Complication from fractional (weighted) scores

[szgyula]

Dear All,

 

This may be obvious but I could not find the solution: Team game (not BAM), TD assigns a weighted score to correct an infraction. The board result difference is 45 points. Law 78B tells you that 20-40 is 1IMP, 50-80 is 2. Assuming that all other boards are 0IMP, what is the result of the game (e.g. 16 boards) in IMP? What is the score in VP *new, 20VP scale)?

 

Keep in mind that the IMP to VP conversion includes the extra 0.01VP corrections (to make it monotonic) so you can not simply say that board result is 1.5IMP and VP is whatever the formula gives.

 

I know, this is such a rare case, but laws should be objective and clear...

 

Gyula

[pran]

Dear All,

 

This may be obvious but I could not find the solution: Team game (not BAM), TD assigns a weighted score to correct an infraction. The board result difference is 45 points. Law 78B tells you that 20-40 is 1IMP, 50-80 is 2. Assuming that all other boards are 0IMP, what is the result of the game (e.g. 16 boards) in IMP? What is the score in VP *new, 20VP scale)?

 

Keep in mind that the IMP to VP conversion includes the extra 0.01VP corrections (to make it monotonic) so you can not simply say that board result is 1.5IMP and VP is whatever the formula gives.

 

I know, this is such a rare case, but laws should be objective and clear...

 

Gyula

Different worlds may have different traditions, but in my world it is customary in such (and similar) situations to round up (i.e. 50 Points = 2 IMPS)

[WellSpyder]

Different worlds may have different traditions, but in my world it is customary in such (and similar) situations to round up (i.e. 50 Points = 2 IMPS)

I wonder what those situations are?

 

It looks to me like this problem is caused by the weighting being applied incorrectly. If the TD uses weights in adjusting the score, the imps from each alternative score should be calculated and then those imps weighted, rather than the raw scores being weighted before imping. So the result may well include fractional imps, but should not involve imping fractional scores.

[campboy]

WellSpyder is right. I suspect Pran was thinking about IMPing against a datum of 45 (for Butler or similar), rather than weighted scores.

[pran]

I wonder what those situations are?

 

It looks to me like this problem is caused by the weighting being applied incorrectly. If the TD uses weights in adjusting the score, the imps from each alternative score should be calculated and then those imps weighted, rather than the raw scores being weighted before imping. So the result may well include fractional imps, but should not involve imping fractional scores.

And even so, the final fractional IMPs be rounded up.

[campboy]

And even so, the final fractional IMPs be rounded up.

Not in the EBU:

Rounding at the end of a calculation is to be done as necessary to the nearest unit of scoring (see 8.12.3), with exact halves rounded away from average.

[pran]

Not in the EBU:

Sorry, I was imprecise - I was all the time thinking of rounding a calculated result exactly midway between two valid results.

[axman]

Not in the EBU:

White Book 4.2.5.1 said:

 

 

Rounding at the end of a calculation is to be done as necessary to the nearest unit of scoring (see 8.12.3), with exact halves rounded away from average.

 

As appealing as it seems at first glance, I suggest that it is misguided in that the law is badly constructed:

 

the imp table is discontinuous- for instance 20-40 =1imp while 50-80 = 2imp; there is a dead space of 40+ thru 50-.

 

if the value of 2imp requires at least 50 then anything less than 50 ought to be worth 1imp. As such, if we mean that 10- 44.99 =1imp then the law ought to so specify because the law's job is to specify the value of the imp.

[barmar]

The possibility of IMPing against a datum seems to be totally ignored by the Laws.

[pran]

The possibility of IMPing against a datum seems to be totally ignored by the Laws.

It is.

 

And with IMP across the field (which for all purposes is far better) there is no such problem.

[mycroft]

To be clear (because people are giving the right answer, but also conflating two or three issues):

 

Weighted scores are something like:

 

50% of 3NT-1, N/S -50

50% of 3NT=, N/S +400

 

This is not (-25 + 200) = 175. It's half the score for -50 plus half the score for +400.

 

So, assuming the other table is in 3♦+1, N/S +130, we don't get a difference of 45 to fail to imp, we get:

 

50% of -180, -5 IMP = -2.5 IMP

+ 50% of +270, 7 IMP = 3.5 IMP

= 1 IMP for the 3NT bidders.

 

On the other topic, there are many reasons why "IMP against datum" is deprecated. The fact that the IMP table doesn't handle non-bridge results is one of them (less important than the fact that the IMP table is carefully created to break on common bridge differences, and the datum blows all of that up, but still important).

[pran]

To be clear (because people are giving the right answer, but also conflating two or three issues):

 

Weighted scores are something like:

 

50% of 3NT-1, N/S -50

50% of 3NT=, N/S +400

 

This is not (-.25 + 200) = 175. It's half the score for -50 plus half the score for +400.

Oh yes, it is precisely the calculation (of the datum) for scoring "IMP against a datum"

 

 

So, assuming the other table is in 3♦+1, N/S +130, we don't get a difference of 45 to fail to imp, we get:

 

50% of -180, -5 IMP = -2.5 IMP

+ 50% of +270, 7 IMP = 3.5 IMP

= 1 IMP for the 3NT bidders.

 

On the other topic, there are many reasons why "IMP against datum" is deprecated. The fact that the IMP table doesn't handle non-bridge results is one of them (less important than the fact that the IMP table is carefully created to break on common bridge differences, and the datum blows all of that up, but still important).

Quite.

[szgyula]

Sorry for the delay (unexpected business trip)...

 

This method (convert to MP/IMP before weighting) still leaves an issue open: How to convert to VP? The new IMP-VP conversion clearly assumes integer IMP results (otherwise the 0.01 adjustments are pointles).

 

Off the record: which scoring programs handle the weighting issue correctly (especially cross imp events)? Just to be devil's advocate: What about more than one table having weighted scores in a cross imp event? How many programs handle that correctly (by hand it is close to impossible)?

 

Weighting the scores (e.g. +600, -50 to +325) certainly can make life much easier.

[RMB1]

To apply the VP scale you have to round fractional IMP to integers.

The WBF say to round the match result to nearest IMP, exact 0.5 away from zero. The EBU round the same but on individual boards.

Jeff Smiths pairsscorer does XIMP and weighted scores.