imp scoring

[Flame]

As far as i know (scoring is one of my weak spots) when making imp results its normal to exclude some extreme results of both sides for the avarage calculation, for example not calculate the best and worth result for NS. I noticed that this isnt done in BBO, and i think its a mistake, especially on a field of large veraity of players level with many pickuped partnerships, there are times when an extreme result ruin the results in all other tables.

sometime its a bridge stupidity like a redouble contract down alot, but sometime it can even be a non bridge like when opponents both leave a table and the couple who left claim all tricks and no one is there to reject the claim.

I think its best to exclude X% of the results.

[1eyedjack]

There are two basic methods of IMPing a large field: (1) identify a "datum" or "average" score and then imp all of the results against that average, or (2) calculate the net sum of each pair's IMP score against every other pair.

 

Method (1), commonly known as "Butler" scoring, is popular in a manually scored face-to-face event because it involves less calculation. In order to identify a sensible datum the extreme scores are excluded.

 

When the entire movement and scoring are handled by computers with virtually limitless number-crunching capacity that restriction is lifted and it becomes no more onerous to do a true IMP comparison against every table, which is regarded as a more accurate and fair calculation by the statistitions (of which I am not one, so I just accept it).

[Flame]

There are two basic methods of IMPing a large field: (1) identify a "datum" or "average" score and then imp all of the results against that average, or (2) calculate the net sum of each pair's IMP score against every other pair.

 

Method (1), commonly known as "Butler" scoring, is popular in a manually scored face-to-face event because it involves less calculation. In order to identify a sensible datum the extreme scores are excluded.

 

When the entire movement and scoring are handled by computers with virtually limitless number-crunching capacity that restriction is lifted and it becomes no more onerous to do a true IMP comparison against every table, which is regarded as a more accurate and fair calculation by the statistitions (of which I am not one, so I just accept it).

Thxs.

I guess you are right.

[hotShot]

Hi!

 

If you score IMP's you have go 2 ways to get the average result:

 

1) You simply calculate the average score for each board scoring every result.

 

2) You cut a number (n) of scores at the top and the bottom and average the rest.

 

The main difference is, that if you are using the second way, the average result of every board should give 0 IMP's.

 

Example we have:

 

4 * 420 and -50

 

Method 1) gives you an average of: 326

So here everybody gets some big +/-IMP scores.

 

Method 2) with n=1 gives an average of: 420

420 leads to 0 IMP's

-50 leads to very big IMP's

 

So with method 2 having 0 IMP's in the end you know that you have a middle position.

 

With method 1 the middle position can have some value different from 0 you don't know.

 

hotShot

[1eyedjack]

Hi!

 

If you score IMP's you have go 2 ways to get the average result:

 

1) You simply calculate the average score for each board scoring every result.

 

2) You cut a number (n) of scores at the top and the bottom and average the rest.

 

The main difference is, that if you are using the second way, the average result of every board should give 0 IMP's.

 

Example we have:

 

4 * 420 and -50

 

Method 1) gives you an average of: 326

So here everybody gets some big +/-IMP scores.

 

Method 2) with n=1 gives an average of: 420

420 leads to 0 IMP's

-50 leads to very big IMP's

 

So with method 2 having 0 IMP's in the end you know that you have a middle position.

 

With method 1 the middle position can have some value different from 0 you don't know.

 

hotShot

I don't think it is that simple. Method 2 distorts the intended logarithmic effect of the IMP scale.

 

To demonstrate this consider an example where excluding the extreme scores makes no difference to the "Butler" datum used in method 2.

 

Consider a board that is played 10 times. On each occasion the contract is 4♠ undoubled by South, vulnerable. On 5 occasions the contract makes on the nose. On the other 5 occasions the contract is defeated by one trick.

 

If you try to identify a "Butler" datum by averaging the scores then the average absolute score is +285 to N/S (and in this example excluding extreme scores makes no difference). Each pair's result differs from that datum by +/-335, for which the IMP score is +/- 8 IMP

 

If instead you score up each comparison as a separate match and average the result you get, for each pair, 4 flat scores and 5 x 12 IMP swings, giving an average of +/- 6.7 IMP.

 

Which result do you reckon more closely resembles the intention of the IMP scoring system? I vote for method 1.