Distribution of scores in duplicate

[Jennifer Murphy]

Has anyone done a study or a mathematical (or statistical) analysis of
the scores in a duplicate event?

In our little three-table social event, people are always complaining
that they got a low score for the right contract, well played, because
the opponents of their opponents screwed up by (a) playing poor defense,
(b) failing to double a bad sacrifice, or (c) making an outrageous bid
that got doubled for a larger score than the right contract.

Clearly things like that happen and they happen in our group more than
in most. But I'd live to be able to cite statistics or calculate
probabilities that the final scores for the night are the result of bad
luck vs bad play.

[Herb]

In a three table game the scores will be much more significantly skewed
than in a larger game. One off-the-wall result can change a pair's score
from a 2 or 1½ to a 0 on the board, or a 75-100% shift. In a 10-table
game, that same result would only change other pairs' scores by 10% or so.

- Herb

[dake50]

I track my % of "field" loses (don't find obvious games, obvious defenses, obvious defensive bids). That % has been a low of 17% to a high of 27%; near 20% seems typical. Our club has usually 7 tables, a small game with 40 of 50 players not much above beginner.
So I would expect a "we just have bad luck" complaint to be 20% true. The rest due to missed inferences.

[Jennifer Murphy]

Right. So what I am looking for is a formula or, better yet, a
spreadsheet, that takes as inputs,

T = the number of tables
B = the number of boards
S = my score total

and, maybe,

L = average skill level of the players

That last variable may be the most significant. Weak players make more
mistakes, which should result in a larger standard deviation and
probably a lower mean.

[Jennifer Murphy]

On Tue, 2 Apr 2013 15:36:10 -0700 (PDT), dake50 <dak...@aol.com> wrote:

>I track my % of "field" loses (don't find obvious games, obvious defenses, obvious defensive bids). That % has been a low of 17% to a high of 27%; near 20% seems typical. Our club has usually 7 tables, a small game with 40 of 50 players not much above beginner.
>So I would expect a "we just have bad luck" complaint to be 20% true. The rest due to missed inferences.

The thing the whiners never mention is the times when they got a top
because the other two pairs sitting *their* direction screwed up or when
they bid a game they shouldn't have and made it for a top due to poor
defense. ;-)

I think that the unlucky low scores are more or less offset by the
unlucky tops. For awhile, I secretly tracked the scores from our group.
The top scorers were more or less same pairs each month as were the low
scorers.

[psug...@gmail.com]

I think that the hand record analysis should include an indication of "luckiness" -- e.g. what percentage of the time contracts "make when they shouldn't" or "go down when they 'should make'". This would be based on statistics like # of tricks compared to total points; etc. This, in turn, is an indicatio0n of the percentage of finesses that are "on", etc. If "most people" would make a particular opening lead against the most common contract, that should be included in the analysis.

[derek]

On Tuesday, April 2, 2013 5:55:48 PM UTC-3, Jennifer Murphy wrote:
>
> In our little three-table social event, people are always complaining
> that they got a low score for the right contract, well played, because
> the opponents of their opponents screwed up by (a) playing poor defense,
> (b) failing to double a bad sacrifice, or (c) making an outrageous bid
> that got doubled for a larger score than the right contract.

I'm not quite sure what Dake's 20% means, but having this happen once or twice a session certainly seems normal in club games. More so on BBO, less at Regional and higher tournaments (our sectionals don't seem much less random than our club games, even in flighted events). So, I'd mostly put it down to random chance.

[Will in New Haven]

Just last night our opponents bid 7NT, which was cold, when other
pairs got to inferior contracts, bad luck. However, declarer had his
brain explode and left six good Hearts on the board, wandering back to
his hand where he had three losers, good luck to us.

On the companion board, they cashed their tricks properly and left me
with no option but a finesse that lost. Several pairs in the field
managed to crash tricks or set up a discard for declarer, bad luck to
us as the eight tricks I took, which is all one could take, gave us a
poor, although not zero, score.

--
Will in New haven

[france...@googlemail.com]

On Tuesday, April 2, 2013 9:55:48 PM UTC+1, Jennifer Murphy wrote:
> Has anyone done a study or a mathematical (or statistical) analysis of the scores in a duplicate event? In our little three-table social event, people are always complaining that they got a low score for the right contract, well played, because the opponents of their opponents screwed up by (a) playing poor defense, (b) failing to double a bad sacrifice, or (c) making an outrageous bid that got doubled for a larger score than the right contract. Clearly things like that happen and they happen in our group more than in most. But I'd live to be able to cite statistics or calculate probabilities that the final scores for the night are the result of bad luck vs bad play.

First the non-useful answer:
-----------------------------

All the things you are describing here are both bad luck and bad play, it just depends on your perspective. The only difference is that "your" bad play is someone else's bad luck.

If a pair at another table defends poorly and hence you get a bad score, that is your bad luck and their bad play. If you defend poorly and present your opponents with a good result, that is your bad play and the other declarers' (at the other tables) bad luck.

So you can't really do the sort of analysis you are suggesting.

Genuine "bad luck" is if you bid well to a good slam that goes off when trumps are 5-0, but the other tables stopped in game. That really is bad luck for you and good luck for your opponents. That always happens, but if you play enough times and over enough evenings it will even itself out with the equivalent bits of good luck.

Now the useful answer
-----------------------
A one-evening 3.5 table duplicate pairs will be somewhat random, because one funny score has such a big effect on everyone else. Also as you have already metnioned, the lower the overall standard the more random the results become. And the variance will be different for different pairs, because some are more 'random' than others - if you have a good player with a novice, then the 'novice' will present a lot of gifts while the good player won't, so it will depend how many hands the novice gets to play (for example).

My personal experience (without a detailed analysis) is that when I play a 24-board evening duplicate with my usual partner at my local club, about 11-15 tables, then
- the majority of the time I get a percentage within the range 58-65%, never lower, with the occasional score in the high 60s or low 70s
- usually we have 'hurt' our result by between 4 and 10% by doing things that are (on reflection) obviously wrong i.e. our bad play
- there's usually another 10% we could do nothing about i.e. our opponents doing something much better than at other tables, or we did something good but were genuinely unlucky. This number varies from about 0 to 20%.

But the really high scores are not when we avoid these two factors, they are more when we are given more gifts by opponents being more stupid than usual.

[vsp...@hotmail.com]

The standard deviations for matchpoint
results is approximately 28% per board.
There's a lot of luck in this game.
Good pairs overcome this. 'Bad' pairs
use this as an excuse to whine.

[Stu Goodgold]

On Wednesday, April 3, 2013 9:02:07 AM UTC-7, vsp...@hotmail.com wrote:
> On Tuesday, April 2, 2013 1:55:48 PM UTC-7, Jennifer Murphy wrote:
>
> > Has anyone done a study or a mathematical (or statistical) analysis of
>
> >
>
> > the scores in a duplicate event?

>
>

> The standard deviations for matchpoint
> results is approximately 28% per board.
> There's a lot of luck in this game.
>
> Good pairs overcome this. 'Bad' pairs
> use this as an excuse to whine.

Where do you get this 28% number?

I would expect the variance depends on a number of factors:

number of tables
number of boards played
skill level of the field (both average and skew)

2 days ago we had a 14 table club game which was a club championship with overall MP awards. I pointed out before the start that the 5 best pairs were all sitting EW, with a major skill advantage over the over pairs. One of our grand lifemasters also noted that. The director agreed but reluctantly didn't want to switch any NS pairs, who reserve their seats. Sure enough, a B-level NS pair finished 1st NS and 2nd overall. So not only does the average skill level of the field matter, it matters how the skill level is skewed.

-Stu Goodgold
San Jose, CA

[vsp...@hotmail.com]

On Wednesday, April 3, 2013 10:49:42 AM UTC-7, Stu Goodgold wrote:
> On Wednesday, April 3, 2013 9:02:07 AM UTC-7, vsp...@hotmail.com wrote:
>
> > On Tuesday, April 2, 2013 1:55:48 PM UTC-7, Jennifer Murphy wrote:
>
> >
>
> > > Has anyone done a study or a mathematical (or statistical) analysis of
>
> >
>
> > >
>
> >
>
> > > the scores in a duplicate event?
>
>
>
> >
>
> >
>
> > The standard deviations for matchpoint
>
> > results is approximately 28% per board.
>
> > There's a lot of luck in this game.
>
> >
>
> > Good pairs overcome this. 'Bad' pairs
>
> > use this as an excuse to whine.
>
>
>
> Where do you get this 28% number?
>

Just made the calcs on many boards from
BBO minis. Most boards were played 50+
times.

>
>
> I would expect the variance depends on a number of factors:
>
>
>
> number of tables
>
> number of boards played
>
> skill level of the field (both average and skew)
>

The variance can be lowered if and only if there
are many ties. Or few unique results. If played
50+ times and there are over 30 unique results,
that would create a high variance. If played
50+ times and only 5 or fewer unique results,
that would be a low variance.
Have never seen the results from a quality field.
Are any of those events played over the internet?

[Barry Margolin]

In article <f1129cc7-8228-4925...@googlegroups.com>,

vsp...@hotmail.com wrote:

> Have never seen the results from a quality field.
> Are any of those events played over the internet?

