[hotShot]

cardsharp, on Nov 3 2006, 03:55 PM, said:

As noted in the link from my previous post, a median variant of Bulter scoring is certainly a viable method scoring and mode (most common result) would work too, but they also have flaws.

http://www.blakjak.demon.co.uk/butler.htm said:

According to David Grabiner of the USA, Robin Barker of England and Jeff Goldsmith of the USA the disadvantage of the median method of scoring [taking the middle result as the datum, or the arithmetic mean between the two middle results with an even number of frequencies] is primarily because of "polarised" boards. If there is a close 3NT that everyone will bid but might make or go off, then you might get a set of scores such as:

+1100 +600 +600 +600 +600 +600 +600 -100 -100 -100 -100 -100 -200

Using the median method the datum is +600 and the imp scores become

+11 0 0 0 0 0 0 -12 -12 -12 -12 -12 -13

Now, if one score is changed from +600 to -100 the datum becomes -100 and the imp scores become

+15 +12 +12 +12 +12 +12 0 0 0 0 0 0 -3

It does not seem reasonable that changing one score should make such an enormous difference. However, when the boards are not polarised this looks quite a fair method. It has the advantage that the datum score is a "real" score for people to imp against, and on non-polarised boards people quite like it.

No method is perfect, we all have our preferences, and most threads on this topic finish with everyone agreeing to disagree

Paul

Paul i don't understand what you are trying to say.

Gerado asked why anybody would want to prefer Butler to CrossIMP's.

The answer to that question is, CrossIMPs are very unfair to almost half of the field, if a board has a very unusual result, especially if there is a dominant score. It does not even matter whether that unusual result was caused by playing perfect bridge or producing utter nonsense.

I was not answering the question why someone would not want to use Butler scoring.

On BBO especially in the MBC we have a unusual score on most boards. So most boards are scored unfair. Since there are only 16 scores available the unfairness is significant. If there were 100 scores it would be much less relevant.
For those boards Butler scoring would definitely be an improvement.

I know this modified Butler scoring method, but i don't think it has enough benefits to use it. Original Butler scoring compares everyones score against the average not against the median.
Using a median only makes sense if there is one. Using the average is more stable even if there are no scores close to the average.

[1eyedjack]

hotShot, on Nov 3 2006, 01:53 PM, said:

If real life bridge my team-mates are my choice and i must accept to get a bad result because of this choice.

And yet you may not have the luxury of choosing who your opponents are at the other table. 50% of the so-called skewed results that arise at the other table, perhaps more, may be attributable to their action.

So I ask you this. In a one-against-one teams match, if you bid and make a normal 4S, and your opponents sitting your way at the other table overreach to bid 6S going down, are you suggesting that you should only get zero IMPs, on the grounds that you took the normal action?

[cherdano]

Guys get a life. One unusual result just gives you less than an IMP. If you compare this to the 3-5 IMPS you get in main room for bidding an obvious game on 23 hcp...

Also, sometimes the one extreme result is just the only table who found a good 6♠ making. Why would you throw this out?

I am like Gerardo, I really fail to understand why anyone would prefer any scoring method over CrossIMPs.

[hotShot]

cherdano, on Nov 3 2006, 07:22 PM, said:

Guys get a life. One unusual result just gives you less than an IMP. If you compare this to the 3-5 IMPS you get in main room for bidding an obvious game on 23 hcp...

Also, sometimes the one extreme result is just the only table who found a good 6♠ making. Why would you throw this out?

I am like Gerardo, I really fail to understand why anyone would prefer any scoring method over CrossIMPs.

When we discuss rating, winning an IMP/board is considered a strong performance.
Loosing an IMP/board by sitting on the "wrong side" of table does not even matter to you?
I will remember that .

[Blofeld]

But the IMP/board is not consistently in the same direction ... in fact it's going to be pretty random.

I can't imagine preferring anything to cross-IMPs. Though a wider field of comparison would be good, if possible.

