BBO Discussion Forums: Hand Evaluation - BBO Discussion Forums

bluejak, on 2012-February-14, 09:44, said:

I don't believe it's a question of 'good players' and 'lesser players'. After removing the large 'oblivious majority' (who come in all standards) It's rather a questions of 'mathematically inclined' players and 'non-mathematically-inclined' players. The ECL is an excellent example of this: it doesn't seem to have changed in the {censored} number of years since I played in it: the CUBC team (full of mathematicians, engineers, scientists) aren't very strong at bridge but don't like the form of scoring; the other 7 counties on average are rather better bridge players but less interested in exactly how the scoring system works. That's an argument for changing it: as most people don't care, you might increase overall happiness by pleasing those who do. But when I was an ECL player I was also, of course, much younger than all the other teams and already aware that the CUBC team were not exactly popular due to being young (and wanting to play all their matches at home).

An internet bridge forum will also have a high preponderance of the mathematically inclined. And an internet forum on bridge laws will automatically exclude the oblivious majority.

bluejak, on 2012-February-21, 13:53, said:

You must be the only person I have "met" that appears completely unable to understand that with cross-imps one compares a pair's score with the score of every other pair in the competition (individually) just as if that other pair had been "the other room" in an ordinary match for teams of four. Then the average of all these IMPs becomes the "normalized" cross-imp score for the pair in question.

The math in this process is that resulting IMPs are averaged exactly the same as when assigning a weighted adjusted score.

What needs some explanation is why averaging total-point scores rather than IMPs is OK for Butler scoring but absolutely not acceptable when assigning (weighted) adjusted scores?

Now I'm going to change the subject a bit, if anyone is still reading this thread, but it comes back to 'Laws & Rulings' at least.

The IMP scale and the VP scale are fundamentally not the same, because one is defined in the Laws & the other isn't.

Is either 'add up all the scores and then IMP them' or 'use the Trinidad/Rainsford adapted IMP scale' legal?
78B defines IMP scoring as "the total point difference between the two scores compared is converted into IMPs according to the following scale...."

This seems to allow cross-imps
If you add up all 4 scores first, you aren't comparing two scores.
If you multiple the scores by 0.7, you aren't comparing two scores.

I think in either case you have to rely on 78D which allows to Regulating Authority to approve other scoring methods.
In the EBU, the White Book refers in passing to add-up-all-4-scores and to Butler pairs in the context of how to calculate VPs but doesn't actually approve them.

Have the EBU "approved" any method of scoring that a Tournament Organiser might like to use? What about other RAs?

pran, on 2012-February-21, 17:00, said:

Don't be silly. We are not talking about what I find best. Ok, you may think that way, but that is what is wrong: you have no interest in doing the best for the customers, just what you personally want. Of course I understand and prefer Cross-imps, but I run events for customers and not for my own benefit.



Frances: all these forms of scoring, for example 2x to4, to8, x-imps for teams of eight, Butler, x-imps for imp pairs, are legal under Law 78D.

Whenever I hear a discussion such as this, I am reminded of the words of John Armstrong. He perfectly understood the logical and statistical arguments behind various forms of scoring; he also perfectly understood the practical reasons why certain changes to established scoring methods would be acceptable to the paying customers while others would not. "But", as he remarked to me after some perceived quirk in the scoring system had led to the elimination of his team from an important event, "truth is there's no point moaning - if you play well you win, and if you play poorly you don't".

bluejak, on 2012-February-21, 19:05, said:

I am not bothered with your view of what you find best, I really don't care.

But I am challenging your assertion that: Anyone who thinks weighted scores are similar to Cross-imps [...] has not thought it through (The class of player involved is irrelevant in this context.)

And I still wonder why Butler scoring is OK when it is totally unacceptable for awarding adjusted (weighted) scores, but I no longer expect any answer to that question.

I have explained, but of course you are not listening.

Players who are playing a form of imp scoring like to be able to calculate the imps themselves, as you can with Butler scoring or a weighted adjustment, but not with Cross-imps.

As to Armstrong's comment, with which I agree 100%, that is why the really important thing when considering these matters is to consider the likes and wishes of the lesser customers, not of clever people who understand the theory.

bluejak, on 2012-February-22, 08:52, said:

I'm not convinced you're right here - I think anyone who can calculate their weighted score themselves should also be able to cope with calculating their Cross-imp score themselves, it just might take a bit longer. Also, I think the number of "lesser customers" who care about being able to quickly calculate their scores by hand is actually not that dissimilar to the number of "clever people" who understand the theory and would prefer it to be done "correctly".

bluejak, on 2012-February-22, 08:52, said:

Do they? I can't say I've ever known anyone except David Martin (who was an EBU scoring buff) do this.

Similarly, although I've heard you say many times that less experienced players find it easier to understand Butler than Cross-IMPs, I've never heard them say it.

Running a weekly IMP game at our club, I do quite often get asked about how the scoring works and the difference between the two, and anyone who is really interested in the answer seems to understand. The thing some of them do have trouble with is the rationale for discarding one or two scores at either end in Butler.

bluejak, on 2012-February-22, 08:52, said:

This just doesn't add up.