On the 28th of every month BBO hosts a "Royal Tournament", which only
allows players with over 1,000 BBO masterpoints to play. Although we
all know the limitations of masterpoints as ratings, this may be the
closest you'll get to that.

You can see the report on the March tournament here:

http://webutil.bridgebase.com/v2/news_fetch.php?id=856

You can then find the play records by looking up one of the players in
BBO's myhands site.

--
Barry Margolin
Arlington, MA

[spenser]

In article <barmar-15BA88....@news.eternal-september.org>,

Barry,

That's probably the closest you'll find in online competition, but I
would argue that certain events (e.g. Blue Ribbon Pairs, Platinum
Pairs, Life Masters Pairs held at the NABCs) would comprise a set of
"quality fields" that surpass that of the Royal Tournament :)

--
Dennis Cohen

[Barry Margolin]

In article <030420131558442149%s...@dogbreath.com>,

Of course. But he asked "are any of those events played over the
Internet", and none of those are.

However, ACBL now posts the recap sheet for all NABC events on their
website, so he should be able to calculate statistics for these events.

[vsp...@hotmail.com]

The recaps aren't ordered. It takes like forever.
On the internet the scores are ordered and I can
sit at home to make the calcs.
But now that I looked at some boards the s.d was
more like 28% to 31%. By the nature of the statistic
it is unlikely a quality field would lower this stat
more than 3%. Rarely do boards have only two outcomes
with over 75% of the pairs receiving the same outcome.

[Travis Crump]

What do you mean by variance? Here is what I think of:

Consider a board played 3 times. One person bids the obvious game and
takes the obvious 10 tricks for +420, one person's opponent's take an
ill advised save for +1100, and one person takes a losing practice
finesse for -50. We have 3 unique results, but no variance. Everyone
receives exactly the matchpoints they were expecting.

Now consider another board. All three pairs overbid to 4S and get
doubled. The defense for all three pairs conspires in idiotic fashion
to blow two tricks on defense allowing it to make for +590. Now despite
only 1 unique result there is a large variance[50%], each pair expects a
cold top, but instead receives only an average.

[Michael Angelo Ravera]

On Wednesday, April 3, 2013 10:49:42 AM UTC-7, Stu Goodgold wrote:

It's amazing that a simple arrow switch for one or two rounds would fix all of this (and your director wouldn't do it)! For 14 tables (13 rounds), you'd like to switch 23 of the table-rounds. That would be tables 6-14 on the penultimate round and all of the tables on the last round. For 12 rounds, you'd like to switch 21 table-rounds. It would be the even (or high half) tables on the penultimate round and everybody on the last.

The matchpoint percentage standard deviation on a 3-table game would be
sqrt(2*.5*.5) = 0.71 matchpoints ~= 35%, if there were no ties. A 17 table game with no ties would have a SD of about sqrt(16*.5*.5) matchpoints ~= 24%. The number of ties would bring it down somewhat. I suspect that a higher skill level would increase the number of ties. Of course, if you add up 25 of these and divide by 25, the SD drops by a factor of 5. This means that a bit more than 95% of the pairs (give or take) won't have a 60% (or 40%) game in 17-table games. But that means that, on the average, about one and a half pairs WILL have a 60% matchpoint score in a 17-table game. The same 95% should be true of a 64% matchpoint score in a 3-table game, but that means that you have to hold about 3 3-table games before even one pair would be expected to have a 64% game.

[patmp...@gmail.com]

I'm a statistician. Variance is defined as the square of the standard deviation. It's a number that is calculated by a standard formula.

Starting with 28% the standard deviation of the average score over 25 boards is 28%/sqrt(25) = 5.6%. Over one hundred boards the standard deviation of the average score is 2.8%. So you can see why serious bridge tournaments are so lengthy.

[patmp...@gmail.com]

On Thursday, April 4, 2013 11:28:08 AM UTC+8, Michael Angelo Ravera wrote:
>
> The matchpoint percentage standard deviation on a 3-table game would be
>
> sqrt(2*.5*.5) = 0.71 matchpoints ~= 35%, if there were no ties. A 17 table game with no ties would have a SD of about sqrt(16*.5*.5) matchpoints ~= 24%. The number of ties would bring it down somewhat. I suspect that a higher skill level would increase the number of ties. Of course, if you add up 25 of these and divide by 25, the SD drops by a factor of 5. This means that a bit more than 95% of the pairs (give or take) won't have a 60% (or 40%) game in 17-table games. But that means that, on the average, about one and a half pairs WILL have a 60% matchpoint score in a 17-table game. The same 95% should be true of a 64% matchpoint score in a 3-table game, but that means that you have to hold about 3 3-table games before even one pair would be expected to have a 64% game.

This analysis is fine as it goes, but it is assuming that all pairs have an equal level of skill. In actual practice you would expect more variation.

[Jennifer Murphy]

On Wed, 3 Apr 2013 11:30:15 -0700 (PDT), vsp...@hotmail.com wrote:

... snip ...

>> > The standard deviations for matchpoint
>> > results is approximately 28% per board.
>> > There's a lot of luck in this game.
>>

>> Where do you get this 28% number?
>>
>Just made the calcs on many boards from
>BBO minis. Most boards were played 50+
>times.

What is the raw data and how are the calculations done?

[Michael Angelo Ravera]

Actually, it assumes that everyone would have no skill at all and that people would simply flip coins and randomly score better or worse than each of the other pairs. Skill of any kind would actually DECREASE this variance. Ties would decrease the variance also.

[Jennifer Murphy]

On Wed, 03 Apr 2013 23:12:00 -0400, Travis Crump
<pret...@techhouse.org> wrote:

...snip...

>> The recaps aren't ordered. It takes like forever.
>> On the internet the scores are ordered and I can
>> sit at home to make the calcs.
>> But now that I looked at some boards the s.d was
>> more like 28% to 31%. By the nature of the statistic
>> it is unlikely a quality field would lower this stat
>> more than 3%. Rarely do boards have only two outcomes
>> with over 75% of the pairs receiving the same outcome.
>
>What do you mean by variance?

The variance of a distribution is a measure of how spread out the values
are.

Suppose I run some test e times and get 3 results each time. Let's say
the results are:

1. 100, 100, 100
2. 75, 100, 125
3. 50, 100, 150

All three sets of data have the same mean (100), but very different
variances (or standard deviations, which is just the square root of the
variance).

For the data above, the variances are:

1. 0
2. 625
3. 2500

>Here is what I think of:

>Consider a board played 3 times. One person bids the obvious game and
>takes the obvious 10 tricks for +420, one person's opponent's take an
>ill advised save for +1100, and one person takes a losing practice
>finesse for -50. We have 3 unique results, but no variance. Everyone
>receives exactly the matchpoints they were expecting.
>
>Now consider another board. All three pairs overbid to 4S and get
>doubled. The defense for all three pairs conspires in idiotic fashion
>to blow two tricks on defense allowing it to make for +590. Now despite
>only 1 unique result there is a large variance[50%], each pair expects a
>cold top, but instead receives only an average.

I'm not sure what you are calculating. It seems to have something to do
with the players' variable expectations. I'm looking at the scores as
compared to each other to try and figure out how much "luck" or "chance"
there is that a pair will not get a top if they bid and play correctly.

For your first board, the raw scores are 1100, 420, -50. When they are
converted to MPs, they become 4, 2, 0, which has a variance of 4.

For your second board, the raw scores are 590, 590, 590, so, as you say,
the MPs are 2, 2, 2 for a variance of zero.

[patmp...@gmail.com]

Hmm. If all the pairs have exactly the same level of skill then each match would depend entirely on luck, with each side having the same chance of winning, namely 50%. So from a statistical point of view the results would be the same as a coin flipping session.

You folks seem to be using variance in some sense that is different than the technical sense in which mathematicians use it. That's fine, but I don't understand it.

[Jennifer Murphy]

On Wed, 3 Apr 2013 20:28:08 -0700 (PDT), Michael Angelo Ravera
<mara...@prodigy.net> wrote:

...snip...

>The matchpoint percentage standard deviation on a 3-table game would be
>sqrt(2*.5*.5) = 0.71 matchpoints ~= 35%, if there were no ties.

Can you help me understand how this is derived?

And I don't understand the "no ties" restriction.

>A 17 table game with no ties would have a SD of about sqrt(16*.5*.5) matchpoints ~= 24%.

I don't understand what you are calculating and, for that matter, how
you can calculate it "theoretically". It seemsd to me that the variance
of a set of data requires a calculation on the datual data points.

If the data points themselves can be calculated, like coin tosses, then
the "expected" SD could be calculated absent any actual data.

>The number of ties would bring it down somewhat.

Why?

>I suspect that a higher skill level would increase the number of ties.

Yes, I would think so.

[Michael Angelo Ravera]

If Matchpoints, for each board played 11 times were uniform at discrete integers for 0-10, it seems like the SD is around 33%. The SD of a distribution like this seems asptmitodically to approach 28.91%. Centrally placed ties would reduce this considerably.

[Jennifer Murphy]

Mr/Ms Statistician (We now have a matrimonial-neutral title, do we have
a gender-neutral one? Mx?):

Can you help me with my probably poorly-worded question?

