[ckmooring]

don't you think we would get better results if the top and bottom board of each deal were eliminate?

[mgoetze]

Isn't that called matchpoints?

[manudude03]

Enjoy butler IMPs, it isn't played much.

[matmat]

yes, and no.

Ideally the software would toss the joke results, i.e. the spite 7NTXX-13 boards, but it seems bad to penalize someone (or give credit to the rest of the field) for making slam on some exotic squeeze when the remainder of the declarers went off.

[manudude03]

Maybe an improvement would be to ignore the 2 extreme results when scoring up without eliminating it completely. The problem with that is in the example matmat gave, if everyone was in 6NT, one pair made it and one went -2 (the other 14 going -1), a pair going -1 would just get a flat board when they should be losing about an imp.

[Siegmund]

Gotta define "better."

If you mean in the statistical sense of estimating as accurately as possible what the distribution of results on the board should be if played many times, no, throwing out results is a Bad Thing.

If you mean some very specialized sense like trying to estimate what par should be on a board, you can make a case for it, but you can make even better cases for all of means, medians, and modes than you can for trimmed means, except in a few really odd situations.

[gwnn]

manudude03, on 2011-January-06, 18:19, said:

Maybe an improvement would be to ignore the 2 extreme results when scoring up without eliminating it completely. The problem with that is in the example matmat gave, if everyone was in 6NT, one pair made it and one went -2 (the other 14 going -1), a pair going -1 would just get a flat board when they should be losing about an imp.

why should they be losing about an imp? the most likely explanation* for this traveller is that:
at one table defence was incredibly stupid and let the contract through
at one table declarer was incredibly stupid and did not even take the 11 tricks they had to take

so why "should" the normal non-stupid people lose an imp because of two or three incredibly stupid people?

*but by no means the only explanation, of course it could be that for example the declarer who went down 2 was playing the best line, or that the declarer who made it made it legitimately and the rest made the mistake

[TylerE]

gwnn, on 2011-January-07, 05:55, said:

so why "should" the normal non-stupid people lose an imp because of two or three incredibly stupid people?

Think of it more as rewarding the defenders who didn't muck it up.

[gwnn]

So the declarers are punished because they didn't play against big idiots?

Anyway I am not even in favour of Butler scoring (it seems artificial to decide that the top 1 and bottom 1 are "bad" scores - why not the top 25%? or top 49%?) - just asking manudude why he's being categorical.

It is my personal belief that you can't say anything certain about what you "deserved" or "should have got" or so on, in bridge, or football, or figure skating etc.

[pooltuna]

ckmooring, on 2011-January-06, 17:39, said:

don't you think we would get better results if the top and bottom board of each deal were eliminate?

I would be more interested in seeing the results of removing each pair's top and bottom score

[hrothgar]

Is there any reason that we limit ourselves to a single score type?

In theory, there's no reason why BBO could display both

1. The normal IMP score
2. The Butler IMP score

I understand that certain scoring types actually significant distort style of play
Its entirely possible that the "optimal" line for a given hand might be different at IMPS, BAM, and MP.

However, other than additional development time/complexity I don't see any real reason not to simulatenously display multiple different score metrics.

With all this said and done, there's a hell of a lot of stuff that I'd rather see than changes to the scoring system

[mgoetze]

hrothgar, on 2011-January-07, 09:29, said:

With all this said and done, there's a hell of a lot of stuff that I'd rather see than changes to the scoring system

And if we're making any changes to the scoring system, I think upping the number of boards compared from 16 to something higher for the MBC would be a good start.

[hotShot]

It would be nice that, if you get the "par" result, both sides at your table get 0 IMPs.
Unfortunately there is often some lunatic result that shifts both sides IMP scores apart.
There is no solution to this, because there are boards that don't have a "par" score, but the problem gets smaller if more scores a taken into consideration.