[cherdano]

hotShot, on Nov 3 2006, 07:44 PM, said:

cherdano, on Nov 3 2006, 07:22 PM, said:

Guys get a life.  One unusual result just gives you less than an IMP. If you compare this to the 3-5 IMPS you get in main room for bidding an obvious game on 23 hcp...

Also, sometimes the one extreme result is just the only table who found a good 6♠ making. Why would you throw this out?

I am like Gerardo, I really fail to understand why anyone would prefer any scoring method over CrossIMPs.

When we discuss rating, winning an IMP/board is considered a strong performance.
Loosing an IMP/board by sitting on the "wrong side" of table does not even matter to you?
I will remember that .

I am happy to lose an IMP a board by some random results due to sitting in the wrong direction every time, if you give me 5 IMPs a board by always giving me the easy-to-bid 23 hcp games that more than half the MBC misses.

But actually, I really don't care about losing an IMP/board in MBC by sitting in the wrong direction.

[Stephen Tu]

Quote

f real life bridge my team-mates are my choice and i must accept to get a bad result because of this choice. With CrossIMPs (especially at BBOs MBC) i win my team-mates by seating.

With CrossIMP's the seating creates a noise of about 1 IMP/Board.
I consider that to much.

But this noise isn't really being caused by the X-imps, its's caused by being pairs instead of teams. You are in effect playing with a bunch of random teammates, with average skill level overall. In your scenario, you have 14 teammates who bring back +420 & 1 brings back -100. You say that this -100 is an abnormal result & should be ignored. But how do you know that? How do you know that it isn't true that if you played this board 3x more times that about 1/15 of your teammates go nuts & overbid? This might be perfectly normal expectation.

Now if that 1 team brings back 7NTxx -5, that's annoying, but those kinds of things can be reported to BBO & the people warned.

Butler magnifies swings (IMP table really designed to compare actual bridge scores, not bridge scores to an average of bridge scores) & nullifies overtricks in many instances. There's really no advantage to it except for hand scoring. You could make argument for a modified cross-imp throwing out high & low but I think it's unnecessary, just make sure boards are played a sufficient amount of times, the random outlier swings are small & will even out. If MBC is only playing boards 16 times, I'd think that's kind of small, double would be better. I think okbridge eventually settled on a 52 max.

Randomness in pair games is inherently unavoidable. If you want to get rid of noise, play team matches only, & preferrably BAM.

[1eyedjack]

Blofeld, on Nov 3 2006, 07:11 PM, said:

But the IMP/board is not consistently in the same direction ... in fact it's going to be pretty random.

I can't imagine preferring anything to cross-IMPs. Though a wider field of comparison would be good, if possible.

This is to my mind the crux of the matter.

I am happy to accept that "noisy" results are more heavily weighted to the extreme scores. Which is not to say that all extreme results are noisy, nor even that the results within the middle body are exempt from noise. But as a statistical generality I would accept that the weighting is at the extremes.

What I cannot accept, however, is that any one orientation is any more prone to contributing such noise than the other.

If you think that a particular random extreme result on "hand A" should be excluded from comparison with your own on the grounds that it was a particular orientation that contributed primarily to the result at the other table, then for every such occasion there will be another "hand B" in which it was caused by the other orientation, on which occasion the exclusion of that comparison would be as much a travesty as would be the inclusion of "hand A".

I also feel that the distortion of the logarithmic IMP scale that is built into Butler by reason of IMPing against a depressed aggregate difference (comparison with mean rather than with the actual results at other table) is a negative feature in the argument for Butler scoring. For sure, it is not "unfair", because all are subjected to the same rules, any more than Matchpoint scoring is "unfair" simply because it is different. But for all that it would be fair, I think it unfortunate that the choice of scoring would mean that you are no longer in an environment in which the optimal strategy precisely mirrors that of a teams game (eg, the break-even odds to justify bidding games, slams etc), as is currently the case.

