[blml] 60% session in IMPs

[John Probst]

----- Original Message ----- 
From: "David Burn" <dalburn at btopenworld.com>
To: "'Bridge Laws Mailing List'" <blml at amsterdamned.org>
Sent: Sunday, December 23, 2007 12:43 AM
Subject: Re: [blml] 60% session in IMPs


> [JP]
>
> We know that over 28 boards, 60% is 49 imps.
>
> [DALB]
>
> I am not sure how we know. Pairs (matchpoints, for the benighted) and IMPs
> are different games, to the extent that comparisons between them at this
> level of granularity (x matchpoints = y IMPs) appear to me to be
> meaningless. In theory, it is possible to score 100% in a 24-board pairs
> game by making a trick more than everyone else in the same contract; if 
> you
> translated this to a match, you would win by 24 IMPs. Similarly, it is
> possible to score 50% in a pairs game by making a trick more than everyone
> else in the same contract on twelve of the boards, and going thirteen down
> in 7NT redoubled on the other twelve. If you did that in a match, you 
> would
> lose by 276 IMPs.
>
> I have read various examples of the rationale for equating matchpoints to
> IMPs, and I am not remotely convinced by any of it. Still, there are
> compelling reasons why some figure needs to be arrived at. I have no
> convincing arguments why the figure I pluck out of the air should be "10% 
> of
> a top = 2 IMPs" rather than the current (in England) 3 IMPs, but JP
> apparently has. If I score 28 60% boards in a pairs tournament, that 
> equates
> to 56 IMPs at 2 IMPs a throw; not quite the 49 IMPs he quotes, but closer 
> to
> that than the 84 IMPs I would have scored at 3 IMPs a throw. I still
> believe, though, that comparing scores achieved at pairs to scores 
> achieved
> at IMPs is to compare apples not with oranges, but with screwdrivers.

David, I entirely agree. A lot of the math associated with VP tables and 
such like is based on observed data. We do *know* that 49 imps is 60% based 
on a staggering volume of observed data; probably 800-1,000 imp games and 
twice the number of mp games from the YC; and the only thing we tested was 
what proportion of players score greater than X imps to be the same as 60+%. 
it turned out to be 49 imps. After that we rely on Gauss tables.  The Friday 
night YC game was still represented as a 54% game, and we "knew" that it was 
about 7 or 8 imps stronger than a 50% game. All these numbers hang to gether 
well based on experience; about 1/4 imp a board is the average extra 
strength of the YC Friday night game. Purists throw up their hands in 
horror. Engineer just build their structure because from experience mostly 
we know it won't fall down.

We observe that the Law recognises that 3 imps in a head to head is A+ at 
pairs. it says nothing more. Fortunately this is very close to the observed 
data, and one can stroll from mps, to Butler, to head-to-head, each time 
based on a lot of obseved data.  ... and the numbers I mention are the ones 
we have teased out of this morass of results.  Without question, the 
"correct" (ie, the best) A+ at Butler is 2 ( think it's a tad more 
mathematically) imps, and head to head 3 imps. If it mattered, I'd factor 
the 2 imps at Butler based against the 28 board 49 imps figure for 
completeness.

so, if you have better data let me know and I'll recompute, but I have had 
access to vast quantities of data over time.  best, John

>
> David Burn
> London, England
>
>
> _______________________________________________
> blml mailing list
> blml at amsterdamned.org
> http://www.amsterdamned.org/mailman/listinfo/blml