🚀 go-pugleaf

Keith Sheppard schrieb:

> This is a question about how you would like your computer scoring program to
> work.
>
> The scenario is that you are computer-scoring a pairs competition.  Because
> some boards have been played a different number of times, the program is
> using Neuberg adjustments to give competitors fractional scores on some
> boards.
>
> The program normally displays competitors' final percentages rounded to two
> decimal places.
>
> Suppose competitor A's final score (before rounding) works out to 48.445%.
> Competitor B's score is 48.450%.  With rounding, both come out at 48.45%.
>
> In the final result, would you adjudicate that these two competitors have
> tied (near as damn it)?  If not, are you happy to see B listed above A with
> both apparently on the same score?
>
> Maybe in an ideal world, the application should recognise this situation and
> extend the number of decimal places displayed, but that's extra code to deal
> with a very rare case and doesn't half muck up the output layouts.
>
> Do you know what your scoring application does in these circumstances?  Any
> views?
>
> Keith

In principle, you would want to list them in the right order but I am
pretty sure at least the Dutch and German regulations describe exactly
how you should round, and that would mean these two pairs will get an
equal score. So my suggestion is to check your local regulations and
build your program accordingly. If you want your program validated in
more countries you will have to put in country options.

Gerben

Keith Sheppard wrote:
> This is a question about how you would like your computer scoring program to
> work.
>
> The scenario is that you are computer-scoring a pairs competition.  Because
> some boards have been played a different number of times, the program is
> using Neuberg adjustments to give competitors fractional scores on some
> boards.
>
> The program normally displays competitors' final percentages rounded to two
> decimal places.
>
> Suppose competitor A's final score (before rounding) works out to 48.445%.
> Competitor B's score is 48.450%.  With rounding, both come out at 48.45%.
>
> In the final result, would you adjudicate that these two competitors have
> tied (near as damn it)?  If not, are you happy to see B listed above A with
> both apparently on the same score?
>
> Maybe in an ideal world, the application should recognise this situation and
> extend the number of decimal places displayed, but that's extra code to deal
> with a very rare case and doesn't half muck up the output layouts.
>
> Do you know what your scoring application does in these circumstances?  Any
> views?

Keith

The EBU white book gives a minimum unit of scoring for MP as 0.0001MP,
i.e. 4 decimal places.

"In general all calculations are to be performed to 4 decimal places
without any rounding during the course of the calculation. Rounding at
the end of a calculation is to be done as necessary to the nearest unit
of scoring, with exact halves rounded away from average.

Results may be displayed to fewer decimal places than the calculations
actually made, as is normal, for example, in MP Pairs."

Ties are only ties if they are equal to 4 decimal places.  If you
display results to only 1 or 2 decimal places, then you should display
the rankings "1=" or "1" and "2" in the output.  I suggest that when
there is an apparent tie (to the displayed accuracy) which is not a
tie, the program should produce a warning and the scorer should write
something on the results, e.g. the scores to 4 decimal places.

Robin

David Stevenson <bridge2@nospam.demon.co.uk> writes:
> Keith Sheppard wrote
> >This is a question about how you would like your computer scoring program to
> >work.
> >
> >The scenario is that you are computer-scoring a pairs competition.  Because
> >some boards have been played a different number of times, the program is
> >using Neuberg adjustments to give competitors fractional scores on some
> >boards.
> >
> >The program normally displays competitors' final percentages rounded to two
> >decimal places.
> >
> >Suppose competitor A's final score (before rounding) works out to 48.445%.
> >Competitor B's score is 48.450%.  With rounding, both come out at 48.45%.
> >
> >In the final result, would you adjudicate that these two competitors have
> >tied (near as damn it)?  If not, are you happy to see B listed above A with
> >both apparently on the same score?
> >
> >Maybe in an ideal world, the application should recognise this situation and
> >extend the number of decimal places displayed, but that's extra code to deal
> >with a very rare case and doesn't half muck up the output layouts.
> >
> >Do you know what your scoring application does in these circumstances?  Any
> >views?
> 
>   The EBU's view is that 48.450% is one place ahead of 48.445%, so
> whether you display more than two decimal places or not they are not
> equal.
> 
>         http://www.ebu.co.uk/laws_ethics/misc/2004whitebook.htm
> 
>   Note that while this is what they recommend - see #12.5 and #78.5.1
> below - they also make it clear that alternative methods are valid - see
> #78.6 below.
> 
> ===================================================================
> EBU White book
> 
> #12.5   "Standard amounts" for various methods of scoring
> 
> Method of scoring
>                  Standard adjustment       Minimum unit of scoring
>   Match Points       10% of top                  0.0001 mp
> 
> ===================================================================
> 
> #78.5.1  General
> 
>   In general all calculations are to be performed to 4 decimal places
> without any rounding during the course of the calculation.  Rounding at
> the end of a calculation is to be done as necessary to the nearest unit
> of scoring, with exact halves rounded away from average.
> 
>   Results may be displayed to fewer decimal places than the calculations
> actually made, as is normal, for example, in MP Pairs.
> 
>   Score changes which are discovered late will not be made if it is
> impracticable to change the score or if the score change in question
> would not make a meaningful difference.
> 
> ===================================================================
> 
> #78.6    General approach to scoring
> 
>   While there are certain statements of how scoring should be done
> software in use does not always follow this.  Anything that is in here
> about methods of scoring is a recommendation only.  If the scoring
> software in use does it differently that does not invalidate the result.
> At the time of writing EBU software did not follow all the
> recommendations.
> 
>   It is recommended that sponsoring organisations be consistent in their
> choice of software.