Perhaps if the day should ever come in which Fred has nothing better to do with his time (and it never will) then Butler scoring might be introduced as an option. It would be stupid to ditch all the code currently in operation, so it could be an additional option rather than a replacement, and the table host or tourney host could choose. Heck, you might even be able to display the scoring under both methods. However in this respect I agree with Cherdano ... what a waste of time and effort that would be. On the other hand, increasing the number of comparisons per hand sounds like a worthwhile goal.

[hrothgar]

1eyedjack, on Nov 3 2006, 10:47 PM, said:

Blofeld, on Nov 3 2006, 07:11 PM, said:

But the IMP/board is not consistently in the same direction ... in fact it's going to be pretty random.

I can't imagine preferring anything to cross-IMPs. Though a wider field of comparison would be good, if possible.

This is to my mind the crux of the matter.

I am happy to accept that "noisy" results are more heavily weighted to the extreme scores. Which is not to say that all extreme results are noisy, nor even that the results within the middle body are exempt from noise. But as a statistical generality I would accept that the weighting is at the extremes.

What I cannot accept, however, is that any one orientation is any more prone to contributing such noise than the other.

If you think that a particular random extreme result on "hand A" should be excluded from comparison with your own on the grounds that it was a particular orientation that contributed primarily to the result at the other table, then for every such occasion there will be another "hand B" in which it was caused by the other orientation, on which occasion the exclusion of that comparison would be as much a travesty as would be the inclusion of "hand A".

I also feel that the distortion of the logarithmic IMP scale that is built into Butler by reason of IMPing against a depressed aggregate difference (comparison with mean rather than with the actual results at other table) is a negative feature in the argument for Butler scoring. For sure, it is not "unfair", because all are subjected to the same rules, any more than Matchpoint scoring is "unfair" simply because it is different. But for all that it would be fair, I think it unfortunate that the choice of scoring would mean that you are no longer in an environment in which the optimal strategy precisely mirrors that of a teams game (eg, the break-even odds to justify bidding games, slams etc), as is currently the case.

Perhaps if the day should ever come in which Fred has nothing better to do with his time (and it never will) then Butler scoring might be introduced as an option. It would be stupid to ditch all the code currently in operation, so it could be an additional option rather than a replacement, and the table host or tourney host could choose. Heck, you might even be able to display the scoring under both methods. However in this respect I agree with Cherdano ... what a waste of time and effort that would be. On the other hand, increasing the number of comparisons per hand sounds like a worthwhile goal.

Here's an even simpler idea:

We all have access to the travellers... If "your" self-esteem is so caught up in the size of your score relative to the rest of the field, just go and recompute your score using whatever method you feel like. For all I care, folks can give themselves +14 IMPs on each and every hand...

Its not the like the score that BBO lists actually count for anything.

[hotShot]

The shortest way to improve the CrossIMP's is taking more results/board.
Because e.g. with 32 scores/board the noise reduces to 0.5 IMP/Board.

Playing in the MBC you never change directions. So if some "unlucky" pair is distorting results it will always be on the same side. (Fortunately there is a good chance that they are not playing the same board as you do again.) That could be optimized by changing directions once in a while.

[Flame]

I dont understand why there is no cut off from both sides on from the result like on most imp bridge tournaments.

[hotShot]

Flame, on Nov 3 2006, 10:08 PM, said:

I dont understand why there is no cut off from both sides on from the result like on most imp bridge tournaments.

CrossIMPS are calculated as the average result of team games teaming you up with every other pair that is sitting at the other direction at the other tables.

You can't leave one away, without loosing the base of your scoring system.

[Stephen Tu]

I don't see why there should be arbitrary exclusion of scores. I don't know about you, but when I play teams my teammates don't bring back par every time, on rare occasions they will do something spectacularly good or awful. Including these extreme scores simulates that. Increasing the number of comparisons to me is easiest & fairest way to reduce the effect of outlying scores without completely ignoring them which to me is also wrong.

OTOH random N/S & EW assignment is reasonable suggestion, it may eliminate any bias of established partnerships sitting in a particular direction.

