Butler scoring

[Jennifer Murphy]

In another thread, Lorne suggest Butler scoring for our little
three-table social pairs events. Having never heard of this method, I
did a little research. I came up with a few interesting discussions,
which I'll post below.

Detailed analysis by David Stevenson:
http://blakjak.org/butler.htm

Hilarious discussion:
http://forums.contractbridge.net/showthread.php?tid=116

Scoring software, which I am evaluating:
http://homepages.nildram.co.uk/~jasmith/

General explanation:
http://www.bridgewebs.com/northfield/news.html

Advantages & disadvantages:
http://bridgeclubsites.blogspot.com/2006/01/try-butler-scoring.html

Simple explanation:

http://www.mrbridge.co.uk/forum/index.php?p=/discussion/45/butler-scoring/p1

I have a couple of questions:


1. Most of the Butler discussions recommend discarding the highest and
lowest score and averaging the rest. If we only have three tables, that
would leave us with just one score to "average". Should I use all three
scores?


2. Are the EW scores equal to the NW scores subtracted from zero:

EW = 0 - NW

Suppose we have a board that is a marginal 4S contract for NS. No one
vulnerable. Let's say one team bids and makes, one team underbids and
makes, and one team bids and goes down. Here are the MP scores:


Pair Contract By Vul Made NS EW NS-MP EW-MP
1-2 4S N None 4 420 0 4 0
3-4 3S N None 4 170 0 2 2
5-6 4S N None -1 0 50 0 4

If I average the NS scores (420, 170, -50), I get 540 / 3 = 180.

To get the Butler scores, I subtract 180 from each NS score giving me:

Pair Butler IMPs
1 240.0 6
3 -10.0 0
5 -230.0 -6

Are the EW scores just the reverse?

Pair Butler IMPs
2 -240.0 -6
4 10.0 0
6 230.0 6

[p***@gmail.com]

Post by Jennifer Murphy
1. Most of the Butler discussions recommend discarding the highest and
lowest score and averaging the rest. If we only have three tables, that
would leave us with just one score to "average". Should I use all three
scores?

Using the average will give you less variance in the scores. Using the median (one score) will result in more "swingy" -10 and 10 scores. I dunno. You might refer to records and see which method gives more pleasing results.

Post by Jennifer Murphy
Are the EW scores just the reverse?

Yes. I'm not willing to check the math, but I'm about 90% sure.

[Dave Flower]

In another thread, Lorne suggest Butler scoring for our little three-table social pairs events. Having never heard of this method, I did a little research. I came up with a few interesting discussions, which I'll post below. Detailed analysis by David Stevenson: http://blakjak.org/butler.htm Hilarious discussion: http://forums.contractbridge.net/showthread.php?tid=116 Scoring software, which I am evaluating: http://homepages.nildram.co.uk/~jasmith/ General explanation: http://www.bridgewebs.com/northfield/news.html Advantages & disadvantages: http://bridgeclubsites.blogspot.com/2006/01/try-butler-scoring.html Simple explanation: http://www.mrbridge.co.uk/forum/index.php?p=/discussion/45/butler-scoring/p1 I have a couple of questions: 1. Most of the Butler discussions recommend discarding the highest and lowest score and averaging the rest. If we only have three tables, that would leave us with just one score to "average". Should I use all three scores? 2. Are the EW scores equal to the NW scores subtracted from zero: EW = 0 - NW Suppose we have a board that is a marginal 4S contract for NS. No one vulnerable. Let's say one team bids and makes, one team underbids and makes, and one team bids and goes down. Here are the MP scores: Pair Contract By Vul Made NS EW NS-MP EW-MP 1-2 4S N None 4 420 0 4 0 3-4 3S N None 4 170 0 2 2 5-6 4S N None -1 0 50 0 4 If I average the NS scores (420, 170, -50), I get 540 / 3 = 180. To get the Butler scores, I subtract 180 from each NS score giving me: Pair Butler IMPs 1 240.0 6 3 -10.0 0 5 -230.0 -6 Are the EW scores just the reverse? Pair Butler IMPs 2 -240.0 -6 4 10.0 0 6 230.0 6

I have two issues with Butler scoring:

1) It compares scores with an artificial datum. This can be avoided by using cross-IMPs, where ones scores are compared separately with every other time the board is played. Perfectly feasible with computer scoring.

2) The boards do not have equal weight. This is quite significant if players do not play all the boards. Making an overtrick in 3NT at matchpoints may gain a top, whereas at Butler it will gain a paltry 1 IMP

Dave Flower

[Mark Brader]

...

