tiebreaker: imp difference or quotient

[geller]

At present the WBF uses imp quotient (imps won/imps lost) as the tiebreaker when vp scores are the same. Here in Japan we currently follow the WBF regulation for purposes of awarding trophies and prizes, etc., although masterpoint awards treat the tie as a tie.

 

Two teams in a Swiss Team event in Tokyo were tied for first place in VPs. Their scores were as follows

 

Team 1: IMP +291 -174

Quotient= 1.67 Net Plus= 117

 

Team 2: IMP +340 -212

Quotient= 1.60 Net Plus= 128

 

 

Under the current rules Team 1 was awarded the trophy, as it had the better IMP quotient. (On the other hand, if IMP difference=net plus= rather than quotient had been used as the tiebreaker, Team 2 would have been awarded the trophy.)

 

My question is, should imp quotient or imp difference (net plus) be used as the tiebreaker? It seems that a long time ago football (soccer) used goal ratio as the tiebreaker, and bridge followed this. Many years ago FIFA switched over from goal quotient to goal difference, but bridge continues to use imp quotient rather than difference. Either is somewhat arbitrary, but which do you think is better?

[Mbodell]

I think quotient is probably slightly better. It punishes the "swingy" team and rewards the steady team. I'm more impressed when someone holds their opponents to a really low score like 50-3 instead of if someone wins by more in a high scoring match like 217-149.

 

It is definitely arbitrary though, and in a swiss format I'd rather use some sense of strength of schedule or cumulative score or something (I.e., did you lose early and then win against the easy teams or did you win early). FIFA changed it to goal differential to encourage higher scoring and more offensive chances, no? Bridge definitely has regulations designed to go the other way (fewer high variance conventions allowed).

[dburn]

For Swiss Teams events, it seems better to use neither total IMPs nor IMP quotient. Instead, break ties by adding together the VP scores of all the teams against whom each tied team has played. The team whose opponents have the highest collective score is the winner, because they played against "more difficult" opponents.

[bluejak]

In England we use the result between the two teams first, failing that the number of rounds won, failing that 'Tie-break points' also known as 'Swiss Points'. This allows for the quality of opposition played.

 

If that fails, straight imps, then point-a-board, then toss a coin.

 

I have always liked the result between the two teams being the first tie-break. Notably they do that in Spanish football, though not elsewhere in football.

[dburn]

In England we use the result between the two teams first, failing that the number of rounds won, failing that 'Tie-break points' also known as 'Swiss Points'.  This allows for the quality of opposition played.

 

If that fails, straight imps, then point-a-board, then toss a coin.

What happened to the method whereby we determined which of two tied teams had been most fixed by the VP scale? It was a long time ago, but I still vividly remember the last qualifier for the Four Stars final at Brighton being manually calculated by that method, which for the uninitiated worked as follows:

 

One IMP was added to each team's score in every match, and the VP results were recalculated to produce a new ranking list - call this L1A.

 

Then, one IMP was subtracted from each team's score, and the VP results were recalculated again to produce list L1B.

 

Each team's VP scores from L1A and L1B were added, and if one team had a higher total than the other(s), that team was the winner.

 

Otherwise, two IMPs were added as above to produce L2A, two IMPs subtracted to produce L2B, and the preceding step was repeated for the new ranking lists.

 

Now that we have computers, it would take only a few minutes longer to split a tie by this method than it took Jim Proctor and Steve Barnfield working with pencil and paper all those years ago. And they had got to plus and minus 4 IMPs before the white smoke rose.

[pran]

In England we use the result between the two teams first, failing that the number of rounds won, failing that 'Tie-break points' also known as 'Swiss Points'.  This allows for the quality of opposition played.

 

If that fails, straight imps, then point-a-board, then toss a coin.

 

I have always liked the result between the two teams being the first tie-break.  Notably they do that in Spanish football, though not elsewhere in football.

We used to break ties with IMP quotient but I believe (so far I have not had any reason to check the current regulation on this question) that in Round Robin leagues we now break ties first by the winner of the match between the two tied teams.

 

There are two reasons why I am sceptic to this principle although admittedely it is both simple and more easily accepted by contestants in general.

 

Say that A and B are tied and A beat B in their match. This implies that A has scored less than B against the rest of the field. Is A better than B because A beat B or is B better than A because B scored better than A against the rest of the field (presumably the same total opposition)?

 

Now say that A, B and C are all tied, A beat B, B beat C and C beat A. Which contestant is the better?

 

IMP quotient has the advantage that one can always see who is ahead of tied teams, and by how much. Also as noted already it favours the "solid" teams that win matches with less total IMPs.

 

I don't think it is possible to show that one rule for breaking ties is "mathematically" better than another.

[mrdct]

The Australian method is:

At the conclusion of an event or stage with a Swiss format, ties are broken by VP of opponents then number of wins in VP then net IMP then lot.

[cherdanno]

I don't think it is possible to show that one rule for breaking ties is "mathematically" better than another.

Double.