Strange regulations.  "Do this and that, but if you don't feel
like doing those, then do whatever you please."

Keith Sheppard wrote
>This is a question about how you would like your computer scoring program to
>work.
>
>The scenario is that you are computer-scoring a pairs competition.  Because
>some boards have been played a different number of times, the program is
>using Neuberg adjustments to give competitors fractional scores on some
>boards.
>
>The program normally displays competitors' final percentages rounded to two
>decimal places.
>
>Suppose competitor A's final score (before rounding) works out to 48.445%.
>Competitor B's score is 48.450%.  With rounding, both come out at 48.45%.
>
>In the final result, would you adjudicate that these two competitors have
>tied (near as damn it)?  If not, are you happy to see B listed above A with
>both apparently on the same score?
>
>Maybe in an ideal world, the application should recognise this situation and
>extend the number of decimal places displayed, but that's extra code to deal
>with a very rare case and doesn't half muck up the output layouts.
>
>Do you know what your scoring application does in these circumstances?  Any
>views?

  The EBU's view is that 48.450% is one place ahead of 48.445%, so
whether you display more than two decimal places or not they are not
equal.

        http://www.ebu.co.uk/laws_ethics/misc/2004whitebook.htm

  Note that while this is what they recommend - see #12.5 and #78.5.1
below - they also make it clear that alternative methods are valid - see
#78.6 below.

===================================================================
EBU White book

#12.5   "Standard amounts" for various methods of scoring

Method of scoring
                 Standard adjustment       Minimum unit of scoring
  Match Points       10% of top                  0.0001 mp

===================================================================

#78.5.1  General

  In general all calculations are to be performed to 4 decimal places
without any rounding during the course of the calculation.  Rounding at
the end of a calculation is to be done as necessary to the nearest unit
of scoring, with exact halves rounded away from average.

  Results may be displayed to fewer decimal places than the calculations
actually made, as is normal, for example, in MP Pairs.

  Score changes which are discovered late will not be made if it is
impracticable to change the score or if the score change in question
would not make a meaningful difference.

===================================================================

#78.6    General approach to scoring

  While there are certain statements of how scoring should be done
software in use does not always follow this.  Anything that is in here
about methods of scoring is a recommendation only.  If the scoring
software in use does it differently that does not invalidate the result.
At the time of writing EBU software did not follow all the
recommendations.

  It is recommended that sponsoring organisations be consistent in their
choice of software.

-- 
David Stevenson               Bridge   RTFLB   Cats  Railways    /\ /\
Liverpool, England, UK        Fax: +44 870 055 7697               @ @
<bridge2@blakjak.com>         ICQ 20039682     bluejak on OKB   =( + )=
           Bridgepage:        http://blakjak.com/brg_menu.htm      ~

"Gerben Dirksen" <gerben47@hotmail.com> schreef in
news:1133872519.647640.9050@o13g2000cwo.googlegroups.com:

>
> Keith Sheppard schrieb:
>
>> This is a question about how you would like your computer scoring
>> program to work.
>>
>> The scenario is that you are computer-scoring a pairs competition.
>> Because some boards have been played a different number of times, the
>> program is using Neuberg adjustments to give competitors fractional
>> scores on some boards.
>>
>> The program normally displays competitors' final percentages rounded to
>> two decimal places.
>>
>> Suppose competitor A's final score (before rounding) works out to
>> 48.445%. Competitor B's score is 48.450%.  With rounding, both come out
>> at 48.45%.
>>
>> In the final result, would you adjudicate that these two competitors
>> have tied (near as damn it)?  If not, are you happy to see B listed
>> above A with both apparently on the same score?
>>
>> Maybe in an ideal world, the application should recognise this
>> situation and extend the number of decimal places displayed, but that's
>> extra code to deal with a very rare case and doesn't half muck up the
>> output layouts.
>>
>> Do you know what your scoring application does in these circumstances?
>> Any views?
>>
>> Keith
>
> In principle, you would want to list them in the right order but I am
> pretty sure at least the Dutch and German regulations describe exactly
> how you should round, and that would mean these two pairs will get an
> equal score. So my suggestion is to check your local regulations and
> build your program accordingly. If you want your program validated in
> more countries you will have to put in country options.