Post by Dave Flower
2) The boards do not have equal weight. This is quite significant if
players do not play all the boards. Making an overtrick in 2NT at
matchpoints may gain a top, whereas at Butler it will gain a paltry 1
IMP

Yes, that's the *idea*.

[Barry Margolin]

Post by Mark Brader
...

Post by Dave Flower
2) The boards do not have equal weight. This is quite significant if
players do not play all the boards. Making an overtrick in 2NT at
matchpoints may gain a top, whereas at Butler it will gain a paltry 1
IMP

Yes, that's the *idea*.

That's true of IMP Pairs in general, whether scored Butler or Cross-IMPs.

[Dave Flower]

Yes, that's what I was saying - two separate issues

Dave Flower

[Jennifer Murphy]

On Mon, 15 Apr 2013 03:56:23 -0700 (PDT), Dave Flower

In another thread, Lorne suggest Butler scoring for our little three-table social pairs events. Having never heard of this method, I did a little research. I came up with a few interesting discussions, which I'll post below. Detailed analysis by David Stevenson: http://blakjak.org/butler.htm Hilarious discussion: http://forums.contractbridge.net/showthread.php?tid=116 Scoring software, which I am evaluating: http://homepages.nildram.co.uk/~jasmith/ General explanation: http://www.bridgewebs.com/northfield/news.html Advantages & disadvantages: http://bridgeclubsites.blogspot.com/2006/01/try-butler-scoring.html Simple explanation: http://www.mrbridge.co.uk/forum/index.php?p=/discussion/45/butler-scoring/p1 I have a couple of questions: 1. Most of the Butler discussions recommend discarding the highest and lowest score and averaging the rest. If we only have three tables, that would leave us with just one score to "average". Should I use all three scores? 2. Are the EW scores equal

to

the NW scores subtracted from zero: EW = 0 - NW Suppose we have a board that is a marginal 4S contract for NS. No one vulnerable. Let's say one team bids and makes, one team underbids and makes, and one team bids and goes down. Here are the MP scores: Pair Contract By Vul Made NS EW NS-MP EW-MP 1-2 4S N None 4 420 0 4 0 3-4 3S N None 4 170 0 2 2 5-6 4S N None -1 0 50 0 4 If I average the NS scores (420, 170, -50), I get 540 / 3 = 180. To get the Butler scores, I subtract 180 from each NS score giving me: Pair Butler IMPs 1 240.0 6 3 -10.0 0 5 -230.0 -6 Are the EW scores just the reverse? Pair Butler IMPs 2 -240.0 -6 4 10.0 0 6 230.0 6
1) It compares scores with an artificial datum. This can be avoided by using cross-IMPs, where ones scores are compared separately with every other time the board is played. Perfectly feasible with computer scoring.

I suppose an average is somewhat "artificial", but it is also a
well-established statistical measure. Why is it a less palatable
comparison reference than the aggregate of the actual scores, which are,
presumably, just data points on some continuum, for which that average
is a well-regarded metric?

Do you have a statistical reason for preferring a cross-IMP type of
averaging, and that is what it is, over the Butler method?

2) The boards do not have equal weight. This is quite significant if players do not play all the boards.

Any rotation in which the players do not all play the same boards has
flaws.

Making an overtrick in 3NT at matchpoints may gain a top, whereas at Butler it will gain a paltry 1 IMP

Are you saying this is a strength or a weakness? It seems to me to be a
strength.

It also seems to me that both the Butler methpd and the Cross-IMP method
would produce very similar results in most situations. It would be
enlightening to see some moderately large tournament scored both ways
(and also at MPs) to see the differences.

[Steve Willner]

Post by Jennifer Murphy
I suppose an average is somewhat "artificial", but it is also a
well-established statistical measure. Why is it a less palatable
comparison reference than the aggregate of the actual scores,

Butler over-emphasizes boards with bimodal scores. That's not its worst
problem, but it's already enough to make me prefer cross-IMPs.

[Barry Margolin]

Post by Steve Willner

Post by Jennifer Murphy
I suppose an average is somewhat "artificial", but it is also a
well-established statistical measure. Why is it a less palatable
comparison reference than the aggregate of the actual scores,

Butler over-emphasizes boards with bimodal scores. That's not its worst
problem, but it's already enough to make me prefer cross-IMPs.

The other argument often given against Butler scoring is that the IMP
table was designed to match common score differences, such as 420 vs 170
(bidding game versus not bidding game). When you IMP against an
average, the score differences don't fall into the same scales that the
IMP table was designed for.

