[paulg]

 bluejak, on 2011-August-26, 10:50, said:

But I have played in Swiss Teams tournaments where the results were shown in both VPs and total imps. This leads to no ties [with incredibly rare exceptions] and a simple method everyone understands and can see. It seems best to me.

This is the method used in Scrabble tournaments these days. SOS used to be more popular but is rarely seen nowadays and spread (net points) is the default tie breaker. Like chess, Scrabble is win, lose or draw but spread is not used in creating the matches (except near the end of a tournament when they use the principle of giving people the best chance of winning the better prize, but that is another topic).

[jnichols]

There are lots of ways to break ties. Some are certainly better than others. Apparently which are better is a matter of opinion.

Most important, IMHO, is that the method(s) to be used are known, in advance, to the players (at least if they bother to read the Conditions of Contest) and that it is easy for the players to see that they are being applied properly. Simple methods like the result in a head to head match, or sum of scores of opponents, or swiss points are easy for a player to verify. Whatever you apply, and in whatever order, will be unpopular with the players who don't win, but that can't be avoided. Loosing a tie breaker by a methond they can't understand will have players grumbling not about loosing, but about the fairness of the tournament and that we need to avoid.

[bluejak]

That is one of the advantages of the method I espouse. If you have a result sheet that shows:

Pos  Team    VP  imps

 1   Green   28  +124
 2   Blue    28  +113
 3   Red     26   +56
 4   Yellow  17    +4
 5   Purple  14   -17
 6   Indigo   4  -217

then players not only understand it but they do not think of it as a tie break at all: it is just the final order.

[pran]

 bluejak, on 2011-August-27, 08:26, said:

That is one of the advantages of the method I espouse. If you have a result sheet that shows:

Pos  Team    VP  imps

 1   Green   28  +124
 2   Blue    28  +113
 3   Red     26   +56
 4   Yellow  17    +4
 5   Purple  14   -17
 6   Indigo   4  -217

then players not only understand it but they do not think of it as a tie break at all: it is just the final order.

Assuming the imps are the net imps (won less lost) I have just one question:

Do you consider Green better than Blue if their victory is for instance 248-124 while Blue has won 150-37?

[campboy]

 pran, on 2011-August-27, 12:49, said:

Do you consider Green better than Blue if their victory is for instance 248-124 while Blue has won 150-37?

Of course. We all -- because we are using a VP scale which says so -- consider a 148-124 victory better than a 50-37 victory; why would your scenario be any different?

[aguahombre]

....

This post has been edited by aguahombre: 2011-August-27, 19:13

[pran]

 campboy, on 2011-August-27, 16:12, said:

Of course. We all -- because we are using a VP scale which says so -- consider a 148-124 victory better than a 50-37 victory; why would your scenario be any different?

In case you didn't notice: The way I understood it Green and Blue ended tied, both with 28VP, the IMPs are their respective grand total IMPs used to resolve the tie.

Maybe I misunderstood something?

[campboy]

Since the difference in IMPs is all that matters for assigning VPs to the individual matches, it is inconsistent to consider some other function of IMPs won to IMPs lost more important than the difference when calculating the tie-break.

[pran]

 campboy, on 2011-August-28, 03:14, said:

Since the difference in IMPs is all that matters for assigning VPs to the individual matches, it is inconsistent to consider some other function of IMPs won to IMPs lost more important than the difference when calculating the tie-break.

Yes, and one traditional way of breaking ties has been to compare the total number of IMPs won and lost over all the matches played in some way. There is little or no correlation between the IMP figures and the VP figures when you use such methods to break ties.

[campboy]

You should subtract the IMPs lost from the IMPs won. This is what you do with the match IMPs to work out the result of each match, so it is what you should do with the overall IMPs to work out your tie break. I don't think I can put it any more simply.

[pran]

 campboy, on 2011-August-28, 08:14, said:

You should subtract the IMPs lost from the IMPs won. This is what you do with the match IMPs to work out the result of each match, so it is what you should do with the overall IMPs to work out your tie break. I don't think I can put it any more simply.

That is exactly what I assumed.

And my question was, and still is: When tiebreaking between Green and Blue, both teams ending at the same VP, which team is considered the better:
The team with a positive overall IMP result of 124 as the difference between 248 won and 124 lost, or
the team with a positive overall IMP result of 113 as the difference between 150 won and 37 lost?

I am still interested in an answer, and honestly I believe that the team with a difference of 113 from a total of only 187 IMPs scored expects to be ranked ahead of the team with a difference of (only) 124 from a total of 372 IMPs scored. Don't you?

[barmar]

What I think you're saying is that rather than looking at the IMP difference, we should look at the ratio of IMPs won to total IMPs. So in the 248 vs. 124 team, they won 50% of their total IMPs, while the other team won 80% of the total IMPs.

This doesn't seem unreasonable.

[pran]

 barmar, on 2011-August-28, 08:43, said:

What I think you're saying is that rather than looking at the IMP difference, we should look at the ratio of IMPs won to total IMPs. So in the 248 vs. 124 team, they won 50% of their total IMPs, while the other team won 80% of the total IMPs.

This doesn't seem unreasonable.

Indeed I do, and that was the standard tie-breaking regulation we had in Norway until recently. (We actually used the quotient between won and lost, which essentially amounts to the same). The regulation has been more elaborated now, I shall not bother this forum with the details.

[campboy]

Do you think winning an individual match 5-1 is better than winning it 16-8?

It is not unreasonable to consider the quotient more important than the difference, but it is inconsistent to consider the difference more important until you get to the tie-break and then suddenly decide to use the quotient.

[pran]

 campboy, on 2011-August-28, 11:05, said:

Do you think winning an individual match 5-1 is better than winning it 16-8?