Calculating a Butler score means adding all the other total-point scores, dividing by the number of scores and convert the difference between the resulting average and the own total-point score to IMPs.

Calculating (normalized) IMPs across the field means for each other total-point score to convert the difference between it and the own total-point score to IMPs, adding all these IMPs and then dividing by the number of scores.

I simply cannot accept an assertion that a person capable to do the one is not also capable of doing the other.

And the only form of "explanation" I have read from your posts is that Butler is (still) accepted because "most players prefer it" (an assertion I doubt if the player's were told what it is all about).

If this explanation is good then why is Butler-style weighting (calculating a total-score average which is then converted to IMPs) completely unacceptable when calculating an adjusted (weighted) score?

Edited to correct an error (the word "other" should not be there for Butler datum calculation)

FrancesHinden, on 2012-February-21, 16:52, said:

Sure, and that seems to be why all of this discussion has had to be based upon suppositions about some set of players who care a bit about the intricacies of the scoring system, but not that much. I've learnt something though, I would never have imagined anyone trying to re-calculate their scores in a pairs game if bluejak hadn't told us of his customers (who apparently don't visit other people's games).

I think this did blow up rather though, I don't think any of the CUBC team are very interested in how the scoring works, I just commented on it as an aside in my OP. It's just one of those things you notice, and realise inertia prevents making a change. Alerting of low-level doubles is a similar one: it's obviously wrong, but making a change would be harder work.

There does seem a lot of animosity towards the CUBC teams, and more so in the ECLs than otherwise. I've never understood why, and it seems unfortunate that a lot of the expression of this animosity comes about as a result of boards like the one in this thread. Most weeks, some pair in the opposition's A or B team will be playing some weird and wonderful bit of system (3 card 1M responses to natural 1m openings whenever you have 10+ and support for the minor being a personal favourite), fail to alert it, and a CUBC pair will enquire for a ruling on whether they were damaged. Generally the cases are more clearcut than this one, the captains/TDs in the teams will amicably reach a ruling, but the pair who failed to alert all too frequently take the request for a ruling as a personal insult and play the rest of the match clearly angry.

Granted, the situation described happens often enough anyway when a ruling is requested, and animosity towards junior pairs is common (unless your partner happens to be female...), but I've never understood why it seems to be so much worse in ECLs.

dburn, on 2012-February-21, 20:53, said:

While it is true that it mostly makes very little difference to the final result which method is used, it is possible for improving your own score on a board to actually reduce your position using Butler scoring. This is not just a theoretical possibility, but has been seen in high-level competition, when a pair who corrected a scoring error gained 30 points on the deal but lost a place (and prize money) as a result. I imagine they felt that moaning would be justified. Note that this only came to light because of the scoring error; similar situations where gaining 30 points would result in losing a place but no-one was aware of it are presumably rather more common than the single example would suggest. This situation is impossible with cross-imps, of course.

bluejak, on 2012-February-22, 08:52, said:

I once read this little thing in "The Best of Reader's Digest" (not my favorite publication, but sometimes everyone needs to visit a doctor). A mother of ten kids said that if one of them wouldn't listen, she would give him a spoonfull of liver oil. If all of them were not listening, she would take a spoonfull herself.

bluejak, on 2012-February-22, 08:52, said:

Can you tell me what the difference in complexity of these two calculations is?

I) Calculation of a weighted adjustment in IMPs:
The score at the other table was: +650 NS
And the ruling is:
40% -100
20% +620
20% +650
20% +710

II) Calculation of Cross-IMPs:
The score at our table was: +650 NS
The other scores are:
2x -100
1x +620
1x +650
1x +710

I would say that whoever is able to calculate the first will be able to calculate the latter.

This doesn't mean that this person is able to calculate a Butler score. For starters, I myself am perfectly capable of calculating the first two (5 IMPs), but I wouldn't even be able to calculate the datum for the Butler scoring (and my math skills are fine).

bluejak, on 2012-February-22, 08:52, said:

Of course, every player should try to win given the conditions of the game, which include the Laws and the method of scoring. No one has any right to blame the scoring method for their results: You knew the scoring method up front. But no one over here is complaining about their results, are they? They are pointing out peculiarities of scoring systems that they perceive as undesired. It may well be that "the complainers" might actually get a worse score if it would be recalculated in the way that the complainers would desire.
Actually, one would expect their scores to get worse. After all, "the complainers" are aware of the peculiarities and they will have adjusted their tactics accordingly. The "oblivious" will just use the team of four tactics. When the scores are recalculated, those who used the tactics according to the original scoring method will be disadvantaged.

Rik

It is all very well making up meaningless figures but let us look at the practicalities of what you and pran are saying.

You play in a Butler pairs of [say] 13 tables. At the end your score on [say] board 13 is +650. How many imps?

Well, the Butler score is +380 you are told, and you probably know, if you have played any imps at all, how to imp +650 against +380.