[Jennifer Murphy]

Post by Barry Margolin

Post by Steve Willner

Post by Jennifer Murphy
I suppose an average is somewhat "artificial", but it is also a
well-established statistical measure. Why is it a less palatable
comparison reference than the aggregate of the actual scores,

Butler over-emphasizes boards with bimodal scores. That's not its worst
problem, but it's already enough to make me prefer cross-IMPs.

The other argument often given against Butler scoring is that the IMP
table was designed to match common score differences, such as 420 vs 170
(bidding game versus not bidding game).

How does the IMP table design "match" common score differences? A score
of 420 is the top end of the 9 IMP range; whereas, 170 is at the bottom
of the 5 IMP range. That's a 4 IMP difference.

Post by Barry Margolin
When you IMP against an
average, the score differences don't fall into the same scales that the
IMP table was designed for.

I don't see that as significant. If I have Excel fit a curve to the IMP
scale (with the middle of the range as the independent variable), I get
a very good fit (R=0.9988) to a third degree polynomial:

y = 0.3284x^3 - 3.0698x^2 + 37.163x

The IMP scale is a way of smoothly and proportionaly damping down the
range of the raw rubber bridge scores: 0-4000+ is mapped onto 0-24.

In any case, when you average IMPs across multiple tables, I bet any
"matching" effect to common scores is lost. In fact, I'd expect that the
results from scoring a pairs event with the two methods would be very
similar and closedr to each other than either would be to MP scoring.

If I had some raw data, I might just write a program to score it various
ways and compare the results.

[Jennifer Murphy]

Post by Steve Willner

Post by Jennifer Murphy
I suppose an average is somewhat "artificial", but it is also a
well-established statistical measure. Why is it a less palatable
comparison reference than the aggregate of the actual scores,

Butler over-emphasizes boards with bimodal scores.

Do you have any data to illustrate this?

Post by Steve Willner
That's not its worst
problem, but it's already enough to make me prefer cross-IMPs.

What is its worst problem?

[Dave Flower]

Consider a board played 11 times. At all the tables, the bidding is 1NT-3NT, and the opening leader has to lead from:

Q J 10 9
Q J 10 9
x x x
x x

as it happens a heart lead beats the contract, a spade does not.

Assume the board has been played ten times, and 3NT has been made (vul) on five of them. If the board is scored Butler, everybody who played the board will stand to gain or lose 12IMPs, depending on the opening lead at the last table. Is this desireable?

Dave Flower

[Barry Margolin]

Post by Dave Flower

wrote: >On 2013-04-15 8:09 PM, Jennifer Murphy wrote: >> I suppose an
average is somewhat "artificial", but it is also a >> well-established
statistical measure. Why is it a less palatable >> comparison reference
than the aggregate of the actual scores, > >Butler over-emphasizes boards
with bimodal scores. Do you have any data to illustrate this? >That's not
its worst >problem, but it's already enough to make me prefer cross-IMPs.
What is its worst problem?

Consider a board played 11 times. At all the tables, the bidding is 1NT-3NT,
Q J 10 9
Q J 10 9
x x x
x x
as it happens a heart lead beats the contract, a spade does not.
Assume the board has been played ten times, and 3NT has been made (vul) on
five of them. If the board is scored Butler, everybody who played the board
will stand to gain or lose 12IMPs, depending on the opening lead at the last
table. Is this desireable?

Before the 11th table plays, the datum is 0. The pairs that made the
contract get +9 IMPs (400-0), the pairs that went down get -2 IMPs
(-50-0).

After the 11th table plays, the two possibilities are:

They make the contract. The datum is 400/11 = 36. The pairs that made
game get 8 IMPs (400-36 = 362), the pairs that went down get -3 IMPs
(-50-36 = -86). So everyone else's score goes down by 1 IMP, because one
more pair was successful.

They go down. The datum is now -36. The pairs that made game get 10
IMPs, the pairs that went down get -1 IMP. So everyone else's score goes
up by 1 IMP.

There's no 12-IMP change, just a 1-IMP shift one way or the other. Note
also that either way, the difference in scores between the pairs that
made and went down is 11 IMPs.

That 11 IMP difference is the reason why Butler is not as good as
cross-IMPs. IMPs were created for team games. The general idea of IMP
pairs is that it simulates a set of team matches, where you have
teammates at each of the other tables playing the board. Cross-IMPs then
gives you the same comparison with each other table that you would in a
team game -- if you make game, you get 10 IMPs for every table where
your "teammates" set them.

