8-board VP scale for 10-board match

[mrdct]

It is quite common in butler-scored IMP-pairs events in Australia to use a VP scale based on less boards than the actual match length. The CTD running a recent event I was covering on BBO rationalised this approach on the basis that a butler is not the same as a head-to-head teams encounter because in teams two real bridge scores are obtained and IMPed against each other, whereas in a butler there is one real score compared against a 'flattened' average of a number of tables.

 

Any thoughts on the merits or otherwise of this approach? It had a profound impact on the outcome the event in question with just 0.01 VPs (new WBF scale) separating 3rd and 4th in an event where the top three pairs form a team to contest a week-long event interstate with airfares, accommodation, entry fees, etc. Using the 10-board scale (new or old) would've reversed the order of 3rd & 4th.

[helene_t]

When you said it had impact on the ranking, presumably multiple sesion results were converted to VP before adding them together? If the fields were different then that would probably be reasonable but if the fields were identical then I think the IMPs should have been added before VPing, in which case the conversion wouldn't affect the order.

 

But given that the decision was made to convert individual sesion results to VP, I think it makes sense to adjust for a lower expected variance in butler than in teams. The number of IMPs that you can accumulate due to luck is smaller so it is in a sense equivalent to a shorter team match. If we simplify things a lot (ignoring the total point to IMP conversion, assuming an infinite size of the butler field, assuming that VPs should be a measure of the statistical significance of pair being above average) then dividing the match length by a factor sqrt(2) would be appropriate.

[paulg]

I believe that the EBU makes the same recommendation. In Scotland they apply a weighting function before the conversion to VPs, which has a similar effect.

[mrdct]

The format was a double round robin of 4 pairs playing 10-board matches for a total of 6 rounds. Each 10-board match was IMPed against predetermined datums (the hands were sourced from a zonal final somewhere in the 80s) and converted to VPs using the new WBF scale for 8-board matched.

[Vampyr]

I believe that the EBU makes the same recommendation.

 

I am not sure that Butler pairs are still used in the EBU. In my experience it is always cross-IMPed nowadays.

 

The format was a double round robin of 4 pairs playing 10-board matches for a total of 6 rounds. Each 10-board match was IMPed against predetermined datums (the hands were sourced from a zonal final somewhere in the 80s) and converted to VPs using the new WBF scale for 8-board matched.

 

If this was the case I assume that rather than the first four pairs you meant the first two in each direction, since the comparison between East-West pairs and North-South pairs would have been entirely random. Even arrow-switching (which I don't think is appropriate in IMP pairs anyway) would not make any difference.

 

Why was such a strange decision (to use predetermined rather than actual datums) taken anyway?

[barmar]

Why was such a strange decision (to use predetermined rather than actual datums) taken anyway?

Sounds like an IMP version of Instant Matchpoints.

[Vampyr]

Sounds like an IMP version of Instant Matchpoints.

 

We have this in the EBU every year -- "Play with the Experts" at the Brighton Congress.

 

The problem with the "top 4 scorers" is that sometimes the top score N/S is eg +180 while the top score E/W is eg -5.

[barmar]

At our club we used to run an IMP Pairs game where there was one table with two expert pairs playing all the boards against each other, while everyone else went through an ordinary pair movement. We used their result as the datum to IMP everyone else against.

[mycroft]

Why would anyone use computed datums? Especially when you're comparing against great bridge results like +147 (or, with better programs, +410)?

[mrdct]

Why was such a strange decision (to use predetermined rather than actual datums) taken anyway?

There were only 4 pairs in the field, so if an internally produced datum was used you would be hugely at the mercy of what happened at the other table. It's very common in Australia to use externally sourced datums for field sizes of 6 or less pairs.

[Trinidad]

There were only 4 pairs in the field, so if an internally produced datum was used you would be hugely at the mercy of what happened at the other table. It's very common in Australia to use externally sourced datums for field sizes of 6 or less pairs.

And now they are at the mercy of an imaginary external field. Unless you can guarantee that the external field is much larger and has the same strength of the internal field I think it is much worse to depend on an external field. You may be removing random variation, but are introducing bias. As a player, I can live with random variation (such is life), but I consider bias unfair.

 

Why not just run 3 (or 2x3) matches, scored at IMPs, converted to VPs (for the correct number of boards)?