I would like to know how much luck or "chance" there is in duplicate
bridge, MP scoring. That is, what are the odds that I will not get a top
if I (and my partner) bid and play correctly?

It seems to me that this is not something that can be done theoretically
or in the abstract. It needs actual data. Is that correct?

If I knew the variance (or SD) of

[Jennifer Murphy]

On Thu, 4 Apr 2013 01:01:36 -0700 (PDT), patmp...@gmail.com wrote:

Damn...hit the wrong key by mistake. Trying again:

Mr/Ms Statistician (We now have a matrimonial-neutral title, do we have
a gender-neutral one? Mx?):

Can you help me with my probably poorly-worded question?

I would like to know how much luck or "chance" there is in duplicate
bridge, MP scoring. That is, what are the odds that I will not get a top
if I (and my partner) bid and play correctly?

It seems to me that this is not something that can be done theoretically
or in the abstract. It needs actual data. Is that correct?

Suppose I had the actual scores from a large duplicate event --
something like 100 scores for each board. Is it true that the variance
(or SD) of those scores is a measure of the randomness of the scores? If
all of the scores were the same, the variabce would be zero, which would
seem to me to indicate zero randomness or chance.

In a 3-table event, the mean of the MP scores is always 2 ((0+2+4)/3).
The mean of the MPs for 20 boards would be 40 (2 x 20). Right?

Now if I knew the SD of those MP scores, could I calculate the odds that
a particular score was skill vs luck?

Maybe I am asking how to calculate the 95% or 90% confidence interval
(of something) around a score. In a 3-table game, a top score is 4. The
maximum possible score in a 20-board event would be 20 x 4 = 80. If I
knew the SD of the scores for those 20 boards (or for 20 average
boards), could I say anything about my score? Say I got a 45. An average
score is 40. Can I say anything about the odds that not getting a higher
score, say a 50, is bad luck vs bad play?

I think I am wording this badly. It's probably because I'm not sure what
I am trying to get. Can you read my mind? ;-)

[Michael Angelo Ravera]

On Thursday, April 4, 2013 1:56:47 AM UTC-7, Jennifer Murphy wrote:
> On Wed, 3 Apr 2013 20:28:08 -0700 (PDT), Michael Angelo Ravera
>

<MAR>
With no ties, it's binomial and the variance is simply npq. This is a worst case assumption based upon equal chances of higher or lower. If there is some possibility of a "the same", the difference from the mean for each of those is zero and this also decreases the number of more extreme values.

For instance, let's say that one-half of the scores are ties and one-fourth are randomly higher and one-fourth lower. What happens to the variance? I beleive that this cuts the variance by half and thus the SD by around 29%.

In another post, I assumed that the Matchpoints were discrete and uniform and got an asptmtotic limiting SD of about 29% of a top per board.

</MAR>

[patmp...@gmail.com]

On Thursday, April 4, 2013 5:17:49 PM UTC+8, Jennifer Murphy wrote:
>
>
>
> In a 3-table event, the mean of the MP scores is always 2 ((0+2+4)/3).
>
> The mean of the MPs for 20 boards would be 40 (2 x 20). Right?
>
>

Right.

>
> Now if I knew the SD of those MP scores, could I calculate the odds that
>
> a particular score was skill vs luck?
>
>

Yes.

The idea is that luck tends to cancel out over a large number of trials, while skill does not. So if the winner of a tournament squeaks by then that could be just luck. Or if the number of boards is small even a strong win could be just luck. Or if the number of entrants is very high then a high score could be just luck. On the other hand if a pair gives everybody a thumping over a large number of boards then the chance of luck might shrink to such a small number that we have to conclude that the result is due to skill. This is all common sense, statistics gives you a precise picture of how much these factors affect things.

It gets tricky. Let's say you have a long tournament with a thousand entrants. That is so many that it is pretty likely that someone will get lucky and run up a high score by chance. But if before the tournament began you picked a single pair to win and they did, then that would be convincing evidence that they really were the best. Statistics lets you calculate the odds in both of those cases.

The standard deviation or variance is essentially a measure of how slowly luck cancels out. The other thing is how close together in skill the pairs are. If one pair is clearly superior that should show quickly. If one pair is 1% better than another you would have to play a great many boards to cancel out the luck factor.

[patmp...@gmail.com]

A complication is that in a duplicate tournament the scores are NOT independent. If Gazzaro-Fishbein gets a very high score then Karo-Kahn gets a lower one. So the number of trials before you get a statistically significant result goes up. But I wouldn't worry too much about that.

[Michael Angelo Ravera]

I believe that Statisticians are referred respectfully and gender neutrally as as "Dr", "Prof", "Master", or "Your most exhaulted omniscience" depending their training and how badly you need a correct answer.

"Correctly" will never fit into it. You can, of course, compare your results to the double-dummy analysis with a general declarer bias (somewhere between one-third and a whole trick per board depending upon skill level and contract level) and get a Win-Lose-Draw from that.

You can replay the boards in a nationally ranked event and see how you did compared to those in the event.

Depending upon which calculation you believe, you get somewhere around 30% of a top on each board as the standard deviation. Over 25 boards, this overall standard deviation drops to 6% of all of the matchpoints available. Two Standard deviations of that make it a 95% (two tailed) probability that such a difference wasn't due to luck.

In club games, we consider it especially noteworthy that someone scores 65% or better at matchpoints or 2.5IMP/Bd or better at IMPs. At our club, which has about 700 games per year, the best game of the year usually ends up at about 77%(although I had the game of the year once with 75%, and by January 15 the next year, 75% had already been beaten 4 times.)

I've heard it said that there are 4 ways to get a top:
1) You do the right thing and it works out for you
2) You do the wrong this, but it works out for you
3) Your opponents do the wrong thing and it works to your advantage
4) Your opponents do the right thing, but it goes bad for them.

On a rerun of an old Charles Goren program (that I saw in the early 1970s but am pretty sure was records 15 years earlier), I heard him say "There is often more than one way to skin a cat and always many ways not to".

[Douglas Newlands]

I'm not a statistician but isn't that formula for the same board being
played 100 times (or whatever) and not for 100 separate hands each
played once?
I thought that if the Std Dev was 's' then the std dev of all scores
at the end of n boards was sqrt(n)*s
I could be entirely wrong of course :(

I thought the reason serious teams were long was that the variance in
scores after n boards was sqrt(n)*6 (IIRC the 6 came from a study of
European teams championship matches) and that at that level an imp a
board is a thrashing even though elsewhere 2 imps a board might be
easily possible.
So if team A is 1 imp per board better than Team B then over 20 boards
the std dev is sqrt(20)*6 ~= 30 so A's natural advantage is not quite
1 std deviation so they are probably 6:4 favourites.
If the match is 100 boards then the std dev is sqrt(100)*6 = 60
but the 100 imp natural advantage is nearly 2 std devs so the odds
on A winning must be approx 85:15 (sorry no tables or calculator
to hand).

Can one of the statisticians sort this please?

doug

[bhmwe...@gmail.com]

Den onsdagen den 3:e april 2013 kl. 17:44:44 UTC+2 skrev france...@googlemail.com:
> But the really high scores are not when we avoid these two factors, they are more when we are given more gifts by opponents being more stupid than usual.

Another way to benefit from good luck is what happened to the club's best pair, who managed to get almost 80% in a normal club evening. Their good luck was that they had a bye during the first round. On board 1 all tables except two played 4S +2, and on board 2 ALL tables played 4S +1. Instead of scoring two average boards they scored 80% on the round...

[vsp...@hotmail.com]

VAR(X)=E(x²)-[E(x)]²

Nowadays you can use EXCEL.
Enter the board outcome results into a column.
On the next empty cell, click the 'fx' function.
Click 'Statistical', find and click 'VAR'.
Under number 1 enter the range of the location
of the outcome results. Use the ':' between
the beginning and end of the range.
Then on the next empty cell
=SQRT(VAR cell)

[patmp...@gmail.com]

On Thursday, April 4, 2013 8:17:55 PM UTC+8, Douglas, wrote:
>
>
>
> I'm not a statistician but isn't that formula for the same board being
>
> played 100 times (or whatever) and not for 100 separate hands each
>
> played once?

It could be done either way. As a statistician the first thing you do is choose the population you are sampling from. In this case the population is the set of all sets of scores resulting from the playing of boards. Each set of scores from the playing of a board is a member of the population. It is a bit unusual.

>
> I thought that if the Std Dev was 's' then the std dev of all scores
>
> at the end of n boards was sqrt(n)*s
>

That is the std dev of the sum of the scores. I was talking about the std dev of the average of the scores. Divide by n to get s/sqrt(n). So they are directly related.

> I could be entirely wrong of course :(
>
>
>
> I thought the reason serious teams were long was that the variance in
>
> scores after n boards was sqrt(n)*6 (IIRC the 6 came from a study of
>
> European teams championship matches) and that at that level an imp a
>
> board is a thrashing even though elsewhere 2 imps a board might be
>
> easily possible.
>

You are talking IMPs. We were talking percentages. I think you are confusing variance with standard deviation. The variance is the square of the standard deviation.

> So if team A is 1 imp per board better than Team B then over 20 boards
>
> the std dev is sqrt(20)*6 ~= 30 so A's natural advantage is not quite
>
> 1 std deviation so they are probably 6:4 favourites.