[Travis Crump]

Post by Barry Margolin

Post by Dave Flower

wrote: >On 2013-04-15 8:09 PM, Jennifer Murphy wrote: >> I suppose an
average is somewhat "artificial", but it is also a >> well-established
statistical measure. Why is it a less palatable >> comparison reference
than the aggregate of the actual scores, > >Butler over-emphasizes boards
with bimodal scores. Do you have any data to illustrate this? >That's not
its worst >problem, but it's already enough to make me prefer cross-IMPs.
What is its worst problem?

Consider a board played 11 times. At all the tables, the bidding is 1NT-3NT,
Q J 10 9
Q J 10 9
x x x
x x
as it happens a heart lead beats the contract, a spade does not.
Assume the board has been played ten times, and 3NT has been made (vul) on
five of them. If the board is scored Butler, everybody who played the board
will stand to gain or lose 12IMPs, depending on the opening lead at the last
table. Is this desireable?

Before the 11th table plays, the datum is 0. The pairs that made the
contract get +9 IMPs (400-0), the pairs that went down get -2 IMPs
(-50-0).

Wouldn't the datum be 250[or 175 NV]? Unless of course you are assuming
a heart lead leads to down 6[down 8 NV] which while possible seems
unlikely. So Vul it would be gain or lose 8[or 6 NV].

[Travis Crump]

Post by Travis Crump

Post by Barry Margolin

Post by Dave Flower

wrote: >On 2013-04-15 8:09 PM, Jennifer Murphy wrote: >> I suppose an
average is somewhat "artificial", but it is also a >> well-established
statistical measure. Why is it a less palatable >> comparison reference
than the aggregate of the actual scores, > >Butler over-emphasizes boards
with bimodal scores. Do you have any data to illustrate this? >That's not
its worst >problem, but it's already enough to make me prefer cross-IMPs.
What is its worst problem?

Consider a board played 11 times. At all the tables, the bidding is 1NT-3NT,
Q J 10 9
Q J 10 9
x x x
x x
as it happens a heart lead beats the contract, a spade does not.
Assume the board has been played ten times, and 3NT has been made (vul) on
five of them. If the board is scored Butler, everybody who played the board
will stand to gain or lose 12IMPs, depending on the opening lead at the last
table. Is this desireable?

Before the 11th table plays, the datum is 0. The pairs that made the
contract get +9 IMPs (400-0), the pairs that went down get -2 IMPs
(-50-0).

Wouldn't the datum be 250[or 175 NV]? Unless of course you are assuming
a heart lead leads to down 6[down 8 NV] which while possible seems
unlikely. So Vul it would be gain or lose 8[or 6 NV].

Whereas at Cross-IMPs, each table that made it would gain 60 Vul, or 6.6
if you divide by the number of comparisons which I think is fairly common.

[Barry Margolin]

Post by Travis Crump

Post by Barry Margolin

Post by Dave Flower

wrote: >On 2013-04-15 8:09 PM, Jennifer Murphy wrote: >> I suppose an
average is somewhat "artificial", but it is also a >> well-established
statistical measure. Why is it a less palatable >> comparison reference
than the aggregate of the actual scores, > >Butler over-emphasizes boards
with bimodal scores. Do you have any data to illustrate this? >That's not
its worst >problem, but it's already enough to make me prefer cross-IMPs.
What is its worst problem?

Consider a board played 11 times. At all the tables, the bidding is 1NT-3NT,
Q J 10 9
Q J 10 9
x x x
x x
as it happens a heart lead beats the contract, a spade does not.
Assume the board has been played ten times, and 3NT has been made (vul) on
five of them. If the board is scored Butler, everybody who played the board
will stand to gain or lose 12IMPs, depending on the opening lead at the last
table. Is this desireable?

Before the 11th table plays, the datum is 0. The pairs that made the
contract get +9 IMPs (400-0), the pairs that went down get -2 IMPs
(-50-0).

Wouldn't the datum be 250[or 175 NV]? Unless of course you are assuming
a heart lead leads to down 6[down 8 NV] which while possible seems
unlikely. So Vul it would be gain or lose 8[or 6 NV].

Oops, you're right. I stupidly forgot that making and going down don't
cancel each other out. But it's still the case that one additional table
doesn't change the datum by very much.

Before: datum is (400*5 + -50*5) / 10 = 175
After win: datum is (400*6 + -50*5) / 11 = 195
After loss: datum is (400*5 + -50*6) / 11 = 155

So either way the datum changes by 20, which will cause at most a 1 IMP
shift in Butler scores.

[Steve Willner]

SW> Butler over-emphasizes boards with bimodal scores.

Post by Jennifer Murphy
Do you have any data to illustrate this?

Isn't it obvious from the non-linearity of the IMP scale? Consider a
board with equal numbers of results of +600 and -100. In normal team
scoring, being on one side or the other is 12 IMPs. In Butler, the
datum is 250, and being on one side or the other is a swing of 16 IMPs.

