Improving hand evaluation - Evolved Binky points (long & technical)
tys...@yahoo.com
evaluation techniques using double-dummy results. Anyone who is
interested in hand evaluation and has never visited his web page
should do so:
http://thomaso.best.vwh.net/bridge/valuations/
One of Thomas's creations is Binky points, an effort to create a hand
evaluation scheme where the sum of a pair's Binky points is an
estimation of the total number of tricks they will take in their best
contract. If you are used to using 4321 points for HCP and 321 for
distribution, each Binky point is worth about 2.7 points. Here is a
nice table that shows what Binky points say you should use for
distribution points (rounding to a half-point and setting the 4333
shape to zero):
Shape Binky Dist Pts
4-3-3-3 3.60 0
4-4-3-2 3.89 1
5-3-3-2 3.94 1
5-4-2-2 4.19 1.5
6-3-2-2 4.26 1.5
4-4-4-1 4.41 2
5-4-3-1 4.46 2.5
6-3-3-1 4.52 2.5
7-2-2-2 4.60 2.5
6-4-2-1 4.75 3
5-5-2-1 4.78 3
7-3-2-1 4.81 3
5-4-4-0 5.12 4
8-2-2-1 5.15 4
6-4-3-0 5.22 4.5
5-5-3-0 5.24 4.5
7-3-3-0 5.29 4.5
6-5-1-1 5.30 4.5
7-4-1-1 5.31 4.5
8-3-1-1 5.39 4.5
7-4-2-0 5.52 5
8-3-2-0 5.55 5
6-5-2-0 5.56 5
My idea to add to this is that the evaluator should be able to evolve
as the auction progresses. It would be too much to have a different
evaluator for every possible scenario, but I think a few common ones
would be worth trying.
I have taken the DD library and found the hands where one partner at
least one 5+ suit. Assume this hand has named his longest suit to
partner. You can then generate a new evaluation profile for Responder
that adds with the "original" Binky points for Opener. You would now
have a new patterns table that has the length in partner's suit
separated from the other 3 suits. I'm defining a pattern convention
that fixes one suit in place, so you will have patterns like 2=5-3-3
and 4=6-3-0. The first number is the amount of support you have for
partner's suit, and then we list the other three suits. You can also
create 2 different holdings tables that show how to evaluate honors in
the suits: one for partner's suit and one for the other 3.
I've gone ahead and determined the new "evolved" Binky points for
Responder. I've only been able to do this for patterns, and not honor
holdings. I'm still learning how to use this database program I'm
using and it doesn't seem easy for me to determine holdings values
yet. Still, I can get the valuations for patterns:
Shape Evol Orig Diff Count
2=4-4-3 3.70 3.89 -0.19 79910
2=5-3-3 3.74 3.94 -0.19 57194
3=4-3-3 3.85 3.60 0.25 80986
1=4-4-4 3.87 4.41 -0.54 15005