>
> If the match is 100 boards then the std dev is sqrt(100)*6 = 60
>
> but the 100 imp natural advantage is nearly 2 std devs so the odds
>
> on A winning must be approx 85:15 (sorry no tables or calculator
>
> to hand).
>
>

There is no simple way to calculate the odds. (The calculation called a convolution, for good reason.) I feel certain that the better team has odds more favorable than those. There must be a table for this somewhere, but it's late.

[Adam Beneschan]

On Thursday, April 4, 2013 2:02:38 AM UTC-7, Jennifer Murphy wrote:

> Mr/Ms Statistician (We now have a matrimonial-neutral title, do we have
> a gender-neutral one? Mx?):

What would that be short for? "Mixture?"

"There was a conversation between Mixture Ravera and Mixture Murphy about ..." ...

Ummm, maybe. Although it does occur to me that "Mixture" might be an appropriate choice for referring to transgenders....

I'd better stop now before I get myself into any more trouble.

-- Adam

[Travis Crump]

On 04/04/2013 04:47 AM, Jennifer Murphy wrote:

> I'm not sure what you are calculating. It seems to have something to do
> with the players' variable expectations. I'm looking at the scores as
> compared to each other to try and figure out how much "luck" or "chance"
> there is that a pair will not get a top if they bid and play correctly.
>
> For your first board, the raw scores are 1100, 420, -50. When they are
> converted to MPs, they become 4, 2, 0, which has a variance of 4.
>
> For your second board, the raw scores are 590, 590, 590, so, as you say,
> the MPs are 2, 2, 2 for a variance of zero.

I guess another way of looking at it is that you are doing a survey of
comparisons. If you play 25 boards with 2 comparisons each, then 95%
confidence would be plus or minus 14%. On the other hand if you had a
26 table game, you'd play 26 boards with 25 comparisons then 95%
confidence would be plus or minus 3.8%. 13 tables would be 5.5%. etc.

[france...@googlemail.com]

"I would like to know how much luck or "chance" there is in duplicate bridge, MP scoring. That is, what are the odds that I will not get a top if I (and my partner) bid and play correctly? "

If there is no luck, then you will get an average, not a top, if you and your partner play correctly, because everyone will do exactly the same as you and everyone will get exactly 50%. In order to get a top, you need 'luck' in that you need someone at another table, or your opponents, to bid and play incorrectly.

[Travis Crump]

On a per board basis, with 2 comparisons with 99% confidence the error
is 91%. If you've ever looked at a bidding contest in the Bridge World
or similar, it is pretty much impossible to do better than 91% on your
own. 90% confidence is 58%. There is hardly any confidence that a
normal 50% board won't score as either 100% or 0%.

[Adam Lea]

On 03/04/13 00:13, Jennifer Murphy wrote:
> On Tue, 2 Apr 2013 15:36:10 -0700 (PDT), dake50<dak...@aol.com> wrote:
>
>> I track my % of "field" loses (don't find obvious games, obvious defenses, obvious defensive bids). That % has been a low of 17% to a high of 27%; near 20% seems typical. Our club has usually 7 tables, a small game with 40 of 50 players not much above beginner.
>> So I would expect a "we just have bad luck" complaint to be 20% true. The rest due to missed inferences.
>
> The thing the whiners never mention is the times when they got a top
> because the other two pairs sitting *their* direction screwed up or when
> they bid a game they shouldn't have and made it for a top due to poor
> defense. ;-)
>
> I think that the unlucky low scores are more or less offset by the
> unlucky tops. For awhile, I secretly tracked the scores from our group.
> The top scorers were more or less same pairs each month as were the low
> scorers.

Yes it is much harder to identify good luck, it more often looks like
something you've earned. :-)

[Adam Lea]

On 04/04/13 16:57, france...@googlemail.com wrote:
> "I would like to know how much luck or "chance" there is in duplicate bridge, MP scoring. That is, what are the odds that I will not get a top if I (and my partner) bid and play correctly?"
>
> If there is no luck, then you will get an average, not a top, if you and your partner play correctly, because everyone will do exactly the same as you and everyone will get exactly 50%. In order to get a top, you need 'luck' in that you need someone at another table, or your opponents, to bid and play incorrectly.
>

That is only true if all the players are of equal skill.

[Barry Margolin]

In article <5K-dnUwuSOKVQ8DM...@bt.com>,

Isn't the skill distribution part of the luck factor? E.g. when you play
in a club game, you may be one of the best pairs. When you play in the
Blue Ribbons, you're not so lucky.

But I guess what you're saying is that given a particular distribution
of skills, how much much of your result is due to your position in the
skill spectrum, and how much is due to varation because of random
factors.

--
Barry Margolin
Arlington, MA

[Travis Crump]

It doesn't matter how skillful a player is, either they find the correct
play or they don't. The less skillful player will be less likely to
find the correct play, but it is never impossible that they will.

There is a way though that different tables playing perfectly can
achieve different results though. System design. A 15-17 1Ner might
get to 3N with 16 opposite 8 while a 14-16 won't. One table might open
and play 1N while another might land in 2M or the opponents might get
into the auction because they opened 1m due to a different 1N range. I'm
sure people can come up with more examples.

[vsp...@hotmail.com]

The world class players increases their chances
of being lucky. 70% games still go down 30% of
the time. The inferior line sometimes is the
only line which makes.

[Eric Leong]

On Apr 4, 8:57 am, franceshin...@googlemail.com wrote:
> "I would like to know how much luck or "chance" there is in duplicate bridge, MP scoring. That is, what are the odds that I will not get a top if I (and my partner) bid and play correctly? "
>
> If there is no luck, then you will get an average, not a top, if you and your partner play correctly, because everyone will do exactly the same as you and everyone will get exactly 50%.  In order to get a top, you need 'luck' in that you need someone at another table, or your opponents, to bid and play incorrectly.

You can get good results by being reasonably different from the field.
Also, most pairs consider pairs bidding as a modification of how they
would bid at imps. I don't think most players don't realize just how
radically different the different form of contest is. I find a lot of
pairs players over compete too much and under double too much.

Eric Leong

[Eric Leong]

On Apr 2, 1:55 pm, Jennifer Murphy <JenMur...@jm.invalid> wrote:
> Has anyone done a study or a mathematical (or statistical) analysis of
> the scores in a duplicate event?
>
> In our little three-table social event, people are always complaining
> that they got a low score for the right contract, well played, because
> the opponents of their opponents screwed up by (a) playing poor defense,
> (b) failing to double a bad sacrifice, or (c) making an outrageous bid
> that got doubled for a larger score than the right contract.
>
> Clearly things like that happen and they happen in our group more than
> in most. But I'd live to be able to cite statistics or calculate
> probabilities that the final scores for the night are the result of bad
> luck vs bad play.

You have to consider the strength of the field. The weaker the field
the variance of results will tend to be much higher.

Eric Leong

[patmp...@gmail.com]

On Friday, April 5, 2013 11:13:31 AM UTC+8, Eric Leong wrote:
> I find a lot of
>
> pairs players over compete too much and under double too much.
>

Yes. For some reason percentage thinking goes out the window with doubling. For example, suppose They are vulnerable and bid 3NT. I think they have a 60% chance of making it on the nose, but I have a raggedy long suit and there is 40% chance they will go down three. So my expectation for doubling is .4(500) - .6(150) = 200 - 90 = 110 and it is a good bet. However the average partner thinks you incompetent whether or not the contract goes down.

Another example is that many players correctly have the attitude that if the contract will either "makes or go down a lot" then they should bid game. In other words, if you don't make game you won't even make part score, so it makes sense to bid a 30% game. If they are doing that then you should double. Unfortunately most partners will be infuriated if they make the contract, even though the expectation is highly favorable. The prevailing attitude is to double only if you feel certain they won't make. I hardly ever do that, because they will play me for all the cards and might pull a Houdini.

[Jennifer Murphy]

On Thu, 4 Apr 2013 08:32:22 -0700 (PDT), Adam Beneschan
<ad...@irvine.com> wrote:

>On Thursday, April 4, 2013 2:02:38 AM UTC-7, Jennifer Murphy wrote:
>
>> Mr/Ms Statistician (We now have a matrimonial-neutral title, do we have
>> a gender-neutral one? Mx?):
>
>What would that be short for? "Mixture?"

"Mr." is short for "mister". I think "Mrs." is (was) short for "Missis"
or "Missus" and "Miss" was itself (did not have an abbreviation).

When "Ms." came along, I don't think it was short for anything. It's
just "Ms." and pronounced "miz".

So, "Mx." doesn't need to be short for anything, but I can't think of
any way to pronouce it so that "Mx. Jones" doesn't sound silly.

I don't imagine that "M?." would get past the language police.

In keeping with the KISS principle, maybe just "M.". I kinda like "M.
Jones", but I have no idea how to pronounce it. Would "Mmmmm" work?
Mmmmm Jones? Naw, too suggestive. ;-)

[Jennifer Murphy]

On Thu, 4 Apr 2013 08:57:13 -0700 (PDT), france...@googlemail.com
wrote:

>"I would like to know how much luck or "chance" there is in duplicate bridge, MP scoring. That is, what are the odds that I will not get a top if I (and my partner) bid and play correctly? "
>
>If there is no luck, then you will get an average, not a top, if you and your partner play correctly, because everyone will do exactly the same as you and everyone will get exactly 50%. In order to get a top, you need 'luck' in that you need someone at another table, or your opponents, to bid and play incorrectly.