Post by Jennifer Murphy
What is [Butler's] worst problem?

That improving your result on a board can hurt your standing in the
event. This is usually only noticed when there's a score correction,
but it can happen any time. (Bastille fixes this.)

Not far behind is that pairs are donated or deprived of random amounts
of IMPs depending on what seat they happen to sit in.

Cross-IMPs solves both of these automatically and also fixes the bimodal
scores problem. Its only real disadvantage is that it's slightly harder
to score than cross-IMPs, but that's trivial if there aren't many tables
or if you score by computer. Your definition of "many" will depend on
how handy you are at scoring, but three tables won't be a problem unless
you are seriously number-challenged.

SW> If you want Butler, just use the middle score as the datum.

Post by Jennifer Murphy
This counter-intuitive for me.

What you have to understand is that Butler is a "quick and dirty" method
to score IMP pairs when cross-IMPs is too much trouble. Nothing about
Butler is mathematically sound. Given that reality, any recommendation
of how many scores to throw out in creating the datum is arbitrary.
With three scores, using the median instead of the arithmetic mean is
simpler and little if at all worse. It has the benefit the median is a
real bridge score, not some arbitrary number.

Post by Jennifer Murphy
Whereas at Cross-IMPs, each table that made it would gain 60 Vul, or 6.6
if you divide by the number of comparisons which I think is fairly common.

Travis is right that dividing by number of comparisons is common;
ACBLscore does that. If each board is played the same number of times,
the divisor won't matter. However, if some boards are played more often
than others (either because of the movement or because some get fouled),
and *if you believe in Neuberg for matchpoints*, the divisor should be
the number of pairs, i.e., one more than the number of comparisons.
Dividing by comparisons is equivalent to factoring at matchpoints.

None of this is new. We've had long threads with detailed explanations
in the past.

[paul]

Post by Jennifer Murphy
In another thread, Lorne suggest Butler scoring for our little
three-table social pairs events...

Our club plays an IMP Pair game once a month I think (on a Wednesday,
when I'm generally at work.) I think they use the datum method, but I
really see no reason not to use cross-IMP's: you get your IMP's vs.
every other pair who plays a hand your way. I also agree with those
who say IMP pairs (I've never heard the name Butler outside of this
newsgroup) is not inherently better than matchpoints for a pairs game:
it makes a lot of hands meaningless and makes a few slam or game hands
swing the results. In a three-table game you're playing five boards
vs. each pair, which is closer to Swiss team conditions, so IMP pairs
seem reasonable.

[Michael Angelo Ravera]

Post by Jennifer Murphy
In another thread, Lorne suggest Butler scoring for our little
three-table social pairs events. Having never heard of this method, I
did a little research. I came up with a few interesting discussions,
which I'll post below.
http://blakjak.org/butler.htm
http://forums.contractbridge.net/showthread.php?tid=116
http://homepages.nildram.co.uk/~jasmith/
http://www.bridgewebs.com/northfield/news.html
http://bridgeclubsites.blogspot.com/2006/01/try-butler-scoring.html
http://www.mrbridge.co.uk/forum/index.php?p=/discussion/45/butler-scoring/p1
1. Most of the Butler discussions recommend discarding the highest and
lowest score and averaging the rest. If we only have three tables, that
would leave us with just one score to "average". Should I use all three
scores?
EW = 0 - NW
Suppose we have a board that is a marginal 4S contract for NS. No one
vulnerable. Let's say one team bids and makes, one team underbids and
Pair Contract By Vul Made NS EW NS-MP EW-MP
1-2 4S N None 4 420 0 4 0
3-4 3S N None 4 170 0 2 2
5-6 4S N None -1 0 50 0 4
If I average the NS scores (420, 170, -50), I get 540 / 3 = 180.
Pair Butler IMPs
1 240.0 6
3 -10.0 0
5 -230.0 -6
Are the EW scores just the reverse?
Pair Butler IMPs
2 -240.0 -6
4 10.0 0
6 230.0 6

Average IMPs is best for small events. Sometimes I use IMPs against the Double Dummy Par Score result.

[Jennifer Murphy]

On Mon, 15 Apr 2013 17:42:22 -0700 (PDT), Michael Angelo Ravera

Post by Michael Angelo Ravera

Post by Jennifer Murphy
In another thread, Lorne suggest Butler scoring for our little
three-table social pairs events. Having never heard of this method, I
did a little research. I came up with a few interesting discussions,
which I'll post below.

