[pran]

pran, on 2015-January-09, 12:43, said:

It is easy to prove that Ascherman and Neuberg are identical.

VixTD, on 2015-January-09, 13:34, said:

If what you describe is Ascherman, then they are indeed identical and my recollection must be false, but in that case why are they referred to as if they were different?

Ascherman and Neuberg are indeed identical, and if you search on Google with keywords "ascherman formula bridge" you will find several articles confirming this.

I have no idea why Ascherman and Neuberg are referred to as if they were different, but I suspect that Neuberg was preferred by people knowing nothing better than to score matchpoints from zero (bottom) to the top.

Once one starts thinking of scores where zero is the average score and below and above average being represented by negative and positive values, then Ascherman has the clear advantage.

However, "Neuberg" has been well established, and very few (if any) apparently dare now speak of Ascherman even when it is his formula they really use with their scoring programs.

And not to forget: I believe it is far more easy for the layman to understand Ascherman than to understand Neuberg.

[Vampyr]

pran, on 2015-January-09, 15:16, said:

And not to forget: I believe it is far more easy for the layman to understand Ascherman than to understand Neuberg.

Vix's instructions make it obvious to the inexperienced director what he is trying to do and why.

[blackshoe]

http://www.bridgeguy...man_system.html

Attributed to a Dutchman named Ascherman in the 1950s, and written up by Herman DeWael in 1996.

Not sure if this is compatible with

Quote

Law 78A: In matchpoint scoring each contestant is awarded, for scores made by different contestants who have played the same board and whose scores are compared with his, two scoring units (matchpoints or half matchpoints) for each score inferior to his, one scoring unit for each score equal to his, and zero scoring units for each score superior to his.

[gordontd]

pran, on 2015-January-09, 12:43, said:

It is easy to prove that Asherman and Neuberg are identical.

That they give identical rankings when used on a single field does not make them identical.

[pran]

gordontd, on 2015-January-10, 02:53, said:

That they give identical rankings when used on a single field does not make them identical.

The mathematical proof is that the algebraic formula defining either method is (easily) transformed to the algebraic formula defining the other.

The main difference between Ascherman and Neuberg is that while Neuman operates on "bottom relative" values (i.e. between zero and TOP,) Ascherman operates on "average relative" values (i.e. between negative "half of TOP" and positive "half of TOP").

The formulae are identical in that they both "expand" "average relative" values by the ratio between original and new number of tables!

Ascherman does this directly, Neuberg must first transform the scores from "bottom relative" to "average relative", then "expand", and finally transform the scores back to "bottom relative".


Sounds Clear, doesn't it?

[pran]

blackshoe, on 2015-January-09, 19:14, said:

http://www.bridgeguy...man_system.html
Attributed to a Dutchman named Ascherman in the 1950s, and written up by Herman DeWael in 1996.
Not sure if this is compatible with

Quote

Law 78A: In matchpoint scoring each contestant is awarded, for scores made by different contestants who have played the same board and whose scores are compared with his, two scoring units (matchpoints or half matchpoints) for each score inferior to his, one scoring unit for each score equal to his, and zero scoring units for each score superior to his.

There is no reason why it should be compatible!

Law 78A deals with scoring within a single group of results. Ascherman and Neuberg handle the question on how to merge scores from two or more groups having played variants of the same boards into a single list of results.

[blackshoe]

pran, on 2015-January-10, 03:35, said:

There is no reason why it should be compatible!

Law 78A deals with scoring within a single group of results. Ascherman and Neuberg handle the question on how to merge scores from two or more groups having played variants of the same boards into a single list of results.

I got the impression from the article I linked that Ascherman, or at least part of it, can be used directly in normal scoring. IAC, the "average relative values" thing would be the sticking point, if I understand it correctly. For example, using this method, how many matchpoints on a particular board does a contestant get compared to each other contestant who beat his raw score? If it's not zero, it's not in accord with 78A as I read it.