Not at all. Different bidding and signalling conventions work
differently with different hands.

[Jennifer Murphy]

You make an excellent point. In our low-intermediate group, most of the
players rarely double unless it's almost guaranteed, meaning they can
see the setting tricks in their hand. This gives away a lot of
information, which could lead to a line of play that would be
ill-advised with an average distribution, but brilliant if you know
where all the trumps or aces are. (Well, if you *know*, then maybe not
brilliant, but at least logical.)

[france...@googlemail.com]

On Friday, April 5, 2013 5:57:53 AM UTC+1, Jennifer Murphy wrote:
> On Thu, 4 Apr 2013 08:57:13 -0700 (PDT), france...@googlemail.com wrote: >"I would like to know how much luck or "chance" there is in duplicate bridge, MP scoring. That is, what are the odds that I will not get a top if I (and my partner) bid and play correctly? " > >If there is no luck, then you will get an average, not a top, if you and your partner play correctly, because everyone will do exactly the same as you and everyone will get exactly 50%. In order to get a top, you need 'luck' in that you need someone at another table, or your opponents, to bid and play incorrectly. Not at all. Different bidding and signalling conventions work differently with different hands.

I was challenging what I saw as an implicit assumption in your question - that if you bid and play "correctly" then the only reason you will not get a top is luck. The point is that your initial expectation on a board is an average, not a top. You can get "lucky" by playing a board better than the other players with your cards. You can get "unlucky" if your opponents play better than other players with their cards.

Different bidding and signalling conventions will increase the variability of scores, but
- if they are all equally good, they won't change your expectation of an average result
- if some are better than others, then you are 'lucky' if your opponents are not playing the best methods

[france...@googlemail.com]

You can get good results by being reasonably different from the field.

Simply being different from the field does not increase your expected result (as opposed to being better than the field, which does).

Being different increases the variability of your results. Whether that means you are more likely to win or not depends on how likely you were to win in the first place.

[derek]

On Thursday, April 4, 2013 12:28:08 AM UTC-3, Michael Angelo Ravera wrote:

> It's amazing that a simple arrow switch for one or two rounds would fix all of this (and your director wouldn't do it)! For 14 tables (13 rounds), you'd like to switch 23 of the table-rounds. That would be tables 6-14 on the penultimate round and all of the tables on the last round. For 12 rounds, you'd like to switch 21 table-rounds. It would be the even (or high half) tables on the penultimate round and everybody on the last.

What's "amazing" about that? It's not required that you understand arrow-switching to pass your ACBL director's exam (I know that, because I had no trouble passing it without knowing anything about the subject), and I'd guess that the great majority of club directors don't know how to do it. In any case, Jennifer called this a "social game" so it likely doesn't even have a qualified director - it certainly doesn't need one.

Even in a sanctioned game, it's not something that ACBLscore makes easy and the /ACBL Club Director's Handbook/ has exactly one reference to a "Scrambled Mitchell", tells you nothing about why you would use it, and never mentions "arrow-switching".

[derek]

On Thursday, April 4, 2013 10:30:59 AM UTC-3, bhmwe...@gmail.com wrote:

> Another way to benefit from good luck is what happened to the club's best pair, who managed to get almost 80% in a normal club evening. Their good luck was that they had a bye during the first round. On board 1 all tables except two played 4S +2, and on board 2 ALL tables played 4S +1. Instead of scoring two average boards they scored 80% on the round...

I don't begin to understand how anyone gets 80% for a bye round, but it seems to me that they got 80% for the whole game, by playing an 80% game.

[vsp...@hotmail.com]

It's the nature of the statistic for matchpoints.
While a quality field will have a lower variance,
it is only marginally lower. 70-75% of the results
must to the same exact outcome before the variance
is significantly lower.
Also in small club games with a 3 or 4 top the
variance is much higher than in 8 or 12 top games.

[Michael Angelo Ravera]

I was commenting on the club championship in which Stu played, not Jennifer's home game. You should know for background that Stu and I happen frequently to play at the same club. I direct there. If he happened to be playing in the nearby club, the director would have known how to run an arrow switch (and, in fact, may have been the one who taught me how). If Stu was playing in the club where I direct and where I see him most frequently, the director doubtless knows how to run an arrow switch. I know this because either the director has done it before, asked me how to do it, or is a director in whose game Stu never plays (and even that one would get help running an arrow switch if the field was badly unbalanced).

[france...@googlemail.com]

I don't begin to understand how anyone gets 80% for a bye round, but it seems to me that they got 80% for the whole game, by playing an 80% game.

They don't get 80% for a bye round exactly.

The point is this is another form of luck. If you are going to sit out two boards in an evening, it's a matter of luck whether these are two very difficult hands on which you, as a good pair, would have got tops (which would put their overall percentage up) or they are two boards on which everyone is going to get exactly average (which would bring their overall percentage down)

Another way of saying that the fewer boards you play, the greater the variability of results.

There's yet another form of luck. When you play against the best pair in the room, do you get two boards on which they can do nothing, and you get averages, or do they get the opportunity to have a scientific auction to the making slam and you get a zero? When you play against the worst pair in the room, what if they have no opportunity to hand you matchpoints because, like everyone else, you bid 1NT-7NT and claim at trick 1? If you don't play against everyone, are the pairs you miss out the strongest in the room or the weakest?

[bhmwe...@gmail.com]

Den fredagen den 5:e april 2013 kl. 15:13:26 UTC+2 skrev derek:

> I don't begin to understand how anyone gets 80% for a bye round

Ater the session all pairs who had a bye gets their actual percentage on the bye round.

[derek]

On Friday, April 5, 2013 1:39:11 PM UTC-3, Michael Angelo Ravera wrote:
> On Friday, April 5, 2013 6:03:50 AM UTC-7, derek wrote:
>
> > On Thursday, April 4, 2013 12:28:08 AM UTC-3, Michael Angelo Ravera wrote:
>
> > > It's amazing that a simple arrow switch for one or two rounds would fix all of this (and your director wouldn't do it)!
>

> > What's "amazing" about that? It's not required that you understand arrow-switching to pass your ACBL director's exam... In any case, Jennifer called this a "social game" so it likely doesn't even have a qualified director - it certainly doesn't need one.

>
>
> I was commenting on the club championship in which Stu played, not Jennifer's home game.

Ah, sorry. I clearly didn't get that.

However, I frequently get very unbalanced fields, as it's impossible to get the older players out of their N/S seats, and I really would like to learn how to do this, but the ACBL doesn't seem interested in helping.

[Michael Angelo Ravera]

You don't have to understand the theory of arrow switching except to know that you need to arrow switch about one-eighth of the boards.

With 8 tables, it's really easy: You just arrow switch the last round.
With 4 tables (in a mitchell), you switch half of the last round.
With 12 tables, you switch the last round and one-half.

With other numbers, you just do the best that you can to switch about one-eighth. If you need to arrow switch and aren't sure about how many rounds you are going to have, do it in the penultimate (and antepenultimate, if necessary) round.

If you are using ACBLScore,
First, go to the movement editor and change the movement to one-winner.
Second, change the movement (pairs only) by selecting the round and table where you are going to start the switch. It turns out that switching pairs is sort of the default action, so you will only need to hit <ENTER> a few times to have the movement switched.
Third, warn E-W pairs to add the correct number to their original table number to obtain their pair numbers.
Fourth, remind people of the switch as it is coming. Peter Townsend gave us the method "The north side of town faced east and the east was facing south."

Have fun!

[Douglas Newlands]

Play a Howell movement.

doug

[pgmer6809]

> I think I am wording this badly. It's probably because I'm not sure what
> I am trying to get. Can you read my mind? ;-)

A man read a woman's mind? you've got to be kidding! :)!

Looking at it from another way, I am guessing that what you are most
interested in, is trying to decide which of 3 table groups are the
best players, and which are the luckiest. :)!

Some approaches:
1. Score as you usually do. Keep track of the results for a few
months. At the end of each period declare a winner based on the
aggregate.
My dad used to play bridge at work this way during lunch hour. There
were 4 of them, they traded partner's every month, and at the end of
the year, the hi man was declared the best player.

2. Score by IMPs versus a 'par' result. If you enter the hands after
each session and then run them through an analysis program, it will
tell you whether you achieved a good result on those cards, or not.
Take the difference of your score versus the 'par' score, and convert
to IMPS. (to minimize the effect of one big swing)
I say do this afterwards, because I am assuming that you do not have
the ability to get deals with hand records made up ahead of time for
you.
(or if you have some good players, who can analyze quickly, you can do
this manually after the play is over.)

3. If you have 3 tables, you could play the evening as a Round Robin
KO with IMP scoring. You could make up new teams every session where
every pair plays with every other pair as team mates, and at the end
of the month (or year or whatever) declare a winning pair.

and if I didn't read your mind correctly, put it down to the usual
lack of male understanding. :)!