It is not unreasonable to consider the quotient more important than the difference, but it is inconsistent to consider the difference more important until you get to the tie-break and then suddenly decide to use the quotient.

No, the VP result from an individual match will be better for the team that wins with 16-8 than for the team that wins with 5-1.

This thread is about breaking ties, and there I believe a high IMP quotient is an indication of better bridge played for the same VP result because it means that the (comparable) difference in IMPs was obtained with less total amount of IMPs issued during the matches.

[bluejak]

 pran, on 2011-August-27, 12:49, said:

Assuming the imps are the net imps (won less lost) I have just one question:

Do you consider Green better than Blue if their victory is for instance 248-124 while Blue has won 150-37?

No. But I don't consider it relevant in any way.

The question is about breaking ties, not who is the better team. A far better method is to have no ties.

[pran]

 bluejak, on 2011-August-28, 18:45, said:

 pran, on 2011-August-27, 12:49, said:

Assuming the imps are the net imps (won less lost) I have just one question:

Do you consider Green better than Blue if their victory is for instance 248-124 while Blue has won 150-37?

No. But I don't consider it relevant in any way.

The question is about breaking ties, not who is the better team. A far better method is to have no ties.

Sorry, my error.

I wasn't aware that you don't care about who is the better contestant when breaking ties, only that the tie should be broken somehow.

[FrancesHinden]

 pran, on 2011-August-29, 00:29, said:

Sorry, my error.

I wasn't aware that you don't care about who is the better contestant when breaking ties, only that the tie should be broken somehow.

None of these methods actually defines who is the 'better' contestant. The main form of scoring has decided that the two contestants are equal. Now you are coming up with something that is always going to be somewhat random.

The problem with imp quotient is at least four fold. First of all the teams won't know who has won (unless you are quoting it each round) and it feels a somewhat random differentiator. Secondly it is completely meaningless unless everyone has played the same boards (which is not the case in many North American swiss events). Thirdly the idea that conceding fewer imps is somehow 'better' doesn't actually hold up when you analyse it carefully. If I bid 10 vulnerable 4H contracts, each of them with a 40% chance of making exactly (and a 60% chance of going one off), four of them make and 6 go one off, and the other team are always in 3M (or 2M+1) then I will lose 36 imps and gain 40 imps. If the other team plays all of these contracts in partscore (but gained 4 imps some other way on one board) then their imp turnover is much much lower than mine, but I think my bidding is better. Fourthly, high variance bridge (leading to bid pluses and minuses) is not by definition worse than low variance bridge; all that matters is the expectation.

The problem with total (net) imps is that early matches are sometimes won very heavily (I have reasonably often won 8-board Swiss matches by 60+ imps) when a good team plays a very weak team. The VP scale cuts this huge margin off... but now you are reintroducing it to split a tie i.e. using your ability to hammer weak teams as the distinguishing factor.

The problem with 'matches won' is that you were playing a VP event, not trying to win individual matches.

I like the head-to-head match result when it's available. While it has the problems already pointed out, it does have the simplistic attraction of saying the better team out of A and B is the one that beat the other one. That somehow seems hard to argue with.

Failing that, I would much rather use swiss points (SOS) than imp quotient or total imps, because you are still using VPs are your discriminator, and it rewards a team who have had a lot of hard matches.

[pran]

 FrancesHinden, on 2011-August-29, 08:24, said:

None of these methods actually defines who is the 'better' contestant. The main form of scoring has decided that the two contestants are equal. Now you are coming up with something that is always going to be somewhat random.

The problem with imp quotient is at least four fold. First of all the teams won't know who has won (unless you are quoting it each round) and it feels a somewhat random differentiator. Secondly it is completely meaningless unless everyone has played the same boards (which is not the case in many North American swiss events). Thirdly the idea that conceding fewer imps is somehow 'better' doesn't actually hold up when you analyse it carefully. If I bid 10 vulnerable 4H contracts, each of them with a 40% chance of making exactly (and a 60% chance of going one off), four of them make and 6 go one off, and the other team are always in 3M (or 2M+1) then I will lose 36 imps and gain 40 imps. If the other team plays all of these contracts in partscore (but gained 4 imps some other way on one board) then their imp turnover is much much lower than mine, but I think my bidding is better. Fourthly, high variance bridge (leading to bid pluses and minuses) is not by definition worse than low variance bridge; all that matters is the expectation.

The problem with total (net) imps is that early matches are sometimes won very heavily (I have reasonably often won 8-board Swiss matches by 60+ imps) when a good team plays a very weak team. The VP scale cuts this huge margin off... but now you are reintroducing it to split a tie i.e. using your ability to hammer weak teams as the distinguishing factor.

The problem with 'matches won' is that you were playing a VP event, not trying to win individual matches.

I like the head-to-head match result when it's available. While it has the problems already pointed out, it does have the simplistic attraction of saying the better team out of A and B is the one that beat the other one. That somehow seems hard to argue with.

Failing that, I would much rather use swiss points (SOS) than imp quotient or total imps, because you are still using VPs are your discriminator, and it rewards a team who have had a lot of hard matches.

But at least we both show some interest in trying to find a criterion for deciding which team "deserves" to win the tie?

As I briefly mentioned IMP quotient was our main tie-breaking method until recently. Our current regulation includes variants of face to face result, swiss points, IMP quotient and/or total IMPs won, each alternative in situations where the method is defined suitable.

And BTW: Our competitons where tie-breaking might be needed are (well almost) always played with identical boards for all the contestants across the entire field during the same round.

[bluejak]

I think that any bridge event automatically and consistently decides who is the best contestant to be very naive.

I still think the advantages of total imps being part of the score very underrated. I agree it is not totally fair, but I expect nothing is, and there is always luck in bridge.