Now you play in a Cross-imps similarly. How do you work out your score? Well, since you have no idea what the scores are at the other tables, I have no idea.

As for going on about weighted scores, not only is it quite irrelevant, but when you get a weighted score, you know what it is being scored against, so it is just like a Butler, and in no way like a Cross-imps.

bluejak, on 2012-February-22, 18:51, said:

If the player is happy about being told that the Butler datum is +380 then fine for him, but in that case he should also be happy about being told that the normalized cross IMPs for him is (say) 9 IMPs, he doesn't calculate his IMP result in either case. (Looking up in a table for converting total-score differences to IMPs is not calculating!)

Now, if he cares about calculating his score he will need to know the 12 other total-point results on the board and either calculate the Butler datum by averaging the all thirteen total-point scores or figure out the 12 resulting IMPs and averaging these. I am surprised that you have no idea how to do this, I have no problem doing it (by hand if necessary). And the one is no more difficult than the other.

There is no way anybody can calculate his score without knowing the 12 other total-point scores whether it is Butler or Cross-IMPs.

"As for going on about weighted scores": Are you now saying that you know what it is being scored against, which can only mean that you calculate a weighted total-point score to obtain this datum? I fully agree that this is just like Butler, but I have had the impression that EBU (like for instance our Regulating Authority) requires the weighting to be done on the IMP results and not on the total-point scores. (Weighting on the IMPs makes it just like normalized Cross-IMPs.)

edited to correct an error (technical) in Butler datum calculation.

Trinidad, on 2012-February-22, 13:11, said:

Just for curiosity I have done the calculation in all three cases (yes there are three cases to consider!)

(Normalized) Cross-IMPs: 40% 13 IMPs and 20% each: 1 IMP, 0 IMPs and -2 IMPs = 5 IMPs
Butler datum no strikeouts: -100, -100, 620, 650, 650 and 710 = 2430/6 = 405 resulting in 6 IMPs
Butler datum 1 strikeout (the "standard"): -100, 620, 650 and 650 = 1820/4 = 455 resulting in 5 IMPs

(And the weighted score is exactly the same procedure as Cross-IMPs, resulting in 5 IMPs here)

Any problem? No

Edit: Yes, I suddenly realized that I had forgotten "own" result when calculating the butler datum, calculation corrected.

pran, on 2012-February-22, 22:57, said:

Q.E.D.

Does this show how hard it is to calculate a Butler score?

Anyhow, you did it wrong. For the calculation of the datum, you forgot to include our own score (which, of course, is an absurd requirement for Butler scoring, since who would want to IMP against his own score?*).

The correct (I hope) averages are:
No strikeouts: (2x-100+620+2x650+710)/6 = 2430/6 = 405
1 strikeout: (1x-100+620+2x650)/4 = 1820/4 = 455

I do not know what the datum should be in these cases, but that is probably my ignorance. At least I have a hunch that the datum is either 410 or 460, but it might be 400 or 450. I have an even bigger problem figuring what the datum would be if the board would be rotated 90 degrees (and the averages would be -405 and -455). Now, I don't even have a hunch anymore. The datum could be any of four.

I don't have these problems when calculating Cross-IMPs.

Rik

*I just realized that in a 3 table Butler there will always be one table that scores 0 IMPs. The other two scores are discarded and they are IMPing against themselves.

Trinidad, on 2012-February-22, 23:40, said:

Right,
I had just closed down and gone back to bed (it is now 6:45AM here) when I realised my error.
So I started up again and corrected it, hoping that I would manage to do so before anybody read my erratic calculation.

A three-table Butler is meaningless although mathematically possible.

Trinidad, on 2012-February-22, 23:40, said:

If you had to omit your own score, you'd have to calculate a different datum for each pair, which would be as much work as Cross-IMPs.

So while it may be mathematically absurd, it's required for practicality.

barmar, on 2012-February-23, 10:56, said:

I used to play several years ago at a club that committed an even greater mathematical absurdity in the interests of practicality - to make it really simple to calculate the "datum" on a board for the monthly "Butler pairs", they used to use the median score rather than the mean (you can probably guess this was in the days before computers were regularly used to score normal club events).

In this case, though, we gradually came to realise that the possibility of the scoring method meaning that ultimately you would be better off getting a "worse" score on the board was just too great. It didn't happen just at the margin, as with "proper" Butler pairs scoring, but time after time. If, for example, there were an odd number of tables, the rest of the field had already played the hand and there were only two potential results - let's say 12 tricks, but half the field bid game and the other half didn't - then it didn't affect your own score at all whether or not you bid the slam (either way your score would become the median, so imping against it would produce a zero swing). But your result did produce a huge swing for every other pair who had already played the board, and depending on which way your rivals were sitting for the hand you could significantly worsen your relative score by doing the "right" thing on the hand.

So ultimately we abandoned this approach and switched to the usual average approach to calculating the datum. I live in hope that eventually the equivalent logic, albeit at a far less serious level, will prevail in the case of Butler vs cross-imp scoring.