[WellSpyder]

szgyula, on 2015-December-03, 03:29, said:

All I ask is a simple quote from anything that explains how to convert IMP to VP that can be applied. So far I saw "VP does not have to add up to 20". There is a huge leap from this to the actual method to be used.

But the regulations already tell you what VPs are awarded for a given net imp score. So just apply this for each team, and there is your "actual method".

[barmar]

A typical VP chart has a row like this:

IMPS VPs
5-7 13-7

The ways this is shown is for the normal case where VPs balance. But it can also be read as: +5 through +7 IMPs gets 13 VP, -5 through -7 IMPs gets 7 VPs. When dealing with split scores, you would use the latter method.

[szgyula]

barmar, on 2015-December-05, 04:40, said:

A typical VP chart has a row like this:

IMPS VPs
5-7 13-7

The ways this is shown is for the normal case where VPs balance. But it can also be read as: +5 through +7 IMPs gets 13 VP, -5 through -7 IMPs gets 7 VPs. When dealing with split scores, you would use the latter method.

This question was more along the lines of "what is the order of differnt operations"? You have "split score", you have "weighting" and you have "IMP to VP". Apparently the correct order is:

1. Weighting
2. IMP to VP
3. Split score

To me this appears to be a "case law", i.e. somebody started to do it this way and we stick to that lacking any compelling reason to do something different, e.g. to calculate the weighting in VP.

[mycroft]

If we have a split and weighted score, we IMP and convert to VP all splits and alternative scores, and apply the assigned weighting.

For instance (totally ignoring the hand in question):

All other boards of the match, Team A is +6 IMPs. Scoring is on the 6-8 board, ACBL 20 point VP scale (because that's the one I know).

Board 7, in the open room, N/S A is 4S+2, +680.
In the closed room, the table result is E/W A 5Dx-2, N/S +500.
However, there's a MI ruling. With the correct information, N/S are never sitting, and we think finding the slam is an outside chance, but not likely. The TD decides to assign:
30% N/S B 6S=, +1430;
70% N/S B 5S+1, +680.
However, again, it is deemed that South doubled as a Gambling Action, "knowing" that she'd been given MI, and counting on the TD to recover to however many spades makes if she didn't take 5♦ for enough. So, E/W get the assigned score, and N/S get something. I'm not an expert at this (by any stretch!), so I'm sure my suggested assigned score is unreasonable, but let's say relief for that part of the damage that was not self-inflicted is deemed to be:
30% N/S B 6S=, +1430;
70% E/W A 5Dx-2, N/S +500.

So:
E/W get a match result of: 30% of -7 (+6 -13 for the slam swing) and 70% of +6: 30% of 7 + 70% of 13 = 2.1 + 9.1 = 11.2 VPs.
N/s get a match result of: 30% of +7 and 70% of -11 = 30% of 13 + 70% of 5 = 3.9 + 3.5 = 7.4 VPs.
Yes, the scores don't add up to 20. This isn't a problem, both sides were "offending".

[szgyula]

mycroft, on 2015-December-08, 18:29, said:

So:
E/W get a match result of: 30% of -7 (+6 -13 for the slam swing) and 70% of +6: 30% of 7 + 70% of 13 = 2.1 + 9.1 = 11.2 VPs.
N/s get a match result of: 30% of +7 and 70% of -11 = 30% of 13 + 70% of 5 = 3.9 + 3.5 = 7.4 VPs.
Yes, the scores don't add up to 20. This isn't a problem, both sides were "offending".

Here you did the weighting in IMP. You could have done it in VP. For the discrete, 30VP scale that would be very difficult of course. For the new, 20VP scale that is trivial. You also reduce the rounding errors. If you weight in IMP, you can introduce 0.5IMP errors (the 20VP scale needs integer IMP results). If you weight in VP, you introduce errors at most 0.005VP. This is already "accepted" as there are 0.01VP corrections in the formula to make the curve behave nicely: As you move away from a tie, each IMP step must result in a non increasing VP step, even after rounding to 0.01VP. This requires small corrections...