[helene_t]

I don't think it is possible to show that one rule for breaking ties is "mathematically" better than another.

Double.

Well, if one is willing to make the right assumptions .....

 

But I think any sensible tie-break method is somewhere between 50% and 51% likely to get it right. So it doesn't matter in practice. Except maybe for very short matches where there is a considerable chance of a tie.

[Free]

What about quotient * net plus or something similar.

 

Bridge is a game of winning imps and losing imps. If you win 1000 imps and lose 900, have you played better than a team who won 100 imps and lost 1? Imo no, so net plus alone is not good because here the team with 1000 imps would win.

What about team 1 winning 400-201 and team 2 winning 200-100? It's clear that team 1 should be winning (imo) because they were able to score a lot more while not giving away too much. Say they played 10 matches, then they averaged a 20 imp difference per game, while team 2 only managed an average of 10 imp difference. But the quotient says team 2 should win.

 

Now consider: quotient * net plus = tiebreaker

In my first example, we'd get team 1 = 111.11 ; team 2 = 9900. Team 2 is a clear winner here.

In my second example, we'd get team 1 = 396.02 ; team 2 = 200. Team 1 is the winner here.

 

This is a very small test with big differences which are probably not realistic. However I think the method has logic behind it, so it might have merrit. In the borderline case above, team 2 would win.

 

Possible modifications could be to use quotient*sqrt(net plus) or sqrt(quotient)*net plus or something similar to have more or less weight for certain parameters.

[bluejak]

I do not agree with the word "should" in the last post.

 

What about team 1 winning 400-201 and team 2 winning 200-100? It's clear that team 1 should be winning (imo) ...

If the rules say team 2 has won, so what? Bridge is not about winning imps, bridge is about winning using the form of scoring in use. You can play a great imp game and get nowhere in a pairs tournament.

 

I do not think that there is a "should". I think you just work something out, publish it, and follow it. Personally, I feel the match between the two teams is a good tie-break and pran's logic leaves me cold. I do not believe it is a matter of logic.

 

To be honest, for Swiss Teams specifically, I like publishing results with VPs and total imps. That way you only have to break a tie when the VPs and the imps are both the same, and that will be incredibly rare. Furthermore, players understand it easily.

[gwnn]

I think the method that explores how much you've been 'fixed by the vp scale' is very cute and simple and probably quite fair. I am not sure if imp difference or quotient is a nicer way of scoring, in practice it shouldn't matter that much. Sure 400:201 vs 200:100 sounds odd but so would 20:10 vs 1000:989.

[duschek]

Using the result between the two (or more) teams as the primary tie-breaking method has one obvious advantage over any other method: The players know the odds instantly, even before the last round (provided of course that the teams met).

 

Opponents' VPs is fine, but has one drawback in that a 25-5 VP win is now better than a 25-0 VP win. Apart from that, surely players understand that scoring 190 VPs against good opposition is better than scoring 190 VPs against weak opposition.

 

Imp difference is very natural, the main problem being a matter of calculation. Players are interested in VPs, not imp differences (except when calculating VPs). Also, slaughtering a weak team works very positively for your imp difference, but does not say as much about your strength as does beating a better team.

 

Imp quotient feels strange indeed. Imp matches and VP scales are all about imp differences and not at all about quotients. A 50% chance of winning 13 imps at the risk of losing 13 imps is at break-even. Imp quotient now says that it is good bridge to make that random proposition at neither or both tables, and bad bridge to make it at either table but not both. Players never think in terms of imp quotients, so using imp quotient for tie-breaking seems very much an academical exercise.

[Free]

I do not agree with the word "should" in the last post.

 

What about team 1 winning 400-201 and team 2 winning 200-100? It's clear that team 1 should be winning (imo) ...

If the rules say team 2 has won, so what? Bridge is not about winning imps, bridge is about winning using the form of scoring in use. You can play a great imp game and get nowhere in a pairs tournament.

 

I do not think that there is a "should". I think you just work something out, publish it, and follow it. Personally, I feel the match between the two teams is a good tie-break and pran's logic leaves me cold. I do not believe it is a matter of logic.

 

To be honest, for Swiss Teams specifically, I like publishing results with VPs and total imps. That way you only have to break a tie when the VPs and the imps are both the same, and that will be incredibly rare. Furthermore, players understand it easily.

The form of scoring presented by the OP was imps, so guess what, this game was in fact about winning imps... So FYI we're talking about tiebrakers in imp games.

 

That being said, the rest of your post seems like a whining old man. The rules apply, ofcourse, but that doesn't mean nobody can question them and try to come up with something better. The weakest argument I've heard lots of times, and here again: "that's the way it is". Personally I have no problem with ordering pairs according to VP scale and imp difference/total/whatever.

 

But if we're looking for "the perfect tiebraker" - if there is one - I think it will be more complicated that method 1 vs method 2. The 2 extreme cases I presented were illustrations as to what I think is best, ignoring current rules, to come up with a theoretical solution. If everyone agrees that a team scoring 100-1 is better than a team scoring 1000-900, then the rule should take that into consideration. Same goes for the quotient, although that's a more difficult case imo.