[neilkaz]

Scores that are obviously from someone or pair screwing around with bidding ..ie 7NTxx with part score cards, or conceding a bunch of tricks that are obvious winners should and must be discarded.

I applaude the BBO staff for their diligence in doing this. .. neilkaz ..

[Flame]

hotShot, on Nov 3 2006, 03:24 PM, said:

Flame, on Nov 3 2006, 10:08 PM, said:

I dont understand why there is no cut off from both sides on from the result like on most imp bridge tournaments.

CrossIMPS are calculated as the average result of team games teaming you up with every other pair that is sitting at the other direction at the other tables.

You can't leave one away, without loosing the base of your scoring system.

I would just cut the best result for each side and do the normal cross imp.
Do not compare the extreme results.
Yes its not prefect but its better the no cut.
And again this isnt my idea, this is how imp results usually done.
You might think i just didnt understand what corossIMPs mean, but you are wrong, i just think its not that hard to elimiate the 2 extreams even at crossimps.

[hotShot]

Stephen Tu, on Nov 3 2006, 10:27 PM, said:

I don't see why there should be arbitrary exclusion of scores. I don't know about you, but when I play teams my teammates don't bring back par every time, on rare occasions they will do something spectacularly good or awful. Including these extreme scores simulates that. Increasing the number of comparisons to me is easiest & fairest way to reduce the effect of outlying scores without completely ignoring them which to me is also wrong.

OTOH random N/S & EW assignment is reasonable suggestion, it may eliminate any bias of established partnerships sitting in a particular direction.

You have 2 scores:
420
430
At IMPs there is no difference at MP there is all the difference you need.

If the results are:
420
420
...
420
-100

You can give (case 1):
+0.4 -0.4
.....
somehow distributed 12

or (case 2)
0 0
....
somehow distributed 12

In case 1 pairs are rewarded / punished for sitting on right / wrong side.
In case 2 these pairs will get the same score as they would get in most of the team games. Different scores for these pairs will be produced at other boards. Hopefully related to their play.

[hotShot]

Flame, on Nov 3 2006, 10:34 PM, said:

I would just cut the best result for each side and do the normal cross imp.
Do not compare the extreme results.
Yes its not prefect but its better the no cut.
And again this isnt my idea, this is how imp results usually done.
You might think i just didnt understand what corossIMPs mean, but you are wrong, i just think its not that hard to elimiate the 2 extreams even at crossimps.

How do you calculate the results for the extreme scores?

What you describe is Butler scoring.

[Flame]

hotShot, on Nov 3 2006, 04:05 PM, said:

Flame, on Nov 3 2006, 10:34 PM, said:

I would just cut the best result for each side and do the normal cross imp.
Do not compare the extreme results.
Yes its not prefect but its better the no cut.
And again this isnt my idea, this is how imp results usually done.
You might think i just didnt understand what corossIMPs mean, but you are wrong, i just think its not that hard to elimiate the 2 extreams even at crossimps.

How do you calculate the results for the extreme scores?

What you describe is Butler scoring.

I agree its very close, the different is what you avarage, on butler you avarage the score, and on crossimps you avarage the imp.
reading all you wrote on this thread i agree with you. On a field like BBO main club there should be a cut off, if you say that cut offs extremes only exist on butler then im with you on suggesting to change it, but if crossimps has some advantage then i think there could be a way to get those advantage and still cut extremes, i think its only a technical problem to do it which isnt that hard.

[inquiry]

Cross Imps is the way to go and I vote it never be changed.

As for "funny results", in tournaments, if someone intentionally does soemthiong wacky, the director can correct the score to protect the field by adjusting the result.

In main room first off, it doesn't really matter. Are you taking those scores THAT serioiusly? Second, each group of 16 hands are compared to 15 different talbles (or at least a lot of different ones) so the "arrow switch" thing is surely maximized. Third, a lot weird results are attempts of people to win bridge. If they get it right, even for the wrong reason, they deserve there good result. You got the "normal one" and their weird one beat you out? Tough luck. IF they get it wrong, you get a better than average result. That is bridge. And we do have three and four table events, imagine those where your throw out the top and bottom score (well with 3 it doesn't matter).