I am not saying this is better. I am not saying it should be done this way. I am just saying that this is more like case law than written law. Historically weighting is done in IMP, full stop.

Rounding is also an issue that must be clearly regulated.

[mycroft]

I don't think I did what you said I did - at least there was nowhere that I "could introduce half-IMP errors". For each assigned result, I:

• calculated the (whole-)IMP total on the board against the result at the other table (and I did it with bridge results, so there's no "compare against 427 datum" problem either);
  
• Added that result to the rest of the match to get a (whole-)IMP result of the match;
  
• converted that result to VPs according to my table;

and then applied the weightings assigned for each result.

If you use the graduated VP scale the same calculation gives the same results - with different numbers of course. Using the USBF 8-board scale the calculation becomes:

E/W get a match result of: 30% of -7 and 70% of +6: 30% of 7.23 and 70% of 12.42 = 10.863 VPs
N/S get a match result of: 30% of +7 and 70% of -11: 30% of 12.77 and 70% of 5.91 = 7.968 VPs.

But the USBF table is friendlier to the loser than the ACBL 20-point (blitz being 43 instead of 28 may have something to do with it).

[szgyula]

mycroft, on 2015-December-09, 11:53, said:

I don't think I did what you said I did - at least there was nowhere that I "could introduce half-IMP errors". For each assigned result, I:

• calculated the (whole-)IMP total on the board against the result at the other table (and I did it with bridge results, so there's no "compare against 427 datum" problem either);
  
• Added that result to the rest of the match to get a (whole-)IMP result of the match;
  
• converted that result to VPs according to my table;

and then applied the weightings assigned for each result.

If you use the graduated VP scale the same calculation gives the same results - with different numbers of course. Using the USBF 8-board scale the calculation becomes:

E/W get a match result of: 30% of -7 and 70% of +6: 30% of 7.23 and 70% of 12.42 = 10.863 VPs
N/S get a match result of: 30% of +7 and 70% of -11: 30% of 12.77 and 70% of 5.91 = 7.968 VPs.

But the USBF table is friendlier to the loser than the ACBL 20-point (blitz being 43 instead of 28 may have something to do with it).

Correct, I missed that. Indeed, you did the weighting in VP, which goes against everything people said so far. Which supports my claim: there is no clear regulation and there should be one.

Of course this assumes that people can agree on something first and everybody can accept that.

[gordontd]

szgyula, on 2015-December-09, 01:15, said:

Rounding is also an issue that must be clearly regulated.

Sure. In the EBU we round up exact halves. In the EBL when I had such a case, I was told just to change the percentages a bit so that I no longer had a half!

[gordontd]

szgyula, on 2015-December-10, 15:52, said:

Correct, I missed that. Indeed, you did the weighting in VP, which goes against everything people said so far. Which supports my claim: there is no clear regulation and there should be one.

Of course this assumes that people can agree on something first and everybody can accept that.

The basic method of scoring a board is IMPs. VPs are the method of scoring the whole match. If I ask you how many VPs you scored on board 5, you couldn't answer. That's why, if we are trying to give board 5 a weighted score, we need to do it in IMPs.

[szgyula]

gordontd, on 2015-December-11, 02:34, said:

The basic method of scoring a board is IMPs. VPs are the method of scoring the whole match. If I ask you how many VPs you scored on board 5, you couldn't answer. That's why, if we are trying to give board 5 a weighted score, we need to do it in IMPs.

Let me quote 12C1(b):

"If, subsequent to the irregularity, the non-offending side has contributed to its own damage by a serious error (unrelated to the infraction) or by wild or gambling action it does not receive relief in the adjustment for such part of the damage as is self-inflicted."

The question is the definition of "damage" in this context. You can define the damage in VP, IMP or even total score. All three are valid, well defined and easy to calculate. You still have to pick one to calculate "such part of the damage". As neither the total score-IMP nor the IMP-VP transformation is linear, the choice does matter. I claim that the choice does not follow from the Laws but there is a precedent to use IMP, i.e. it is a case law. I suggest making this a written law.