Average IMPs is best for small events. Sometimes I use IMPs against the Double Dummy Par Score result.

Why do you think so?

Is this the same as the Cross-IMPs method recommended by Dave Flower?

[Michael Angelo Ravera]

Post by Jennifer Murphy
On Mon, 15 Apr 2013 17:42:22 -0700 (PDT), Michael Angelo Ravera

Post by Michael Angelo Ravera

Post by Jennifer Murphy
In another thread, Lorne suggest Butler scoring for our little
three-table social pairs events. Having never heard of this method, I
did a little research. I came up with a few interesting discussions,
which I'll post below.

Average IMPs is best for small events. Sometimes I use IMPs against the Double Dummy Par Score result.

Why do you think so?
Is this the same as the Cross-IMPs method recommended by Dave Flower?

Average IMPs is effectively the same as Cross-IMPs ,if everybody plays the same number of boards and all of the boards get played the same number of times and Cross-IMPs saves a division, if you are scoring manually.

However, if you ever have to throw out a board, can't play it due to time constraints, find out that two tables played it one way and the third some way different, you will want to have a way to fix it. Average IMPs give you a way. (Two tables that played the board the same way get IMP'd against each other and the other two pairs get a 0, +3, or -3, or the average of all of their other scores, according to why it is unplayable).

Another reason for average IMPs is that it puts results into a normal IMP range (-24.00 to +24.00). The result is like playing a Team game with the field as teammates.

[Steve Willner]

Post by Jennifer Murphy
In another thread, Lorne suggest Butler scoring for our little
three-table social pairs events.

If you want Butler, just use the middle score as the datum. Butler has
some problems -- not least that improving your score on one board may
harm your placing in the event -- but the method is easy to understand
and works reasonably well.

However, with only three tables, it's almost as easy to use cross-IMPs
as Butler. Just IMP each score against the two others and add the IMPs.
I've cross-IMPed 7- or 8-table games by hand, and it's not that hard.
Three should be easy even if you are not all that fast with the
pencil. Cross-IMPs eliminates all the problems with Butler.

As others have written, an IMP pairs game will be more random than
matchpoints with the same number of boards, but that's part of the fun.

[Jennifer Murphy]

Post by Steve Willner

Post by Jennifer Murphy
In another thread, Lorne suggest Butler scoring for our little
three-table social pairs events.

If you want Butler, just use the middle score as the datum.

This counter-intuitive for me. The "literature" recommends tossing the
top & bottom 10% of the scores. One site even offered a formula for
doing fractional discards. Eliminating 2/3 of the scores seems draconian
and more than triple the recommended 10%.

In any case, I would think we would want all of the scores. If those
outliers are anomolies, they will occur rarely, and should not be too
much of a problem. If they are not, then they are good data.

A much better solution, I would think, would be to have some data about
the shape of the score distribution curve -- which may well vary by the
skew of the points or cards in the hands. If we had that, we could
"adjust" a set of actual data to more closely fit the typical
distribution and then calculate a Butler score.

Ah, so many formulas, so little time.

Post by Steve Willner
Butler has
some problems -- not least that improving your score on one board may
harm your placing in the event -- but the method is easy to understand
and works reasonably well.

I read that complaint a couple of times, but no one gave an example and
I couldn't think of one.

Post by Steve Willner
However, with only three tables, it's almost as easy to use cross-IMPs
as Butler.

I'd say it might be easier. Just subtract and lookup. No averaging.

I'd just set up a little spreadsheet for either.

Post by Steve Willner
Just IMP each score against the two others and add the IMPs.
I've cross-IMPed 7- or 8-table games by hand, and it's not that hard.
Three should be easy even if you are not all that fast with the
pencil. Cross-IMPs eliminates all the problems with Butler.
As others have written, an IMP pairs game will be more random than
matchpoints with the same number of boards, but that's part of the fun.

My question is which is a more accurate assessment of the quality of the
play.

[p***@gmail.com]

On Mon, 15 Apr 2013 21:28:36 -0400, Steve Willner

Post by Steve Willner

Post by Jennifer Murphy
In another thread, Lorne suggest Butler scoring for our little
three-table social pairs events.

If you want Butler, just use the middle score as the datum.

This counter-intuitive for me. The "literature" recommends tossing the
top & bottom 10% of the scores. One site even offered a formula for
doing fractional discards. Eliminating 2/3 of the scores seems draconian
and more than triple the recommended 10%.

Yes, I too would go for the average. With only three boards you aren't going to get stability no matter what you do, so there is no need to worry about it much. I think the fractional discards wouldn't be worth bothering with.