[derek]

On Tuesday, April 9, 2013 7:57:06 PM UTC-3, Douglas, wrote:
>
> > However, I frequently get very unbalanced fields, as it's impossible to get the older players out of their N/S seats, and
>
> >I really would like to learn how to do this, but the ACBL doesn't seem interested in helping.
>
> Play a Howell movement.

Yeah, right. Which part of "impossible" is unclear? I have the best attended afternoon game in the city. It would quickly become the worst if I made them play a Howell - not to mention that it would be as least as difficult to score a Howell with 4 stationary pairs, as a 12 table Mitchell with arrow switches.

[Jennifer Murphy]

On Wed, 10 Apr 2013 05:12:05 -0700 (PDT), derek <de...@pointerstop.ca>
wrote:

This kind of stubborn entitlement by "older players" is one of the
reasons that my friends and I do not participate in very many of the
local ACBL events. The atmosphere is often not welcoming to newcomers
and/or occasional players. And there are usually a few cranky "veterans"
whose primary job it seems to be to snarl at anyone who makes any
procedural mistakes. Who needs it?

[Barry Margolin]

In article <ef597052-dd1a-4a74...@googlegroups.com>,

derek <de...@pointerstop.ca> wrote:

> However, I frequently get very unbalanced fields, as it's impossible to get
> the older players out of their N/S seats, and I really would like to learn
> how to do this, but the ACBL doesn't seem interested in helping.

Institute a new policy: the director assigns seats, not the players
themselves. A pair may only request N/S seats if one of them is
physically disabled.

[vsp...@hotmail.com]

What's the advantage of online bridge.
Nearly all procedural errors are eliminated.
In club games I give all procedural errors
a mulligan. It has annoyed some partners.

[derek]

As I said to Douglas, I have the most popular afternoon game in town - I don't intend to make it the least popular. In any case, I regularly have 4 disabled Norths.

[derek]

I realize that my wording suggests they're all to blame, and we do have our share of "cranky veterans", but I'm not kidding about needing 4 stationary pairs. While a couple of my oldest and frailest players choose to play East/West because their joints freeze up if they sit too long, I regularly have 3 or 4 disabled players (one who can barely move, one with a new hip, one with an old hip replacement that went bad - and one who many have suggested doesn't actually have a disability: would you be the director to ask for proof? Not me!)

Douglas' suggestion is simply impossible. We have two local clubs. Their games don't usually overlap, but when they do, I've known players to show up at one, find they're going to play a Howell, and race over to the other. They hate Howells. I have exactly one player who would choose a Howell.

[Dave Flower]

On Thursday, 11 April 2013 14:51:54 UTC+1, derek wrote:
> On Wednesday, April 10, 2013 11:59:57 AM UTC-3, Jennifer Murphy wrote: > On Wed, 10 Apr 2013 05:12:05 -0700 (PDT), derek <de...@pointerstop.ca> > wrote: > > >On Tuesday, April 9, 2013 7:57:06 PM UTC-3, Douglas, wrote: > >> > >> > However, I frequently get very unbalanced fields, as it's impossible to get the older players out of their N/S seats, and > >> > >> >I really would like to learn how to do this, but the ACBL doesn't seem interested in helping. > >> > >> Play a Howell movement. > > >Yeah, right. Which part of "impossible" is unclear? I have the best attended afternoon game in the city. It would quickly become the worst if I made them play a Howell - not to mention that it would be as least as difficult to score a Howell with 4 stationary pairs, as a 12 table Mitchell with arrow switches. > > This kind of stubborn entitlement by "older players" is one of the > reasons that my friends and I do not participate in very many of the > local ACBL events. The atmosphere is often not welcoming to newcomers > and/or occasional players. And there are usually a few cranky "veterans" I realize that my wording suggests they're all to blame, and we do have our share of "cranky veterans", but I'm not kidding about needing 4 stationary pairs. While a couple of my oldest and frailest players choose to play East/West because their joints freeze up if they sit too long, I regularly have 3 or 4 disabled players (one who can barely move, one with a new hip, one with an old hip replacement that went bad - and one who many have suggested doesn't actually have a disability: would you be the director to ask for proof? Not me!) Douglas' suggestion is simply impossible. We have two local clubs. Their games don't usually overlap, but when they do, I've known players to show up at one, find they're going to play a Howell, and race over to the other. They hate Howells. I have exactly one player who would choose a Howell.

Have you considered the possibility of atationary E/W pairs (whan they are supposed to go to a table, the North/South pair at that table move to them) ?

Dave Flower

[Wes Powers]

What do you do when a stationary E/W pair meets a stationary N/S pair?

[Barry Margolin]

In article <kk6fi6$4it$1...@dont-email.me>,

Is that anything like the proverbial immovable object and unstoppable
force?

[Bertil]

On Friday, April 5, 2013 12:39:11 PM UTC-4, Michael Angelo Ravera wrote:
> On Friday, April 5, 2013 6:03:50 AM UTC-7, derek wrote:
>
> > On Thursday, April 4, 2013 12:28:08 AM UTC-3, Michael Angelo Ravera wrote:
>
> >
>
> >
>
> >
>
> > > It's amazing that a simple arrow switch for one or two rounds would fix all of this (and your director wouldn't do it)! For 14 tables (13 rounds), you'd like to switch 23 of the table-rounds. That would be tables 6-14 on the penultimate round and all of the tables on the last round. For 12 rounds, you'd like to switch 21 table-rounds. It would be the even (or high half) tables on the penultimate round and everybody on the last.
>
> >
>

Arrow switch is a new concept to me, so I wonder what would happen if after
seating all pairs the director switches the NS arrow to EW on every other board.
How would the computer score the results? I'm assuming that the original outside pairs continue to move.

Stig

[derek]

On Thursday, April 11, 2013 10:56:36 AM UTC-3, Dave Flower wrote:
> On Thursday, 11 April 2013 14:51:54 UTC+1, derek wrote:
>
> Have you considered the possibility of atationary E/W pairs (whan they are supposed to go to a table, the North/South pair at that table move to them) ?

I can't say I have, but what then happens when a stationary E/W is supposed to play a stationary N/S?

[Travis Crump]

Everyone managed to get themselves to the game. They ought to be able
to manage to move for 1 round.

[derek]

So which disabled player do you suggest I force to move?

I honestly can't understand why so many people seem to think it's wrong to cater to the disabled, in a game where the participants are usually of an age that disabilities are common.

[Steve Foster]

No-one (I think) is suggesting not catering for infirmities, but
there's catering and then there's pandering. IMHO, deliberately skewing
movements so that a disabled player never has to face another disabled
player is pandering.

What would you do if over half the field were disabled?

--
Steve Foster
For SSL Certificates, Domains, etc, visit.:
https://netshop.virtual-isp.net

[vsp...@hotmail.com]

Can only happen when there are more disabled
pairs than there are tables in play.

[Bertil]

At my local club game one player is taken by car to the building and then must use a walker to get to the game room and sit down with the walker at his side.
It would be a major effort for him to get up every two deal and go to the
next table. Other players have other impediments to walking around.

Stig

[Wes Powers]

Nope, it will happen with only two disabled pairs, assuming a complete
Mitchell movement.

Suppose N/S pair 3 and E/W pair 1 are disabled. What do you do on the
third round, when they are scheduled to play each other?

Depending on the number of tables, you might be able to avoid this once
by running an incomplete movement, but add another one or two disabled
pairs, and you won't be able to avoid it without running a horribly
unfair movement.

Wes

[Dave Flower]

On Saturday, 13 April 2013 18:22:34 UTC+1, Wes Powers wrote:
> On 4/13/2013 10:18, vsp...@hotmail.com wrote: > On Friday, April 12, 2013 6:06:46 AM UTC-7, derek wrote: >> On Thursday, April 11, 2013 10:56:36 AM UTC-3, Dave Flower wrote: >> >>> On Thursday, 11 April 2013 14:51:54 UTC+1, derek wrote: >> >>> >> >>> Have you considered the possibility of atationary E/W pairs (whan they are supposed to go to a table, the North/South pair at that table move to them) ? >> >> >> >> I can't say I have, but what then happens when a stationary E/W is supposed to play a stationary N/S? > > Can only happen when there are more disabled > pairs than there are tables in play. > Nope, it will happen with only two disabled pairs, assuming a complete Mitchell movement. Suppose N/S pair 3 and E/W pair 1 are disabled. What do you do on the third round, when they are scheduled to play each other? Depending on the number of tables, you might be able to avoid this once by running an incomplete movement, but add another one or two disabled pairs, and you won't be able to avoid it without running a horribly unfair movement. Wes

Note that you can reduce the amount disabled players have to move by arranging then to play on the first (or last) round.

Dave Flower

[Lorne]

On 02/04/2013 21:55, Jennifer Murphy wrote:
> Has anyone done a study or a mathematical (or statistical) analysis of
> the scores in a duplicate event?
>
> In our little three-table social event, people are always complaining
> that they got a low score for the right contract, well played, because
> the opponents of their opponents screwed up by (a) playing poor defense,
> (b) failing to double a bad sacrifice, or (c) making an outrageous bid
> that got doubled for a larger score than the right contract.
>
> Clearly things like that happen and they happen in our group more than
> in most. But I'd live to be able to cite statistics or calculate
> probabilities that the final scores for the night are the result of bad
> luck vs bad play.
>
With only 3 tables MP scoring is always going to be a lottery. I
suggest you change the scoring to Butler IMPs which is much more likely
to give a fair result when there are very few tables.