It is like reading tea leaves but there is 86D:

"In team play when the Director awards an adjusted score (excluding any award that ensues from application of Law 6D2), and a result has been obtained* between the same contestants at another table, the Director may assign an adjusted score in IMPs or total points (and should do so when that result appears favorable to the non-offending side)."

you may argue that the MAY here means free choice between total score and IMP (i.e. excludes VP) but you can also argue that these (total score, imp) are just options. Using the "Expressio unius est exclusio alterius" and "Ejusdem generis" concepts from roman era favors the first interpretation.

Ceterum censeo: Improve the laws so this is settled for all and not only for people with a PhD in law or people initiated by some Guru in the correct interpretation of a vague law. Rounding is even worse (EBL actually has statements about this in some CoC: round toward NOS).

[gordontd]

szgyula, on 2015-December-11, 03:55, said:

I claim that the choice does not follow from the Laws but there is a precedent to use IMP, i.e. it is a case law. I suggest making this a written law.

It may be based on precedent but the fact that IMPs & total points are mentioned in this context but VPs are not is certainly relevant. I was simply giving one other explanation as to why things might have developed the way they have - namely that it's based on the basic method of scoring a board rather than the whole match. By all means try to get it in the laws, but this discussion is going nowhere because you just keep repeating yourself and clearly have no interest in listening to anyone with a different view from yours. I suppose the warning was in your original post when you said "I have a long argument with the national authorities and we seem to disagree fundamentally".

[RMB1]

gordontd, on 2015-December-11, 02:29, said:

Sure. In the EBU we round up exact halves. In the EBL when I had such a case, I was told just to change the percentages a bit so that I no longer had a half!

As part of the definition of the WBF "continuous" VP scale, it assumes fractional IMPs are be retained on individual boards but then the match result is rounded before conversion to VP: exact halves in the match result are rounded away from zero.

(GordonTD: mamosTD and I had to dig this out after a weighted ruling in a Camrose event.)

[szgyula]

gordontd, on 2015-December-11, 04:31, said:

By all means try to get it in the laws, but this discussion is going nowhere because you just keep repeating yourself and clearly have no interest in listening to anyone with a different view from yours. I suppose the warning was in your original post when you said "I have a long argument with the national authorities and we seem to disagree fundamentally".

It is clear from the beginning that you have a fundamental issue with this discussion because you assume something about my hidden reasons, motivation, etc. You keep derailing the discussion through attacking me and my motivation. Fine. Your choice.

I will offer one more argument for those who are interested: Rounding. If you work in IMPs, you sooner or later have to deal with fractional IMPs. Just the way you can not convert 15 Total Points to IMP, you can not handle fractional IMPs. The Total Points are assumed to be integer multiples of 10. The Total Point to IMP conversion needs this. The IMPs are assumed to be integers. The IMP to VP scale (new one) needs this. Thus, there is a need to round the numbers.

In the past, using the 30VP scale, this problem would have lead to a disaster. If you did the weighting in VP, you ended up with fractional VPs, which you had to round. Thus, in some cases no change (rounding ate the change) in some cases outrageous change. Thus, you did the rounding in IMP to minimize the distortions. At most 0.5IMP error. This was still bad with the 30VP scale but that was life.

Now we have the 20VP scale with a quanta of 0.01VP. This still needs integer IMPs. Thus, you can weight and round in IMPs or you weight in VPs and round to 0.01. Former has 0.5IMP randomness, which translates to 0.05...0.1 typical randomness in VP for a typical 32 board result. Weighting in VP has a randomness of 0.005VP. An order of magnitude less.

Thus, it would be worth considering, no matter what MY motivation is, to switch to VP based weighting. I understand that tradition is very important and it was done in IMP for ages, but now, with the 20VP scale it would be worth thinking it over. That is all I wanted to say.

I already said the rest: I learnt that there is no clear law. I understood that for some, this is not obvious, either. I understood that there is a traditional way to do this. I also understood that there were very good reasons to choose the IMP based method. I also understood that this was the right choice at that time.