[blml] Thai braking

[Steve Willner]

[getting back to this thread after long delay]

> From: richard.hills at immi.gov.au
>  Cross imps limits
> one "luck problem" of a single random imp landing NS or EW
> at the cost of a huge "luck problem" of double-digit imps
> due to the luck of sitting NS or EW when a ridiculous
> result happens at another table.

Or when it happens at your own.  Or facing opponents who have perfect 
methods for the deal.  Or bad ones.  There is inherent randomness in 
bridge.  Why should the scoring system increase it?

> Suppose...
> -1700 (ridiculous result caused by South insisting on an
> esoteric convention and North having a memory lapse)

The scoring system doesn't know or care whether a result is ridiculous 
or comes from excellent play.  If you want to take the _nature_ of the 
result into account, you will need human judges.  (Sounds a bit like ski 
jumping, which combines an objective measure of distance with subjective 
"form points.")

[flat board except for one wild result]
> At olympic-scored Butler-with-a-datum (top and bottom
> scores excluded before striking the datum), the EW players
> at the six normal tables get their fair result of a flat
> board.

You left out "totally un-" just before "fair."  If a player holding the 
same cards I do gets a worse result than mine, any objective scoring 
system will give me some IMPs.  In your example, Butler deprives me of 
my due.  In effect, Butler is giving the entire benefit of a ridiculous 
result to the single lucky pair who happen to be table opponents. 
Cross-IMP scoring spreads the benefit around more evenly.

If you don't like large swings on a single deal, change the scoring 
system.  For example, some IMP pair events truncate the IMP scale at 18. 
You could change the scale more drastically or use a different form of 
scoring altogether.  But don't think using a datum solves the problem of 
deviant results.  It just puts the problem in a different place.

> Aussie national champion and mathematician, Warren Lazer,
> has devised an olympic-scored cross imps method which
> fixes this problem.  For each pair, their imps are crossed
> only against the middle four of the other six scores.

Better than a datum but still unfair.  If he's really a mathematician, 
either he hasn't thought things through or he is trying to accomplish 
something different than you suggest.

From: Herman De Wael <hermandw at skynet.be>
> I can easily create a third "bastille" system ...
 > by defining the average as that number for
> which the total imps becomes zero.

That would be fair.  You will need different averages for NS and EW, but 
that's no problem.  When you do the math, you'll find it differs very 
little from cross-IMPs.  (The main difference will be larger swings on 
bimodal deals.)  I personally like this system less than cross-IMPs, but 
that's a matter of taste, not right or wrong.