[derek]

On Saturday, April 13, 2013 11:08:31 AM UTC-3, Steve Foster wrote:
> derek wrote:
>
> > On Saturday, April 13, 2013 1:34:36 AM UTC-3, Travis Crump wrote:
> > > On 04/12/2013 09:06 AM, derek wrote:
> > >
> > > > On Thursday, April 11, 2013 10:56:36 AM UTC-3, Dave Flower wrote:
> > >
> > > >> Have you considered the possibility of atationary E/W pairs
> > > (whan they are supposed to go to a table, the North/South pair at
> > > that table move to them) ?
> > >
> > > > I can't say I have, but what then happens when a stationary E/W
> > > > is supposed to play a stationary N/S?
> > >
> > > Everyone managed to get themselves to the game. They ought to be
> > > able to manage to move for 1 round.
> >
> > So which disabled player do you suggest I force to move?
> > I honestly can't understand why so many people seem to think it's
> > wrong to cater to the disabled, in a game where the participants are
> > usually of an age that disabilities are common.
>
> No-one (I think) is suggesting not catering for infirmities, but
> there's catering and then there's pandering. IMHO, deliberately skewing
> movements so that a disabled player never has to face another disabled
> player is pandering.

I beg to differ. Quite a few people seem to think it's completely improper for me to let people have a stationary table unless I set it myself. They don't care that I have a bunch of disabled players, and are calling me lazy, and them "cranky veterans".

> What would you do if over half the field were disabled?

So far I've been talking about 4 players in a 12-table field. I'm honestly not worried about it happening, but if it did, I'd probably just throw up my hands and admit I'm never going to get to a fair movement. I don't get paid enough to direct this game (it's well under a tenth of my usual half-day rate...) to worry about making it perfect - I was simply asking for some advice for making it better than it is.

[derek]

Really? I'd need to see the movement, but that doesn't seem intuitive. For instance, consider a "perfect" movement - 8 tables, with a bye-stand and relay where every North/South meets every East/West and they play 24 boards - then if you have just one stationary East/West, they're going to have to meet our stationary North/South. Now, perhaps you need to use a different movement, if you have a stationary East/West, but that's the sort of thing I'd need to know if I was doing this...

[Dave Flower]

On Sunday, 14 April 2013 00:32:31 UTC+1, derek wrote:
> On Saturday, April 13, 2013 11:08:31 AM UTC-3, Steve Foster wrote: > derek wrote: > > > On Saturday, April 13, 2013 1:34:36 AM UTC-3, Travis Crump wrote: > > > On 04/12/2013 09:06 AM, derek wrote: > > > > > > > On Thursday, April 11, 2013 10:56:36 AM UTC-3, Dave Flower wrote: > > > > > > >> Have you considered the possibility of atationary E/W pairs > > > (whan they are supposed to go to a table, the North/South pair at > > > that table move to them) ? > > > > > > > I can't say I have, but what then happens when a stationary E/W > > > > is supposed to play a stationary N/S? > > > > > > Everyone managed to get themselves to the game. They ought to be > > > able to manage to move for 1 round. > > > > So which disabled player do you suggest I force to move? > > I honestly can't understand why so many people seem to think it's > > wrong to cater to the disabled, in a game where the participants are > > usually of an age that disabilities are common. > > No-one (I think) is suggesting not catering for infirmities, but > there's catering and then there's pandering. IMHO, deliberately skewing > movements so that a disabled player never has to face another disabled > player is pandering. I beg to differ. Quite a few people seem to think it's completely improper for me to let people have a stationary table unless I set it myself. They don't care that I have a bunch of disabled players, and are calling me lazy, and them "cranky veterans". > What would you do if over half the field were disabled? So far I've been talking about 4 players in a 12-table field. I'm honestly not worried about it happening, but if it did, I'd probably just throw up my hands and admit I'm never going to get to a fair movement. I don't get paid enough to direct this game (it's well under a tenth of my usual half-day rate...) to worry about making it perfect - I was simply asking for some advice for making it better than it is.

I assume you mean 4 disabled pairs.

If you are playing an 11-round skip Mitchell, then plac the disabled pairs at tables 1 and 7 (or 2/8 etc.). E/W 1 and 7 would have to move to stationary E/W tables at the end of round 1, but that is all, as the skip conveniently avoids the problem of NS1 playing EW7 etc

Dave Flower

[Jennifer Murphy]

On Sun, 14 Apr 2013 00:09:49 +0100, Lorne <lorne_a...@hotmail.com>
wrote:

I had never heard of Butler Scoring, so I did a little research. I
didn't find a lot of information, but I think I have the general idea.

1. First calculate a "par" score. One site said to throw out the highest
and lowest score and average the rest. This suggested to me that Butler
scoring is not intended for a small event. In our three-table event, par
would be the middle score.

2. Calculate the difference between each score and the par.

3. Convert to IMPs.

Is that how it works?

For a three-table event, I can't see throwing out two scores, so I'd
average all three.

I'm not sure how this is more fair. It seems like it would result is
larger score differences. In a three-table MP event, a top is 4 and a
bottom 0.

Do you have references to more detailed discussions?

[derek]

On Sunday, April 14, 2013 5:15:10 AM UTC-3, Dave Flower wrote:
> On Sunday, 14 April 2013 00:32:31 UTC+1, derek wrote:
>
>> What would you do if over half the field were disabled?

> So far I've been talking about 4 players in a 12-table field.
>

> I assume you mean 4 disabled pairs.

No, I mean four disabled _players_. But I've never had the good fortune to get two of them to play together.

> If you are playing an 11-round skip Mitchell, then plac the disabled pairs at tables 1 and 7 (or 2/8 etc.). E/W 1 and 7 would have to move to stationary E/W tables at the end of round 1, but that is all, as the skip conveniently avoids the problem of NS1 playing EW7 etc

OK

[Charles Brenner]

On Apr 3, 9:02 am, vspo...@hotmail.com wrote:

> On Tuesday, April 2, 2013 1:55:48 PM UTC-7, Jennifer Murphy wrote:
> > Has anyone done a study or a mathematical (or statistical) analysis of
>
> > the scores in a duplicate event?
>
> > In our little three-table social event, people are always complaining
>
> > that they got a low score for the right contract, well played, because
>
> > the opponents of their opponents screwed up by (a) playing poor defense,
>
> > (b) failing to double a bad sacrifice, or (c) making an outrageous bid
>
> > that got doubled for a larger score than the right contract.
>
> > Clearly things like that happen and they happen in our group more than
>
> > in most. But I'd live to be able to cite statistics or calculate
>
> > probabilities that the final scores for the night are the result of bad
>
> > luck vs bad play.
>

> The standard deviations for matchpoint
> results is approximately 28% per board.
> There's a lot of luck in this game.
> Good pairs overcome this.  'Bad' pairs
> use this as an excuse to whine.

That's an interesting number. Where is it from? Obviously it would
vary according to the number of scores, so I assume it is based on
some particular number of scores -- each board played 13 times maybe?
Anyway, it seems to me useful in that with this one fact plus a few
plausible modeling assumptions, one could compute distributions of
scores for a session.

I propose an alternative, arguably more basic, datum which I think
would do the trick about equally well: The probability that two
randomly selected results for a board will be the same. Of course this
is high for routine deals and lower for exciting or problematic ones,
but I'm guessing that we can settle for the overall probability. I
like my statistic because it is fundamental both in that we can
compute the per-board standard deviation from it, for any specified
number of plays, and in that it's closer to purely mathematical and
less statistical, which suits my inclination.

[Michael Angelo Ravera]

On Thursday, April 11, 2013 8:29:24 AM UTC-7, Bertil wrote:
> Arrow switch is a new concept to me, so I wonder what would happen if after
> seating all pairs the director switches the NS arrow to EW on every other board.
> How would the computer score the results? I'm assuming that the original outside pairs continue to move.

First, you don't switch half the boards or anything like it. Your aim is for one-eighth.

Second, if you hypothetically set the movement that you described up properly in ACBLScore, it would work just fine. (although ACBLScore might have a problem with more than 25 rounds).

Third, usually you switch whole rounds of boards.

