the team strategy and, when I've been NPC, I've told the team that they need to consider it as a 320 board tournament. The conversion to VPs make little difference in the long run.

 

Thanks for reminding me!

I think that this question is very well suited to a Monte Carlo simulation.

 

Create a population of bridge teams that differ in terms of their average performance on a board as well as the board variance. In theory, for a given tournament format, you should be able to define an "iso-win" curve that shows the different combination performance / variance that produce an identical chance of winning said tournament. For example, you might determine that an average performance of

 

+ X IMPS per board with a variance of Y is exactly equivalent to

+ A IMPS per board with a variance of B

 

Repeat the same experiment using the second tournament format.

Identify whether you observe any significant differences in the shape of the iso-win curves.

 

If the curves are shaped the same, run whatever format you want. If the curves are shaped differently, then you will (probably) discover that you want to mirror the Conditions of Contest of the real event.

[TimG]

Create a population of bridge teams that differ in terms of their average performance on a board as well as the board variance.

I don't think this will work so well. Players will presumably be taking into account "state of the match" so the average performance and board variance will not remain constant.

[hrothgar]

Create a population of bridge teams that differ in terms of their average performance on a board as well as the board variance.

I don't think this will work so well. Players will presumably be taking into account "state of the match" so the average performance and board variance will not remain constant.

Hi Tim:

 

There's a couple different ways to interpret your post:

Interpretation 1: Neither of the tournament formats that is being discussed is a particularly good approximation for a round robin tournament. I think that this is a fair comment.

 

Interpretation 2: My argument requires an assumption of constant performance and variance across boards. Not sure that I agree with this.

 

My argument is based on a (hypothetical) analysis of the shape of iso-win curves. The only thing that my argument requires is that the shape of said curves are different.

 

Let's assume that we performed said analysis and the curves do, indeed, have different shapes. I would be shocked if a "State of the Match" set of adjustments to bidding / play style would cause the shape of the two sets of curves to converge...

[cherdanno]

The format of all the European Team Championships are based a round-robin of 16- or 20- board matches scored by VPs.

 

My question is whether it is better to have a single 128 board match, or to have eight 16-board matches each scored by VPs.

Do you think team strategy should be different depending upon whether a team is playing a series of 16-board matches as opposed to a single 128 board match?

 

If you think there should be a difference in the way teams approach these two possibilities, then I think a good case can be made for your trials to mimic the championship format.

I knew that posting the question would help.

 

I think that the 16-board match rounds are pretty much irrelevant to the team strategy and, when I've been NPC, I've told the team that they need to consider it as a 320 board tournament. The conversion to VPs make little difference in the long run.

 

Thanks for reminding me!

Since you asked the question, I was automatically assuming you have the opposite opinion! The only reason I could see to play it as 8 matches of 16 boards would have been to give players practice with the state of the match considerations. (When you are down, there is a bit of advantage to swinging.)

[TimG]

We've all been in a position where we know that we need a big win to have a chance of winning a Swiss and we've probably taken some gambles that we would otherwise not in order to give ourselves (what we perceived as) a better chance to win.

 

Our average IMPs and variance will not be the same during such a match as it would be earlier in the event. Unless your model takes into account this possible change in average/variance, I don't think it will be particularly useful. Especially since the debate over format is very much about whether this sort of change in play is measurable or effective. It seems to me that your model is assuming constant average/variance.

[ArtK78]

As a practical matter, I doubt that there is any significant difference between one 128-board match and 8 16-board matches scored at VPs. There will be differences in the result but those differences will essentially be random. I am sure that you can construct scenarios where one team wins a close 128 board match and the other team would have won had the match been scored at VPs in 8 16-board matches.

 

Still, if you want to have the best team win, I would choose a 128 board match as opposed to 8 16-board matches.

 

The fact that the European Team Championship is a round robin event should not influence the choice of the format. A better team will win more often in any format.

[MFA]

The format of all the European Team Championships are based a round-robin of 16- or 20- board matches scored by VPs.

 

My question is whether it is better to have a single 128 board match, or to have eight 16-board matches each scored by VPs.