[f***@googlemail.com]

Butler has >some problems -- not least that improving your score on one board may >harm your placing in the event -- but the method is easy to understand >and works reasonably well.

I read that complaint a couple of times, but no one gave an example and I couldn't think of one.


See, for example, http://www.bridgeguys.com/sec/glossary/b/bastille_movement.html which explains how this cost a top pair some substantial prize money.
(Note I am not recommending the Bastille method described here, but it explains the problem)

[Lorne]

Post by Jennifer Murphy
In another thread, Lorne suggest Butler scoring for our little
three-table social pairs events. Having never heard of this method, I
did a little research. I came up with a few interesting discussions,
which I'll post below.
http://blakjak.org/butler.htm
http://forums.contractbridge.net/showthread.php?tid=116
http://homepages.nildram.co.uk/~jasmith/
http://www.bridgewebs.com/northfield/news.html
http://bridgeclubsites.blogspot.com/2006/01/try-butler-scoring.html
http://www.mrbridge.co.uk/forum/index.php?p=/discussion/45/butler-scoring/p1
1. Most of the Butler discussions recommend discarding the highest and
lowest score and averaging the rest. If we only have three tables, that
would leave us with just one score to "average". Should I use all three
scores?
EW = 0 - NW
Suppose we have a board that is a marginal 4S contract for NS. No one
vulnerable. Let's say one team bids and makes, one team underbids and
Pair Contract By Vul Made NS EW NS-MP EW-MP
1-2 4S N None 4 420 0 4 0
3-4 3S N None 4 170 0 2 2
5-6 4S N None -1 0 50 0 4
If I average the NS scores (420, 170, -50), I get 540 / 3 = 180.
Pair Butler IMPs
1 240.0 6
3 -10.0 0
5 -230.0 -6
Are the EW scores just the reverse?
Pair Butler IMPs
2 -240.0 -6
4 10.0 0
6 230.0 6

Whatever you do there will be those who can point out something that
looks unfair or something they do not like. My view is that the average
player finds it tough to accept that an overtrick can be worth more than
bidding a slam or that playing a suspect 4-3 fit in a major is better
than a solid 4-4 fit in a minor or that risking your contract for a 60%
overtrick can be the "correct" play so going to Butler in a small field
where the overtrick is often a top/bottom decision with MP scoring is
likely to be more acceptable because the big swings can only be through
big penalties, or bidding slams/games others miss. People seem to
accept they deserve a big swing on slam/game decisions but not on
overtrick decsions.

I think your best strategy is to tell the group you are going to play 5
sessions scored as Butler IMPs as an experiment. Explain to them they
are playing teams with team-mates that always bring back the average
score made by the 3 pairs sitting the other way, then after 5 sessions
ask them how they like it and go with whatever response you get.

Alternatively do the same with cross-IMPs if you like that better (I
personally prefer Butler having played a lot at a club that does it once
a week, but others have described the good/bad points of both methods).
Do not get too complex and try both unless the disussion after 5
sessions suggests they want to do that.

[Steve Willner]

Post by Lorne
Alternatively do the same with cross-IMPs if you like that better (I
personally prefer Butler having played a lot at a club that does it once
a week, but others have described the good/bad points of both methods).

I'm curious what you think are the good points of Butler, other than its
relative simplicity of calculation. Or is that enough to sway you, even
against the obvious disadvantages?

What do you say to the pair (Hamman-Wolff, I think most famously), who
say "If only we hadn't scored that overtrick in 2D, we'd have placed
higher in the event?"

[Lorne]

Post by Steve Willner

Post by Lorne
Alternatively do the same with cross-IMPs if you like that better (I
personally prefer Butler having played a lot at a club that does it once
a week, but others have described the good/bad points of both methods).

I'm curious what you think are the good points of Butler, other than its
relative simplicity of calculation. Or is that enough to sway you, even
against the obvious disadvantages?
What do you say to the pair (Hamman-Wolff, I think most famously), who
say "If only we hadn't scored that overtrick in 2D, we'd have placed
higher in the event?"

I said what I thought were the plus points in the part of my post you
removed. Basically with 3 tables you will get volatile results whatever
you do but with Butler the big swings revolve around game/slam bidding
and big penalties which most players accept deserve big swings and not
around the odd overtrick.

What happened to Hamman/Wolff was a rarity. I played Butler IMPs almost
every Friday for 10+ years without ever noticing any anomoly like that.
In MP's you could equally have a situation where they say if only I
had not tried for the 75% overtrick on the last board.