[pran]

blackshoe, on 2015-January-10, 12:36, said:

I got the impression from the article I linked that Ascherman, or at least part of it, can be used directly in normal scoring. IAC, the "average relative values" thing would be the sticking point, if I understand it correctly. For example, using this method, how many matchpoints on a particular board does a contestant get compared to each other contestant who beat his raw score? If it's not zero, it's not in accord with 78A as I read it.

Law 78A describes "bottom relative" scoring. But only a fanatic SB would discard "average relative" scoring on this ground.

Direct "average relative" scoring is very simple: You gain one MP for each other raw score your raw score beats and lose one MP for each other raw score that beats your raw score. (You ignore every other raw score you match or with which you cannot compare - for whatever reason).


There is one consideration that requires a little brainwork: How much is 10% of a top? (Needed when awarding Ave+ or Ave- etc.). The answer is that it is 20% of the count of all other scores in the group.

[blackshoe]

pran, on 2015-January-10, 13:27, said:

Law 78A describes "bottom relative" scoring. But only a fanatic SB would discard "average relative" scoring on this ground.

Direct "average relative" scoring is very simple: You gain one MP for each other raw score your raw score beats and lose one MP for each other raw score that beats your raw score. (You ignore every other raw score you match or with which you cannot compare - for whatever reason).


There is one consideration that requires a little brainwork: How much is 10% of a top? (Needed when awarding Ave+ or Ave- etc.). The answer is that it is 20% of the count of all other scores in the group.

If I'm reading this right, 10% of a top in a ten table field is 1.8 matchpoints? So Ave+ would be +1.8 and Ave- would be -1.8?

This may or may not be easier than "bottom relative" scoring, but I'm sure of one thing: it's not going to happen in the ACBL.

[pran]

blackshoe, on 2015-January-10, 13:41, said:

If I'm reading this right, 10% of a top in a ten table field is 1.8 matchpoints? So Ave+ would be +1.8 and Ave- would be -1.8?

This may or may not be easier than "bottom relative" scoring, but I'm sure of one thing: it's not going to happen in the ACBL.

Isn't this rather simple for the players?
With "bottom relative" scoring you usually have no idea whether your MP score is poor, fair or good, you need the percentage value to know.
With "average relative" scoring you know that zero is average, minus is below and plus is above.

Your calculation was correct:
In a ten table field the "bottom relative" top is 18 MP, average is 9MP and Ave+ is 10,8 MP
The corresponding "average relative" scores are: Top is +9, average is 0 and Ave+ is +1,8 MP

ACBL has long ago ceased to surprise me, they still require their own version of the laws (as the only organisation in the world?).

(Anyway "average relative" has for many years now been the standard way of presenting MP results here in Norway and I seriously doubt we shall ever go back to "bottom relative".)

[gordontd]

pran, on 2015-January-10, 03:24, said:

The mathematical proof is that the algebraic formula defining either method is (easily) transformed to the algebraic formula defining the other.

The main difference between Ascherman and Neuberg is that while Neuman operates on "bottom relative" values (i.e. between zero and TOP,) Ascherman operates on "average relative" values (i.e. between negative "half of TOP" and positive "half of TOP").

The formulae are identical in that they both "expand" "average relative" values by the ratio between original and new number of tables!

Ascherman does this directly, Neuberg must first transform the scores from "bottom relative" to "average relative", then "expand", and finally transform the scores back to "bottom relative".


Sounds Clear, doesn't it?

Having looked into this a little more, it seems to me that while what you describe is a very neat implementation of Neuberg, it is not Ascherman as originally described by it's promoter Herman de Wael, since that does differ from Neuberg in making all scores closer to average.

I also found another document, which not only documents Ascherman but also mentions your variation on it.

[gordontd]

pran, on 2015-January-10, 14:38, said:

Isn't this rather simple for the players?
With "bottom relative" scoring you usually have no idea whether your MP score is poor, fair or good, you need the percentage value to know.
With "average relative" scoring you know that zero is average, minus is below and plus is above.