Do you think team strategy should be different depending upon whether a team is playing a series of 16-board matches as opposed to a single 128 board match?

 

If you think there should be a difference in the way teams approach these two possibilities, then I think a good case can be made for your trials to mimic the championship format.

I knew that posting the question would help.

 

I think that the 16-board match rounds are pretty much irrelevant to the team strategy and, when I've been NPC, I've told the team that they need to consider it as a 320 board tournament. The conversion to VPs make little difference in the long run.

 

Thanks for reminding me!

Since you asked the question, I was automatically assuming you have the opposite opinion! The only reason I could see to play it as 8 matches of 16 boards would have been to give players practice with the state of the match considerations. (When you are down, there is a bit of advantage to swinging.)

Uh oh, sounds like very dangerous tactics.

 

I strongly prefer to ignore such things. Swinging because of speculative match considerations would just open myself to the tilt factor, turning normal losses into blood baths.

 

It's entirely possibile that a scientist with an appropriately effective microscope would be able to spot profitable swing options in a long round robin tournament.

 

But for practical purposes this issue is much better ignored. I know (also very good) players who are willing to swing in a shortish round robin match. I consider this to be nothing but a giant leak in their game.

[655321]

The only reason I could see to play it as 8 matches of 16 boards would have been to give players practice with the state of the match considerations. (When you are down, there is a bit of advantage to swinging.)

Uh oh, sounds like very dangerous tactics.

 

I strongly prefer to ignore such things. Swinging because of speculative match considerations would just open myself to the tilt factor, turning normal losses into blood baths.

 

It's entirely possibile that a scientist with an appropriately effective microscope would be able to spot profitable swing options in a long round robin tournament.

 

But for practical purposes this issue is much better ignored. I know (also very good) players who are willing to swing in a shortish round robin match. I consider this to be nothing but a giant leak in their game.

Strongly agree with MFA.

[aguahombre]

only two teams? maybe their captains agree.

[paulg]

probably eight 16-boards matches. Even better would be 20-board matches, if you look at Italy's record on boards 17-20.

We will consider this, depending on logistics etc, and make it a 120-board match rather than 128.

[paulg]

only two teams?  maybe their captains agree.

I think this shows a fundamental lack of understanding of Scottish bridge and politics :) :)

[MFA]

... if you look at Italy's record on boards 17-20.

Did anyone actually make an analysis of this or is it just a hunch that they do well on the last boards?

[OleBerg]

Another vote for one long one.

"That's what she said".

[OleBerg]

A slight argument for one long match, could be to avoid trouble.

 

If you play 8 matches, and it is very close, somebody could feel cheated if they lost 4½ - 3½ but scored most IMP's.

 

I would consider it much less likely, that a team that lost by, say, 4 IMP's in one long match, would complain that they would have won, if it had been 16 board matches. (They probably wont even do the math.)

[MickyB]

5,6,4,4,5,5,5,5,5,6,6//6,6,7,7,inf - 20 board matches

5,5,4,4,4,4,4,4,5,5,5//6,6,6,7,inf - 16 board matches

 

^----------------------^------------^

15-15..................25-5..........25-0

 

Those are the IMP ranges for a particular VP score (apologies for the formatting). Unlike the EBU scale, the WBF scale isn't far off being linear, so there isn't much reason to swing when you are losing unless you think you are already losing 25-0 - hopefully a rare occurance!

 

The main thing that you might wish to consider is that, when you are in 25 VP territory, it is desirable to reduce variance. This presents a problem, as the team that is losing shouldn't have as great a desire to increase variance - basically, the matches in the Euros aren't zero-sum, but obviously your trial does have to be zero-sum.

 

The closest you could get to emulating this would be to score purely by IMPs, but with two caps -

 

Upto the IMP total required for 25-5, IMPs count 100%

Between the IMP totals required for 25-5 and 25-0, IMPs count 50%, to reflect the fact that you only wish the IMPs to count for one of the two teams.

Beyond the IMP total for 25-0, extra IMPs do not count.

 

Compared with scoring by VPs, the most likely effect of this method would be to reduce the chance of a close match being decided by who got [to steal a phrase from MGB] "value for money" on the VP scale, which is clearly desirable.