[f***@googlemail.com]

On 20/04/2013 02:10, Steve Willner wrote: > On 2013-04-17 6:18 AM, Lorne wrote: >> Alternatively do the same with cross-IMPs if you like that better (I >> personally prefer Butler having played a lot at a club that does it once >> a week, but others have described the good/bad points of both methods). > > I'm curious what you think are the good points of Butler, other than its > relative simplicity of calculation. Or is that enough to sway you, even > against the obvious disadvantages? > > What do you say to the pair (Hamman-Wolff, I think most famously), who > say "If only we hadn't scored that overtrick in 2D, we'd have placed > higher in the event?" > I said what I thought were the plus points in the part of my post you removed. Basically with 3 tables you will get volatile results whatever you do but with Butler the big swings revolve around game/slam bidding and big penalties which most players accept deserve big swings and not around the odd overtrick. What happened to Hamman/Wolff was a rarity. I played Butler IMPs almost every Friday for 10+ years without ever noticing any anomoly like that. In MP's you could equally have a situation where they say if only I had not tried for the 75% overtrick on the last board.

You are comparing Butler IMPs to matchpoints.
Steve (and others, including me) think that Ximps is superior to Butler imps and cannot see any disadvantages.

[Michael Angelo Ravera]

Post by f***@googlemail.com

On 20/04/2013 02:10, Steve Willner wrote: > On 2013-04-17 6:18 AM, Lorne wrote: >> Alternatively do the same with cross-IMPs if you like that better (I >> personally prefer Butler having played a lot at a club that does it once >> a week, but others have described the good/bad points of both methods). > > I'm curious what you think are the good points of Butler, other than its > relative simplicity of calculation. Or is that enough to sway you, even > against the obvious disadvantages? > > What do you say to the pair (Hamman-Wolff, I think most famously), who > say "If only we hadn't scored that overtrick in 2D, we'd have placed > higher in the event?" > I said what I thought were the plus points in the part of my post you removed. Basically with 3 tables you will get volatile results whatever you do but with Butler the big swings revolve around game/slam bidding and big penalties which most players accept deserve big swings and not around the odd overtrick. What happened to Hamman/Wolff was a rarity. I played Butler IMPs almost every Friday for 10+ years without ever noticing any anomoly like that. In MP's you could equally have a situation where they say if only I had not tried for the 75% overtrick on the last board.

You are comparing Butler IMPs to matchpoints.
Steve (and others, including me) think that Ximps is superior to Butler imps and cannot see any disadvantages.

I'm with you. I always use either "Cross-IMP average" or "Against the Double Dummy Par Score".

[Lorne]

Post by f***@googlemail.com

On 20/04/2013 02:10, Steve Willner wrote: > On 2013-04-17 6:18 AM, Lorne wrote: >> Alternatively do the same with cross-IMPs if you like that better (I >> personally prefer Butler having played a lot at a club that does it once >> a week, but others have described the good/bad points of both methods). > > I'm curious what you think are the good points of Butler, other than its > relative simplicity of calculation. Or is that enough to sway you, even > against the obvious disadvantages? > > What do you say to the pair (Hamman-Wolff, I think most famously), who > say "If only we hadn't scored that overtrick in 2D, we'd have placed > higher in the event?" > I said what I thought were the plus points in the part of my post you removed. Basically with 3 tables you will get volatile results whatever you do but with Butler the big swings revolve around game/slam bidding and big penalties which most players accept deserve big swings and not around the odd overtrick. What happened to Hamm

an/Wolff was a rarity. I played Butler IMPs almost every Friday for 10+ years without ever noticing any anomoly like that. In MP's you could equally have a situation where they say if only I had not tried for the 75% overtrick on the last board.

Post by f***@googlemail.com
You are comparing Butler IMPs to matchpoints.
Steve (and others, including me) think that Ximps is superior to Butler imps and cannot see any disadvantages.

I did not realise that was the comparison. I do not have a strong view
about cross IMPs to Butler, my slight preference for Butler is probably
due to the fact that is what I played at the Young Chelsea for many
years so it is familiar.

[Steve Willner]

What happened to Hamman/Wolff was a rarity. I played Butler IMPs almost every Friday for 10+
years without ever noticing any anomoly like that.

Did you look? I'd expect it to be fairly common -- far from a "rarity"
-- but only noticed if there's a scoring correction.

I'm happy to play IMP pair games; our club schedules them a few times a
year. No problem with that. The issue is Butler scoring specifically.

As to familiarity as a reason to use Butler rather than cross-IMPs, I
can sort of sympathize. All of us tend to prefer what is familiar.
Being mathematically inclined myself, though, I think the flaws of
Butler are so obvious that I can't understand why anyone would prefer it
unless ease of scoring makes it the only practical choice.