Do complicated methods pay off? Some answers!
tys...@yahoo.com
difference does it really make? I have seen other articles that list
different methods and then show their correlation coefficients with
the number of tricks actually taken. I don't know about you, but if
someone told me "this method" had an r-squared value of 0.74 and "that
one" was 0.76, I couldn't tell if it was worth the effort. So let's
convert that ugly coefficient to something we can understand: TRICKS
and SCORES.
If we use different methods of counting our points, this is what we
get for initial valuations:
R2 Ave Error Score
HCP 0.66 1.23 -0.49
Bergen 0.74 1.08 -0.01
HCP + 321 0.74 1.07 0.00
Zar 0.75 1.05 0.08
BUM RAP + 321 0.76 1.03 0.14
Binky 0.78 0.99 0.32
R2 is the r-squared value
Ave Error is the average number of tricks we are off by in our
estimation of the number of tricks we should take. Note that this
will never get to zero because we don't know the distribution of the
opponent's cards.
Score is the predicted number of IMPs/board that we expect to gain (or
lose) when our opponents use the simple HCP+123. This is determined
by having both sides use their method to estimate the number of tricks
they will take. An assumed par contract is determined from that
estimate and compared to the other evaluation. This isn't 100%
realistic, and the actual score could be higher or lower, but I think
it's a fair order of magnitude. I'm just trying to put a stake in the
ground where no one has tried before.
HCP is A=4, K=3, Q=2, J=1
Bergen is HCP + length of longest two suits
HCP+321 is HCP + 3 per void + 2 per singleton + 1 per doubleton
Zar is HCP + Controls + twice the length of longest suit + once the
length of second-longest suit minus length of shortest suit.
http://public.aci.on.ca/~zpetkov/
BUM RAP is a substitute for HCP: A=4.5, K=3, Q=1.5, J=0.75, T=0.25
Binky is Thomas Andrew's evaluator:
http://thomaso.best.vwh.net/bridge/valuations/
*** For all these valuations (except Binky) 1 point is deducted for
the following holdings: J,Q,K,QJ,Jx,Qx ***
I'm surprised at Zar's poor performance. I guess more complicated
doesn't always mean more accurate. BUM RAP is a simpler method that
gives better results.
We can now look at a subset of hands where opener has a 5+ suit to
examine how the methods "revaluate."
R2 Ave Error Score
HCP 0.65 1.21 -0.33
Bergen 0.71 1.12 -0.03
HCP + 321 0.71 1.11 0.00
Zar 0.72 1.10 0.04
HCP + fit 0.73 1.07 0.14
BUM RAP + 321 0.73 1.07 0.14
Zar + fit 0.74 1.05 0.22
BUM RAP + fit 0.75 1.03 0.32
Binky 0.75 1.02 0.33
Evolved Binky 0.78 0.97 0.54
HCP+fit and BUM RAP+fit revaluate so that if there is an 8-card fit,
responder's distribution points become 5,3,1. Also 2 points are added
for each trump beyond 8.
Zar+fit revaluates giving +3 points for each trump beyond 8 and +3
points for each card over 4 in a side suit.
Evolved Binky is like "Binky+fit" and is my contribution, described in
previous rgb articles:
http://tinyurl.com/25huc
http://tinyurl.com/383e6
Tysen
redspot
article on "True Value" point count in the March
2004 ACBL Bridge Bulletin.
The complexities of TV (e.g.: dealing with 1/4
points) reminds me of KNR analysis that was
being talked about a couple of years ago, although
TV appears to be intended only for balanced hands.
Does anyone actually use TV or KNR (or similar
systems), other than possibly in computer programs?
Carl Ritner
Don't know about other people, but I found the "Truer Value" in the
March ACBL Bulletin to be very cumbersome for human use. It's better
to rely on your "feel" of the hand which is developed through
practice.
Computer programs have the capability to look at every possible card
combination in each suit and assign true trick-taking potential (TTP)
to the hand, so that approximations such as HCP, TV, rule of 20, etc
are not needed. However, there needs to be a conversion routine since
the ACBL card can't be filled out in TTP.
There will never be a quick and dirty method available to assess the
strength of a hand at the table - at least none better than what we
have now. You can "tweak" each method but unless you are measuring
trick potential, you are using a correlative method, which will always
fail somewhere along the curve.
Cheers,
Carl
www.carlritner.com
Eric Sandberg
> Does anyone actually use TV or KNR (or similar
> systems), other than possibly in computer programs?
I can speak to the use of KNR. Although the actual calculations are
fairly complex, with a little trial and error practice, one can begin
to make rapid estimates at the table that are fairly close. As others
have suggested in the past, I have found it quite useful in
determining opening bids, but much less so in other situations. It is
rather conservative for balanced hands and very aggressive for shapely
hands and blind adherence results in many anti-field actions. There
are two areas where the KNR has been extremely helpful to me. One is
for balanced hands of a variety of strengths. I think it does a good
job distinguishing between hands that sit on the cusp of opening 1m
vs. pass. The KNR also identifies balanced hands with 15-17 hcp that
might not be opened 1NT as well as hands outside a traditional 15-17
that could open 1NT. These anti-field decisions are often profitable.
A second area of opportunity for the KNR is with borderline fourth
seat openers. For balanced hands in this situation a modified cassino
count where a sum of the KNR plus the number of spades totaling more
than 14 (rather than 15) identifies good openers in many cases.
Eric
redspot
news:c1918743.0403...@posting.google.com...
I've seen several variations of KNR. The version that was (is?)
on an interactive web site was different from that which was in
the original article. What calculations do you use at the table?