Removing outliers from DATUM

[Bbradley62]

It would be nice to have a par score, but many boards don't have one.

(I remember a board where I could make 6♠ while our opponents could make 6♣.)

All boards have a par. On the one that you referenced, the par would be 7♣x down one.

Using the theoretical par has two problems:

(1) Someone has to do the extra work (and take the extra time) to examine each hand closely enough to determine that theoretical par.

(2) There may be times when all results are on one side of the theoretical par. In the given example, if all EW club-holders allowed themselves to be outbid by the NS spade holders (below the 7-level), then all NS pairs will score + game or slam, but par would be set at +100 or +200, so all NS pairs would win IMPs and all EW pairs would lose IMPs. That's not how it's supposed to work.

[Siegmund]

I'd prefer to call that a difference between comparing against other players and comparing against a datum. It is indeed very possible for all the pairs one direction to fail to beat par.

 

That's a feature, not a flaw, if you choose to take ddpar as your datum. :D

 

Whether it would be popular with players or not, I am not sure - but given how popular the hand records with making contracts printed on them are, I think it'd be a lot more popular than IMPing against +185s ever was.

[helene_t]

Using the theoretical par has two problems:

(1) Someone has to do the extra work (and take the extra time) to examine each hand closely enough to determine that theoretical par.

Scoring software usually does this already. Yes, the DD par is sometimes silly (requiring dcl to drop a sec king instead of trying the finesse) but you don't want the subjectivity of human judgement involved. What you could do is to let GIB or Jack or w/e play the hands and use such artificial SD results as par.

 

(2) There may be times when all results are on one side of the theoretical par.

That could be an advantage (in a small field it sometimes happens that all EW pairs play well and all NS pairs play badly on a particular board) or it could be a disadvantage (it could be that the DD par is just unrealistic). It would work better in small fields than it large fields.

 

I don't think that a DD par datum would be better than XIMPs but I might be wrong so I think it would be worth doing some studies on.

 

But even if it turns out to be inferior to XIMPs in normal tournaments, it would still be attractive in some special cases:

- You can use it for Mitchell movements. In Mitchell, EW scores are not comparable to NS scores, especially when pairs chose their direction themselves so there may be a bias. DD par datum solves much of this problem.

- You can use it in very weak fields where the table results are generally less meaningful than SD par.

- You get you results immediately, don't have to wait for the frequency tables. Your IMP score will not be affected by adjustments if you are not involved.

- You can use it when duplication is not feasible for whatever reason.

[whereagles]

how many boards per round Nuno? I'd be reluctant to turn imps to VP unless there was enough boards per round,

We'll be playing 8, 10 or 12 boards/round, depending on the number of pairs.

 

By the way, since the division of cross-imps by number of comparisions can generate fractional imps, can anyone give me a link to a fractional VP table?

[helene_t]

By the way, since the division of cross-imps by number of comparisions can generate fractional imps, can anyone give me a link to a fractional VP table?

I am pretty sure there is no such thing. You would have to round off.

[Vampyr]

Some people seem to have suggested that XIMPs is more difficult to understand than Butler. I think it is easier, especially for anyone who has ever played matchpoints. You are compared with evey other pair who played in your direction.

 

Anyway, to the original poster -- apparently most Australian pairs tournaments are IMP Swiss pairs. I assume that they use both XIMPs and VPs, so you can probably get a scale from them (Look for the Swiss Pairs thread on "offline bridge").

[Mbodell]

how many boards per round Nuno? I'd be reluctant to turn imps to VP unless there was enough boards per round,

We'll be playing 8, 10 or 12 boards/round, depending on the number of pairs.

 

By the way, since the division of cross-imps by number of comparisions can generate fractional imps, can anyone give me a link to a fractional VP table?

If you have a fractional VP table like the USBF uses then you can extrapolate from the IMP differences. So if one pair has 3.54 you'd give them 11.486 VP using that table (VP for 3 + .54 * (VP for 4 - VP for 3)).

 

One other option though if you are playing IMP pairs is to just not convert to VP. That is what we do locally for IMP pair events. I realize that if you are doing some sort of swiss pairs this may not work, but there is no reason why IMP pairs automatically means that.

[barmar]

What you could do is to let GIB or Jack or w/e play the hands and use such artificial SD results as par.

We did something like that a couple of times several years ago at my club. We didn't use robots, we had one table with two expert pairs playing all the boards, and everyone else was IMPed against them. And after the session, one of the experts stuck around to analyze the interesting hands and answer questions.

 

I wonder if BBO would be interested in implementing this type of tournament. Perhaps the Cayne team could be recruited to play as the datum table occasionally in place of one of their daily team games.

[helene_t]

I wonder if BBO would be interested in implementing this type of tournament. Perhaps the Cayne team could be recruited to play as the datum table occasionally in place of one of their daily team games.

Interesting idea.

 

One problem is that those interested in playing such tournament would often also be interested in watching the Cayne match itself.

[mycroft]

Another way of limiting one match (which is what IMP->VP does; it's designed also to mitigate the "one blowout determines the event" factor, while not eliminating the "big win better than small win" factor completely (as would win-loss)) is simply to score by total IMPs, but cap the maximum IMP/match to something reasonable, like whatever the 25-5 (on WBF "25"-point scale) floor is (like the Cavendish Pairs rule that IMPs on a board are capped to 17*comparisons).

[Trinidad]

By the way, since the division of cross-imps by number of comparisions can generate fractional imps, can anyone give me a link to a fractional VP table?

You can easily calculate the table yourself. All VP tables are derived from the table for 32 boards. (see e.g. http://www.ecatsbridge.com/documents/Scores/imps_to_vps.asp)

 

You can derive the VP table for another number of boards, N, by multiplying the upper limits of the 32 board table by sqrt (N/32). You can see this by comparing the upper boundaries for 8 boards with the upper boundaries for 32 boards. The upper boundaries for 8 boards are exactly half of the upper boundaries for 32 boards.

 

You don't need to round anything. You just need to realize that the upper boundary is the fixed number and that the lower boundary of the next range is not 1 IMP higher, but 0.000000....1 IMP higher.

 

Rik

 

P.S. Whether it is wise to use fractional IMPs and fractional VP tables is another question. I would just round the IMP total and use the standard (rounded) VP table that people are comfortable with. But I wanted to point out that it isn't necessary to do that and that you can calculate VP tables with fractional IMP boundaries.