Now I think you are really over selling the method. When scores are not displayed as percentages they are usually shown with both match points side by side like this:

Quote

====================================================
BOARD 1
NS EW Contract Dec Lead NS+ NS- MP MP
====================================================
1...25... 3NT+3...S...S6...490............51...15

which makes it simple to see how good your score is, unlike:

Quote

====================================================
BOARD 1
NS EW Contract Dec Lead NS+ NS- MP MP
====================================================
1...25... 3NT+3...S...S6...490............18... -18

[pran]

gordontd, on 2015-January-11, 03:10, said:

Having looked into this a little more, it seems to me that while what you describe is a very neat implementation of Neuberg, it is not Ascherman as originally described by it's promoter Herman de Wael, since that does differ from Neuberg in making all scores closer to average.

I also found another document, which not only documents Ascherman but also mentions your variation on it.

I Accept that there may be variations of Ascherman, what I favour is apparently a variant which mathematically is proven equivalent to Neuberg, simpler to use and simpler to understand by the layman.

A bit puzzled by the assertion that Ascherman makes all score percentages closer to average I suspect this solely depends on the range within which the percentages are calculated, i.e. whether the (total) range used is twice the number of tables, this total reduced by one or reduced by two. I fail to see any significance in this question.

[pran]

gordontd, on 2015-January-11, 03:26, said:

[...]

And round reports (handed out to each pair after each round) with details and then summary like:

"Your result for the round was -2, Your current standing is +10"

is IMHO much easier to grasp than:

"Your result for the round was 34, Your current standing is 532"

[gordontd]

pran, on 2015-January-11, 03:50, said:

And round reports (handed out to each pair after each round) with details and then summary like:

"Your result for the round was -2, Your current standing is +10"

is IMHO much easier to grasp than:

"Your result for the round was 34, Your current standing is 532"

Ours give a lot more detail than that - rank, matchpoints, percentages and VPs if applicable.

[gordontd]

pran, on 2015-January-11, 03:41, said:

A bit puzzled by the assertion that Ascherman makes all score percentages closer to average I suspect this solely depends on the range within which the percentages are calculated, i.e. whether the (total) range used is twice the number of tables, this total reduced by one or reduced by two. I fail to see any significance in this question.

The significance is that a 68% result under one method is better than a 68% result under the other.

[pran]

gordontd, on 2015-January-11, 03:56, said:

Ours give a lot more detail than that - rank, matchpoints, percentages and VPs if applicable.

This is what a typical round report at events for pairs contains in Norway:

===============================================================
Pair; 10 Tom Smith - John Doe
Points: 80,5 53,6% Rank: 5 (-86,1)

Round/Table Ver Board Contract Lead Result Pts Round
10 / 3 N-S 7 46 5CX W -3 S5 500 17,0
47 2H E = SJ -110 -9,0
48 3C N = S3 110 7,0
49 4S S = DJ 420 11,0
50 3N S -1 H3 -100 -8,0 18,0
==============================================================

( "80,5" and "53,6%" refer to the current total for this pair
"-86,1" is how far they are behind the current leader)

We have had no negative comments on this, nobody seems to miss the information on current average or "bottom relative" scores

Edit: Sorry I was unable to make the presentation with proper spacing between fields

[pran]

gordontd, on 2015-January-11, 03:58, said:

The significance is that a 68% result under one method is better than a 68% result under the other.

Well, comparing apples with oranges should be discouraged.

[gordontd]

pran, on 2015-January-11, 04:40, said:

Well, comparing apples with oranges should be discouraged.

If, as you stated above, they are identical, they wouldn't be apples & oranges.

[pran]

gordontd, on 2015-January-11, 05:12, said:

If, as you stated above, they are identical, they wouldn't be apples & oranges.

They are identical so long as you think matchpoints, but if you suddenly begin calculating MP percentages off a different scale range then the calculated percentages will of course be different.