I am not aware of any Dutch rules for rounding, and I can't find any in
the competitiereglement and wedstrijdreglement at www.bridge.nl. The
usual practice is to write down MP's with one decimal and percentages with
two decimals. But I don't think there is a rule against computer
programmes retaining more significant figures.

"Keith Sheppard" <keith.sheppard@tesco.net> wrote in message
news:1Aelf.12743$a15.10714@newsfe5-win.ntli.net...
> This is a question about how you would like your computer scoring program
> to work.
>
> The scenario is that you are computer-scoring a pairs competition.
> Because some boards have been played a different number of times, the
> program is using Neuberg adjustments to give competitors fractional scores
> on some boards.
>
> The program normally displays competitors' final percentages rounded to
> two decimal places.
>
> Suppose competitor A's final score (before rounding) works out to 48.445%.
> Competitor B's score is 48.450%.  With rounding, both come out at 48.45%.
>
> In the final result, would you adjudicate that these two competitors have
> tied (near as damn it)?  If not, are you happy to see B listed above A
> with both apparently on the same score?
>
> Maybe in an ideal world, the application should recognise this situation
> and extend the number of decimal places displayed, but that's extra code
> to deal with a very rare case and doesn't half muck up the output layouts.
>
> Do you know what your scoring application does in these circumstances?
> Any views?

Easiest thing for you to do as a programmer is to put a box on your setup
screen saying "results table rounded to ?? decimals",  calculate to whatever
accuracy the variable type you are using can handle, but then round the
results to whatever the user sets up in your setup screen.  Hence in your
example if the user selects 2 decimals you show both results as 48.45% and
their position as =8th but if the user selects 3 decimals show 48.500% &
48.445% and their position as 8th & 9th.

As others have said some regulators define the number of decimals that
should be used so maybe a drop down can be provided with EBU, ACBL etc as
choices and if the user chooses EBU you print to 4 places (assuming that is
the correct EBU calculation).

I would suggest if the position is based on 4 decimal places you should
print 4, but if it is based on 2 you should print 2 and thereby make it
clear how the result is arrived at.  However if matchpoints as well as
percentages are printed you can make it easier to read with a 2 decimals %
printout as long as the matchpoints are printed to an accuracy that makes it
clear why one pair is above another.

If 2 pairs are equal you need a simple rule to dictate order - I suggest
pair number.

Adding this functionality should not be a programming overhead.  I do not
know what language you use but most languages like C++/Delphi allow you to
define a const array for different format strings and then your print
command can just pick up the 1st/2nd/3rd array element according to how many
decimals the user selected on the setup screen.

Thanks David

I should have know that you would have the definitive answer.  Four decimal
places it is then.

Keith

>>I do not know what language you use but most languages like
>>C++/Delphi allow you to define a const array for different
>>format strings and then your print
The bulk of my application is in Visual Basic, although I drop into C++
whenever I really need to.

The main problem is lining stuff up in columns, particularly when printing
because then I'm dealing in actual pixel positions on the printed page.
It's not difficult, just cumbersome, to cope with varying column widths.

Regards
Keith

"Keith Sheppard" <keith.sheppard@tesco.net> wrote in message
news:gXilf.13785$a15.9370@newsfe5-win.ntli.net...
>>>I do not know what language you use but most languages like
>>>C++/Delphi allow you to define a const array for different
>>>format strings and then your print
> The bulk of my application is in Visual Basic, although I drop into C++
> whenever I really need to.
>
> The main problem is lining stuff up in columns, particularly when printing
> because then I'm dealing in actual pixel positions on the printed page.
> It's not difficult, just cumbersome, to cope with varying column widths.
>
> Regards
> Keith

I only use VB for Macros in Microsoft Office applications so I know little
about it but can't you solve this by setting tab positions and printing at
the tab stops?  Then you just adjust the tabs once at the beginning of the
code according to the user settings chosen but all you print code is
unchanged by the decimals chosen or what that means to the position of your
tab stops.

Anyway from what others have said it looks like you should print MP's to 4
decimals so everyone can see the correct score used to define the results
but the right column should be a % result to 2 decimals so it is easy to
read and obvious that two equal percentages may not be 2 equal scores.

>>I only use VB for Macros in Microsoft Office applications so I know little
>>about it but can't you solve this by setting tab positions and printing at
>>the tab stops?
I don't know about that but my printing code does pixel-accurate positioning
of text, grid lines etc.  Maybe it is overly complex but I don't really feel
like rewriting it at this late stage.

Keith

Everett M. Greene wrote

>Strange regulations.  "Do this and that, but if you don't feel
>like doing those, then do whatever you please."

   Same sort of approach as in North America, for example, the difference
being that we say so.

--
David Stevenson               Bridge   RTFLB   Cats  Railways    /\ /\
Liverpool, England, UK        Fax: +44 870 055 7697               @ @
<bridge2@blakjak.com>         ICQ 20039682     bluejak on OKB   =( + )            Bridgepage:        http://blakjak.com/brg_menu.htm      ~