Unfortunately the Windows client can only handle 16 scores playing in the MBC. This restriction is a heritage from the time when the number of players logged to BBO had only 2 digits and not 5 as it is now. So unless the windows client is dropped or updated(there is no plan to do that), it will not get more.

[matmat]

hotShot, on 2011-January-07, 10:37, said:

There is no solution to this, because there are boards that don't have a "par" score, but the problem gets smaller if more scores a taken into consideration.

double.

[barmar]

gwnn, on 2011-January-07, 07:32, said:

So the declarers are punished because they didn't play against big idiots?

That's how duplicate bridge works. If someone gets a gift, everyone else playing that direction is "punished" just as much as if the pair with the good result had done it by finding a great squeeze. Matchpoints and IMPs are based on the score, not how it was achieved.

[FrancesHinden]

gwnn, on 2011-January-07, 07:32, said:

So the declarers are punished because they didn't play against big idiots?

Anyway I am not even in favour of Butler scoring (it seems artificial to decide that the top 1 and bottom 1 are "bad" scores - why not the top 25%? or top 49%?) - just asking manudude why he's being categorical.

It is my personal belief that you can't say anything certain about what you "deserved" or "should have got" or so on, in bridge, or football, or figure skating etc.

Really? I've always found that when I win, I deserved it because I played well. When I don't win, I was either unlucky (pairs), or my teammates had a bad day (teams). Seems to be true all the time.

[gwnn]

barmar, on 2011-January-08, 22:24, said:

That's how duplicate bridge works. If someone gets a gift, everyone else playing that direction is "punished" just as much as if the pair with the good result had done it by finding a great squeeze. Matchpoints and IMPs are based on the score, not how it was achieved.

Like I said, I wasn't arguing for or against any particular scoring system. All I took exception to was why manudude seems to think that one side of the tables who got 6NT-1, an apparently 100% normal result, "should get" 1 IMP and the other "should get" -1 IMP. It is not clear to me that this is true. In fact, it seems to be that the whole idea of Butler is to give 0 IMP's to all these pairs because the 6NT-1 result is so normal that no one ought to get any imps for it. So it seemed to me that manudude's argument was "Butler is bad because pairs should get the scores from ximps, not Butler", an apparently harsh and categorical statement. It might be, however, that I missed his argument on why they "should get" 1 IMP, or I missed someone else's argument. I still don't understand exactly what merit in bridge is, other than in the long run when one plays worse bridge one gets worse results, cheaters never prosper, and the force be with you.

[barmar]

gwnn, I think maybe we're interpreting "should get" differently. You seem to be interpreting it as meaning that they "deserve" to lose an IMP. I think manudude just meant that it's more consistent with the spirit of IMP scoring.

[mycroft]

And why should defenders be punished because we are the only ones in the room to have a blackwood error and bid grand off a finessible trump king? Because that's the way it works. Swings and roundabouts.

Butler scoring is bad because improving your bridge score can decrease your IMP result. The question of whether to throw out outliers for any calculation (Butler datum or cross-imp comparison, or matchpoints, for that matter) is a different question.

I agree that more results to compare against is better, and that 16 is too few, and that Fred has explained why before and why it's not going to change soon before as well. But IMP pairs is a crapshoot to begin with, in the MBC the craps table is slightly flatter than a mountain range, and it isn't the world championships, or qualifying for it. Anything from +2 to -2 IMPs is a zero. Deal with it.

[gwnn]

barmar, on 2011-January-10, 19:39, said:

gwnn, I think maybe we're interpreting "should get" differently. You seem to be interpreting it as meaning that they "deserve" to lose an IMP. I think manudude just meant that it's more consistent with the spirit of IMP scoring.

It's more consistent with the spirit of XIMP scoring, while it's less consistent with the spirit of Butler scoring.

Would you agree with this:

The problem with Matchpoints is that when the whole room has -140 and I have -150, then I get a zero, but I should get about average.

?