Lorne
2004-04-01 21:40:06 UTC
If some boards are played less often than others in a movements it is
unfair on those playing the less played boards and some factoring of scores
is required. In the days of hand scoring each board was correctly scored,
and each pair got a % score equal to actual MP's divided by the max they
could have got and factoring was achieved because the max possible score for
a pair reflected which boards they played.
Some computer software, including the EBU program, does not work like that.
It takes the actual MP's for the boards played less often and factors each
score up. This seems OK but the results are not what you first expect - for
example if some boards get played 13 times so a top=24, but others are
played 12 times so a top=22 you would expect each score on boards played 12
times to be factored up by 24/22. The table below shows possible scores on
a board played 12 times, the "expected" factored score, and what the
software uses:
MPs (hand played 12 times) Factored to 13 times Results from EBU
software
0 0.00 0.1
1 1.09 1.2
2 2.18 2.3
3 3.27 3.3
4 4.36 4.4
5 5.45 5.5
6 6.55 6.6
7 7.64 7.7
8 8.73 8.8
9 9.82 9.8
10 10.91 10.9
11 12.00 12
12 13.09 13.1
13 14.18 14.2
14 15.27 15.2
15 16.36 16.3
16 17.45 17.4
17 18.55 18.5
18 19.64 19.6
19 20.73 20.7
20 21.82 21.7
21 22.91 22.8
22 24.00 23.9
As you can see an outright top on a board played 12 times only gets 23.9
MP's rather than the expected 24 (ps it may calculate to more decimal places
but only 1 place is printed out).
Is this deliberate?
It may be deliberate because a top is harder to get the more often a board
is played. If you have a top on a board played 12 times and it gets played
1 more time I think you have a 92.3% chance of keeping the top so the
software may be saying your fair score when factored up = 92.3% of 24 +
(say) 3% of 23 for a shared top + (say) 4.7% of 22 when the last pair beat
you = 23.88. Obviously how you split up 7.7% between a shared top and a
second is open to debate and my figures just illustarte a point, but if this
is the methodology working out an algorythm that works for all the
complicated scenarios when you start with 12,13,14 etc. MP's is very
complex.
Does anyone know what the correct way to do it is? Is there a published
methodology? Is the EBU program doing what people expect?
I did ask the same question of the EBU but nobody replied!
unfair on those playing the less played boards and some factoring of scores
is required. In the days of hand scoring each board was correctly scored,
and each pair got a % score equal to actual MP's divided by the max they
could have got and factoring was achieved because the max possible score for
a pair reflected which boards they played.
Some computer software, including the EBU program, does not work like that.
It takes the actual MP's for the boards played less often and factors each
score up. This seems OK but the results are not what you first expect - for
example if some boards get played 13 times so a top=24, but others are
played 12 times so a top=22 you would expect each score on boards played 12
times to be factored up by 24/22. The table below shows possible scores on
a board played 12 times, the "expected" factored score, and what the
software uses:
MPs (hand played 12 times) Factored to 13 times Results from EBU
software
0 0.00 0.1
1 1.09 1.2
2 2.18 2.3
3 3.27 3.3
4 4.36 4.4
5 5.45 5.5
6 6.55 6.6
7 7.64 7.7
8 8.73 8.8
9 9.82 9.8
10 10.91 10.9
11 12.00 12
12 13.09 13.1
13 14.18 14.2
14 15.27 15.2
15 16.36 16.3
16 17.45 17.4
17 18.55 18.5
18 19.64 19.6
19 20.73 20.7
20 21.82 21.7
21 22.91 22.8
22 24.00 23.9
As you can see an outright top on a board played 12 times only gets 23.9
MP's rather than the expected 24 (ps it may calculate to more decimal places
but only 1 place is printed out).
Is this deliberate?
It may be deliberate because a top is harder to get the more often a board
is played. If you have a top on a board played 12 times and it gets played
1 more time I think you have a 92.3% chance of keeping the top so the
software may be saying your fair score when factored up = 92.3% of 24 +
(say) 3% of 23 for a shared top + (say) 4.7% of 22 when the last pair beat
you = 23.88. Obviously how you split up 7.7% between a shared top and a
second is open to debate and my figures just illustarte a point, but if this
is the methodology working out an algorythm that works for all the
complicated scenarios when you start with 12,13,14 etc. MP's is very
complex.
Does anyone know what the correct way to do it is? Is there a published
methodology? Is the EBU program doing what people expect?
I did ask the same question of the EBU but nobody replied!