Here is what the guide cards look like for a switch with a Bye and Share 6-table game:

PAIR 1 PAIR 2 PAIR 3 PAIR 4
6 TABLES 6 TABLES 6 TABLES 6 TABLES
RD SIT VS BOARDS RD SIT VS BOARDS RD SIT VS BOARDS RD SIT VS BOARDS
1 1N 7 1- 4(R) 1 2N 8 5- 8 1 3N 9 9-12 1 4N 10 17-20
2 1N 12 5- 8(R) 2 2N 7 9-12 2 3N 8 13-16 2 4N 9 21-24
3 1N 11 9-12(R) 3 2N 12 13-16 3 3N 7 17-20 3 4N 8 1- 4
4 1N 10 13-16(R) 4 2N 11 17-20 4 3N 12 21-24 4 4N 7 5- 8
5 1N 9 17-20(R) 5 2N 10 21-24 5 3N 11 1- 4 5 4N 12 9-12
6 1N 8 21-24(R) 6 2E 9 1- 4 6 3E 10 5- 8 6 4E 11 13-16


PAIR 5 PAIR 6 PAIR 7 PAIR 8
6 TABLES 6 TABLES 6 TABLES 6 TABLES
RD SIT VS BOARDS RD SIT VS BOARDS RD SIT VS BOARDS RD SIT VS BOARDS
1 5N 11 21-24 1 6N 12 1- 4(R) 1 1E 1 1- 4(R) 1 2E 2 5- 8
2 5N 10 1- 4 2 6N 11 5- 8(R) 2 2E 2 9-12 2 3E 3 13-16
3 5N 9 5- 8 3 6N 10 9-12(R) 3 3E 3 17-20 3 4E 4 1- 4
4 5N 8 9-12 4 6N 9 13-16(R) 4 4E 4 5- 8 4 5E 5 9-12
5 5N 7 13-16 5 6N 8 17-20(R) 5 5E 5 13-16 5 6E 6 17-20(R)
6 5E 12 17-20 6 6N 7 21-24(R) 6 6E 6 21-24(R) 6 1E 1 21-24(R)


PAIR 9 PAIR 10 PAIR 11 PAIR 12
6 TABLES 6 TABLES 6 TABLES 6 TABLES
RD SIT VS BOARDS RD SIT VS BOARDS RD SIT VS BOARDS RD SIT VS BOARDS
1 3E 3 9-12 1 4E 4 17-20 1 5E 5 21-24 1 6E 6 1- 4(R)
2 4E 4 21-24 2 5E 5 1- 4 2 6E 6 5- 8(R) 2 1E 1 5- 8(R)
3 5E 5 5- 8 3 6E 6 9-12(R) 3 1E 1 9-12(R) 3 2E 2 13-16
4 6E 6 13-16(R) 4 1E 1 13-16(R) 4 2E 2 17-20 4 3E 3 21-24
5 1E 1 17-20(R) 5 2E 2 21-24 5 3E 3 1- 4 5 4E 4 9-12
6 2N 2 1- 4 6 3N 3 5- 8 6 4N 4 13-16 6 5N 5 17-20

[vsp...@hotmail.com]

You can make the calculations yourself.
I already posted it once.

VAR(X)=E(x²)-[E(x)]²
Those were old pencil and paper days.

Nowadays we can use EXCEL.
For each board.
Enter the outcome results(matchpoints) into a column.
On the next empty cell, click the 'fx' function.
Click 'Statistical', find and click 'VAR'.
Under number 1 enter the range of the location
of the outcome results. Use the ':' between
the beginning and end of the range.
Then on the next empty cell
=SQRT(VAR cell)
May need an adjustment. Multiply by 100 and
divide by the max matchpoints for the board.

[Jennifer Murphy]

On Sun, 14 Apr 2013 00:09:49 +0100, Lorne <lorne_a...@hotmail.com>
wrote:

After a little more research, I found several sites, one of which is
hilarious, especially if you have ever been a director. I'm going to
start a new thread on Butler scoring. Thanks for the tip.

[Charles Brenner]

Yes, that is how to compute the standard deviation and the coefficient
of variation. I meant, where is your data from? What kind of, and
what, events? How much data? And you overlooked another question that
I asked, the top per board. (Surely you aren't claiming that the stdev
is 28% independent of what top is, are you?)

[derek]

On Sunday, April 14, 2013 6:48:18 PM UTC-3, Michael Angelo Ravera wrote:
>
> Here is what the guide cards look like for a switch with a Bye and Share 6-table game:

Does ACBLScore actually have that movement, have you created it as an external movement, or do you actually have to manually switch all of those (well, just the N/S's - I know ACBLscore is smart enough to automatically switch the corresponding pairs)?

Obviously, if I have to create an external movement, I only have to do it once (not something I've ever done, but I'm not scared to do it).

If I'm getting the hang of this, you have in fact switched 1/9th of the boards - which is as close to 1/8th as you're going to get by switching for whole rounds.

[derek]

On Monday, April 15, 2013 9:58:27 AM UTC-3, derek wrote:

> If I'm getting the hang of this, you have in fact switched 1/9th of the boards - which is as close to 1/8th as you're going to get by switching for whole rounds.

Ah-hah! And arrow-switching one more table would put you just as far on the other side of 1/8th, so not worth the confusion.

Do the Bridgemates understand the switch?

What's the significance of 1/8th? I mentioned elsewhere that I have exactly one player who would choose to do a Howell when I don't have to. Specifically, we'll usually use a bumping N/S if we have 7 1/2 tables, and she doesn't feel it's fair - I have always suspected she's wrong, but can't explain why, but it seems like it should have a similar effect to arrow-switching about an 1/8th of the boards

[vsp...@hotmail.com]

If the top is 8 or higher, the std dev is
approximately 28%. Lower tops from small
club games have a higher s.d.

[Michael Angelo Ravera]

The "Hayashi" 7.5 table movement isn't as good as if there were some additional arrow switches (we'll work out how many in a minute).

BridgeMates and BridgePads don't have any problem with an arrow-switch movement, but BridgePads don't give a good overall percentage when you do an arrow switch. I haven't run an arrow switched pairs game with BridgeMates lately (Our club uses Pad).

It looks to me that if each of 28 boards is played 7 times, you would want to switch about 24 of them to make the "fairest" movement. It sounds like the typical Hayashi movement is about 12 board switches short of that. The solution would be to switch half of the last round (or the whole nominal last round [which is actually a half-round] except at the table where people are being bumped).

[derek]

On Tuesday, April 16, 2013 4:29:41 AM UTC-3, Michael Angelo Ravera wrote:
>
> BridgeMates and BridgePads don't have any problem with an arrow-switch movement, but BridgePads don't give a good overall percentage when you do an arrow switch. I haven't run an arrow switched pairs game with BridgeMates lately (Our club uses Pad).

That's OK - my players mostly haven't figured out the overall percentage function on the BridgeMate, and don't totally trust it anyway, so as long as it's handling the movement properly, it's cool.

>
> It looks to me that if each of 28 boards is played 7 times, you would want to switch about 24 of them to make the "fairest" movement. It sounds like the typical Hayashi movement is about 12 board switches short of that. The solution would be to switch half of the last round (or the whole nominal last round [which is actually a half-round] except at the table where people are being bumped).

Very good. I'll try that next time it comes up! And refer my critic to you :)

[Lorne]

You are correct about how it works and I think that with 3 tables you
should average all 3 scores although there is a case for using the
middle score.

As to volatilty of results the problem with MP's and 3 tables is that a
hand where 2 people make +110 and the other makes +140 the pair with
+140 get 100% on the board and the other 2 get 25%. Assuming 24 boards
are played this means the board is worth 4% overall for one pair and 1%
for the other 2 playing - a very big difference for just an overtrick.

Butler scoring would give a par result of +110 or +120 depending on how
you get the par so just 1 IMP for the player making an overtrick.
Typically a Butler winner would have around +50IMPs so this is small and
unlikely to affect the ranking but in pairs a 3 or 3% swing is huge.

You do ofcourse still get some big swings but even bidding a slam missed
at two tables is only 11-13IMPs or so which is still less than the
efffect of an overtrick at MP scoring on the overall ranking for this
field size.

[Grant Robinson]

On Wednesday, April 3, 2013 10:49:42 AM UTC-7, Stu Goodgold wrote:
> On Wednesday, April 3, 2013 9:02:07 AM UTC-7, vsp...@hotmail.com wrote:

>
> > On Tuesday, April 2, 2013 1:55:48 PM UTC-7, Jennifer Murphy wrote:
>
> >
>
> > > Has anyone done a study or a mathematical (or statistical) analysis of
>
> >
>
> > >
>
> >
>
> > > the scores in a duplicate event?
>
>
>
> >
>
> >
>

> > The standard deviations for matchpoint
>
> > results is approximately 28% per board.
>
> > There's a lot of luck in this game.
>
> >
>
> > Good pairs overcome this. 'Bad' pairs
>
> > use this as an excuse to whine.
>
>
>

> Where do you get this 28% number?
>
>
>
> I would expect the variance depends on a number of factors:
>
>
>
> number of tables
>
> number of boards played
>
> skill level of the field (both average and skew)
>
>
>
> 2 days ago we had a 14 table club game which was a club championship with overall MP awards. I pointed out before the start that the 5 best pairs were all sitting EW, with a major skill advantage over the over pairs. One of our grand lifemasters also noted that. The director agreed but reluctantly didn't want to switch any NS pairs, who reserve their seats. Sure enough, a B-level NS pair finished 1st NS and 2nd overall. So not only does the average skill level of the field matter, it matters how the skill level is skewed.
>
>
>
> -Stu Goodgold
>
> San Jose, CA

It is not valid to say that "a N-S pair finished second overall" since the movement did not accomplish a single winner. It is not valid to compare N-S and E-W scores in a two-winner movement. However, I think we can fairly say that the masterpoints awarded probably did not reflect the relative playing skill of the entrants on that day. That isn't best, but isn't horrid, in my opinion.