 

Thinking about it, this all makes little sense: We invented a Butler method to imitate the scoring in a teams match for pairs play in events with more than 4 pairs. Here, we have a team match and we try to imitate Butler scoring (which was intended to imitate a team match)?!?

 

Given that Butler is a bad method to imitate a team match to begin with, particularly for small fields, the last thing I would worry about is the victory point scale that is used.

 

Rik

[Vampyr]

There were only 4 pairs in the field, so if an internally produced datum was used you would be hugely at the mercy of what happened at the other table. It's very common in Australia to use externally sourced datums for field sizes of 6 or less pairs.

 

How did the event get to the point where four pairs played to eliminate one?

[mrdct]

And now they are at the mercy of an imaginary external field. Unless you can guarantee that the external field is much larger and has the same strength of the internal field I think it is much worse to depend on an external field. You may be removing random variation, but are introducing bias. As a player, I can live with random variation (such is life), but I consider bias unfair.

 

Why not just run 3 (or 2x3) matches, scored at IMPs, converted to VPs (for the correct number of boards)?

 

Thinking about it, this all makes little sense: We invented a Butler method to imitate the scoring in a teams match for pairs play in events with more than 4 pairs. Here, we have a team match and we try to imitate Butler scoring (which was intended to imitate a team match)?!?

 

Given that Butler is a bad method to imitate a team match to begin with, particularly for small fields, the last thing I would worry about is the victory point scale that is used.

 

Rik

I'm not 100% sure where the datums came from, but I believe they were from a multi-zone open event from the mid-80s (I think WBF Zones 5 & 6) and they certainly weren't from an "imaginary external field". I'm guessing that there would've been 10-15 or so teams involved in such an event with some variability in standard smoothed out by excluding some outlier results.

 

This particular event was a youth selection event and whilst it was disappointing that only 4 pairs entered, this was for selection of a state team, not a national team, and in recent years there have been three or less pairs entering so we are heading in the right direction.

 

I think external datums are ideal for this sort of situation but an option may have been to use the same boards as the 6-table seniors event being played in the room next door and score-up against their datums or an 8-table datum including the 2 youth tables. Problematic though as the quality of the youth field was certainly quite a bit stronger than the seniors and we had slightly different session times.

 

My main query though is not the use of butler scoring, but which VP scale should be used or should it just be on IMPs with some cut-off (which is how NSW and the national selection event do it).

[mrdct]

How did the event get to the point where four pairs played to eliminate one?

We only had four pairs enter, but we do have other similar event in Australia where a field of 10 pairs reduces to one winner (straight on to the team) and places 2 to 5 fight it out for the remaining two spots on the team. It actually got to a point with just two pairs left as first and second were so far ahead of third with a round to go that the TD allowed them to not play the last last roudn and go home early (the draw had been organised to have 1v2 and 3v4 in the last round).

[Vampyr]

I think that playing three stanzas, with each pair teaming up with each of the other pairs might be a sensible solution to getting an IMP result from four pairs.

[gordontd]

We have this in the EBU every year -- "Play with the Experts" at the Brighton Congress.

 

The problem with the "top 4 scorers" is that sometimes the top score N/S is eg +180 while the top score E/W is eg -5.

That's why in "Play with the Experts" there are separate ranks for EW & NS.

[Vampyr]

That's why in "Play with the Experts" there are separate ranks for EW & NS.

 

Quite. But it is not clear how the artificial datum was used in the OP case.

[mrdct]

Quite. But it is not clear how the artificial datum was used in the OP case.

I'm not sure where you are getting "artificial" from. The datums used were derived from taking some sort of average, rounded to the nearest 10 and excluding an appropriate number of outliers, taken from actual results from around a dozen tables of open national teams in a multi-zonal tournament. There's nothing artificial about it.

 

In terms of how they were used, normal IMP scoring against the datum. Mechanically, at the end of each match the players were handed a piece of paper with the datums and they scored up. This is pretty much how all of my home games are organised where I always keep a few sets of boards on hand from old events that I didn't play in or read about and have the datums in an envelope to score up after play.

[helene_t]

AFAICS you could make a joined ranking for NS and EW as long as the external datum is not from a mitchell but from a reasonably balanced movement such as a Howell.