Lets imagine a worse case (and there are no where near that many) where one pair produces a weird result that gives the players sitting your direction 24 imps (the maximum). This is compared across 16 tables (divide by 15), so each other table gets 24/15 or -1.6 imps. That might not make us happy, but in a tourmaent, that 24 imps would give that pair a commanding lead (assuming the director doesn't fit it), but in the main room? So what? Your overall score is only compared with yourself...

And lets take that tragic -1.6 imps you get credited towards you when some really does something stupid and sufferes -24 imps. How big is that? I would be much more upset about the - fact that if your opponents have 26 to 29 hcp and opener has 4333, 4432, 5332, or 5422 distribtuion and they bid and make a pedestrian 3NT you will average -2.77 imps, and that is little unchanges rather you use butler or crossimps.

And really, what does throwing out the highest and lowest really do for you? Not much, but often over reward incompetence. Check this hand. NS ahve 12 tricks in clubs (4D, 4C, 2H, 1S, 1H ruff). Look at all the 6NT biddes... (8), 7NT ibdders (1) and amazingly a 6S bidder. Only one pair bid the correct slam to play (and even that is 50-50 as you have to find the club). Throw out the high and low, and average the results, imp with what is left, and the underbidders are rewarded (their scores go up), when in fact, they have very little to be proud of. If heart hook was on, they would have lost tons of imps. Cross imps look fairer to me. Also note, and this is important, you ge the same score as anyone sitting your direction with the same result (true of either method). (on this hand, your weird result (6C making) is my idea of good bridge).

IMP-291 [space] [space] notaio [space] [space] [space]Dlr: North 
Board 6596 [space]S Q763 [space] [space] [space]Vul: N-S
 [space] [space] [space] [space] [space] [space]H AJ5 [space] [space] [space] 
giancarlod [space]D KQ [space] [space] [space] [space]Aris 61 [space] [space] 
S T8542 [space] [space] C AJ32 [space] [space] [space]S KJ [space] [space] [space] [space]
H 3 [space] [space] [space] [space] [space] [space] [space] [space] [space] [space] H QT9864 [space] [space]
D T932 [space] [space] [space]BDP30 [space] [space] [space] D 754 [space] [space] [space] 
C Q84 [space] [space] [space] S A9 [space] [space] [space] [space]C 76 [space] [space] [space] [space]
 [space] [space] [space] [space] [space] [space]H K72 [space] [space] [space] 
 [space] [space] [space] [space] [space] [space]D AJ86 [space] [space] [space]
 [space] [space] [space] [space] [space] [space]C KT95 [space] [space] [space]

# Contr                  Ld         Decl         Tr         Score         Cross Imps         

1 6C         N         H3         Playe1         12         1370         15.27         15
2 3N         S         HT         Playe2         12         690         8.6         11
3 3N         S         H9         Playe3         11         660         8.27         11
4 3N         N         S4         Playe4         11         660         8.27         11
5 4N         S         HQ         Playe5         11         660         8.27         11
6 4N         N         S4         Playe6         10         630         7.6         11
7 6N         N         S4         Playe7         11         -100         -3.2         -6
8 6N         N         S4         Playe8         11         -100         -3.2         -6
9 6N         N         S4         Playe9         11         -100         -3.2         -6
10 6N         N         DT         Playe10         11         -100         -3.2         -6
11 6N         N         S4         Playe11         11         -100         -3.2         -6
12 6N         N         H3         Playe12         10         -200         -5.13         -8
13 6S         N         H3         Playe13         10         -200         -5.13         -8
14 7N         N         S4         Playe14         11         -200         -5.13         -8
15 6N         N         S4         Playe15         7         -500         -9.87         -8
16 6NXX         N         S2         Playe16         10         -1000         -15         -15

[matmat]

While we're at it, can we adjust for competent opponents?
I am tired of losing 13 imps when opps get to 6S= and the field is in 4S+2