[CamHenry]

Opponents' VPs is fine, but has one drawback in that a 25-5 VP win is now better than a 25-0 VP win. Apart from that, surely players understand that scoring 190 VPs against good opposition is better than scoring 190 VPs against weak opposition.

My understanding is that it's opponents' VPs against the rest of the field. That means 25-5 is no different from 25-0 for calculating that team's weighting for you, but makes them contribute more to the tie-break for other teams. In short, 25-0 remains better than 25-5, and I think that's one of the few areas everyone will agree is desirable! :rolleyes:

[campboy]

The problem, in my opinion, with using the head-to-head results as a tie-breaker is that adding extra teams to the tie-break changes the result -- even if A would win a tie-break between A and B it is possible for B to win a tie-break between A, B and C. Methods which are determined by an absolute statistic for each team do not suffer this problem; if A has a better IMP ratio than B it is irrelevant how many other teams are involved in the tie, A will always have a better IMP ratio than B.

 

Number of rounds won seems like a good first tie-break to me. Ok, it will often be inconclusive but it takes no time to calculate and since the masterpoint awards (at least in the EBU) already focus on winning matches it shouldn't involve any change of strategy.

[bluejak]

The form of scoring presented by the OP was imps, so guess what, this game was in fact about winning imps...  So FYI we're talking about tiebrakers in imp games.

Perhaps you should take the trouble to read the OP. The form of scoring was VPs, not imps.

 

That being said, the rest of your post seems like a whining old man.  The rules apply, ofcourse, but that doesn't mean nobody can question them and try to come up with something better.  The weakest argument I've heard lots of times, and here again: "that's the way it is".  Personally I have no problem with ordering pairs according to VP scale and imp difference/total/whatever.

Thanks for your charming comments. I like to do the best I can for the players, and suggestions as to what is best are just that.

 

But if we're looking for "the perfect tiebraker" - if there is one - I think it will be more complicated that method 1 vs method 2.  The 2 extreme cases I presented were illustrations as to what I think is best, ignoring current rules, to come up with a theoretical solution.  If everyone agrees that a team scoring 100-1 is better than a team scoring 1000-900, then the rule should take that into consideration.  Same goes for the quotient, although that's a more difficult case imo.

Now that is where you are completely wrong. What is best for the players is what is seen by the players as best, not some academic exercise little understood by them and is thus not a perfect solution.

[Free]

The form of scoring presented by the OP was imps, so guess what, this game was in fact about winning imps...  So FYI we're talking about tiebrakers in imp games.

Perhaps you should take the trouble to read the OP. The form of scoring was VPs, not imps.

~snip~

Oh riiiiiiight, sorry, my mistake! You look at the results of each table and from there you calculate the VP score immediately. :)

 

Me and the rest of the world look at the difference of total points per board, convert it to imps, calculate the imp totals of each team, calculate the difference, and then convert that to a VP score. If the total VP's are the same between 2 teams then they're irrelevant. In that case it's probably most logical to look at some difference in the imp scores (quotient, difference, imps won,...).

 

Are you a lawyer who needs every word and letter spelled correctly, or can you read between the lines? (alert! retorical question! please don't answer!) Welcome to the "CHANGING Laws & Regulations" forum btw...

[blackshoe]

Leave the sarcasm at home, please.

[bluejak]

The form of scoring presented by the OP was imps, so guess what, this game was in fact about winning imps...  So FYI we're talking about tiebrakers in imp games.

Perhaps you should take the trouble to read the OP. The form of scoring was VPs, not imps.

~snip~

Oh riiiiiiight, sorry, my mistake! You look at the results of each table and from there you calculate the VP score immediately. :)

 

Me and the rest of the world look at the difference of total points per board, convert it to imps, calculate the imp totals of each team, calculate the difference, and then convert that to a VP score. If the total VP's are the same between 2 teams then they're irrelevant. In that case it's probably most logical to look at some difference in the imp scores (quotient, difference, imps won,...).

 

Are you a lawyer who needs every word and letter spelled correctly, or can you read between the lines? (alert! retorical question! please don't answer!) Welcome to the "CHANGING Laws & Regulations" forum btw...

No, I am someone whose prime aim is to do the best for the players, which is why I have suggested practical solutions.

 

As for the rest of the world, you seem to have no idea what they want: they want something easy, obvious, transparent. You are the one suggesting lawyer-type solutions. I disapprove strongly of solutions that only are believed correct by lawyers, which is why my solutions are the exact opposite of lawyerly solutions. To me, tie-break points and imp quotients are lawyer-type solutions, so I do not approve.

[pran]

My humble experience is that players here have been very happy with IMP quotient as the tie-breaking factor.

 

Now that we have "who won the match between tied contestants" I am just awaiting the first event when among three tied teams each team beat one and lost the other tie-braking match.

 

(In Swiss events we still maintain the calculation of which team had the stronger opposition.)