2=5-4-2 3.89 4.19 -0.30 77350
1=5-4-3 3.95 4.46 -0.51 64477
2=6-3-2 4.01 4.26 -0.24 40813
1=6-3-3 4.06 4.52 -0.45 17135
3=4-4-2 4.07 3.89 0.17 54781
3=5-3-2 4.09 3.94 0.15 78419
4=3-3-3 4.11 3.60 0.51 16934
1=6-4-2 4.26 4.75 -0.49 23303
1=5-5-2 4.27 4.78 -0.51 15885
3=6-2-2 4.34 4.26 0.08 14296
2=7-2-2 4.37 4.60 -0.22 5605
0=5-4-4 4.39 5.12 -0.73 7880
2=6-4-1 4.43 4.75 -0.32 17003
4=4-3-2 4.44 3.89 0.55 69258
1=7-3-2 4.45 4.81 -0.35 9155
2=5-5-1 4.46 4.78 -0.32 11599
0=5-5-3 4.53 5.24 -0.71 5691
3=5-4-1 4.56 4.46 0.09 32609
0=6-4-3 4.57 5.22 -0.66 8351
5=3-3-2 4.58 3.94 0.64 14008
2=7-3-1 4.58 4.81 -0.23 6727
3=6-3-1 4.62 4.52 0.11 17279
4=5-2-2 4.74 4.19 0.55 16818
1=6-5-1 4.81 5.30 -0.49 6815
0=7-3-3 4.82 5.29 -0.47 1625
5=4-2-2 4.85 4.19 0.66 9394
3=7-2-1 4.87 4.81 0.06 4649
6=3-2-2 4.91 4.26 0.65 2365
0=6-5-2 4.93 5.56 -0.63 3922
1=7-4-1 4.95 5.31 -0.36 3724
0=7-4-2 4.98 5.52 -0.54 2216
4=4-4-1 5.06 4.41 0.66 14289
2=8-2-1 5.06 5.15 -0.09 1401
4=5-3-1 5.14 4.46 0.67 20730
1=8-3-1 5.14 5.39 -0.25 1161
5=4-3-1 5.29 4.46 0.83 11543
3=5-5-0 5.29 5.24 0.05 2275
2=6-5-0 5.30 5.56 -0.26 2332
4=6-2-1 5.33 4.75 0.58 7420
6=3-3-1 5.34 4.52 0.82 1487
2=7-4-0 5.36 5.52 -0.16 1287
3=6-4-0 5.37 5.22 0.15 3312
3=7-3-0 5.46 5.29 0.17 1329
5=5-2-1 5.59 4.78 0.81 5605
6=4-2-1 5.62 4.75 0.87 1928
4=5-4-0 5.92 5.12 0.80 3883
4=6-3-0 6.00 5.22 0.78 2045
5=4-4-0 6.26 5.12 1.15 1126
5=5-3-0 6.28 5.24 1.04 1604
I cut the list off at a hand count of 1000, as there is some error in
the figures of some of the rarer hand types. However, the standard
error of the valuations with a count of ~1000 is less than 0.02. It's
even less for patterns with a larger count.
Here are the results translated into Milton Work values:
Shape MWC Orig Diff
2=4-4-3 0.5 1 -0.5
2=5-3-3 0.5 1 -0.5
3=4-3-3 0.5 0 0.5
1=4-4-4 0.5 2 -1.5
2=5-4-2 1 1.5 -0.5
1=5-4-3 1 2.5 -1.5
2=6-3-2 1 1.5 -0.5
1=6-3-3 1 2.5 -1.5
3=4-4-2 1 1 0
3=5-3-2 1.5 1 0.5
4=3-3-3 1.5 0 1.5
1=6-4-2 1.5 3 -1.5
1=5-5-2 2 3 -1
3=6-2-2 2 1.5 0.5
2=7-2-2 2 2.5 -0.5
0=5-4-4 2 4 -2
2=6-4-1 2 3 -1
4=4-3-2 2 1 1
1=7-3-2 2.5 3 -0.5
2=5-5-1 2.5 3 -0.5
0=5-5-3 2.5 4.5 -2
3=5-4-1 2.5 2.5 0
0=6-4-3 2.5 4.5 -2
5=3-3-2 2.5 1 1.5
2=7-3-1 2.5 3 -0.5
3=6-3-1 2.5 2.5 0
4=5-2-2 3 1.5 1.5
1=6-5-1 3 4.5 -1.5
0=7-3-3 3 4.5 -1.5
5=4-2-2 3.5 1.5 2
3=7-2-1 3.5 3 0.5
6=3-2-2 3.5 1.5 2
0=6-5-2 3.5 5 -1.5
1=7-4-1 3.5 4.5 -1
0=7-4-2 3.5 5 -1.5
2=8-2-1 4 4 0
4=4-4-1 4 2 2
4=5-3-1 4 2.5 1.5
1=8-3-1 4 4.5 -0.5
5=4-3-1 4.5 2.5 2
3=5-5-0 4.5 4.5 0
2=6-5-0 4.5 5 -0.5
4=6-2-1 4.5 3 1.5
6=3-3-1 4.5 2.5 2
2=7-4-0 4.5 5 -0.5
3=6-4-0 4.5 4.5 0
3=7-3-0 5 4.5 0.5
5=5-2-1 5.5 3 2.5
6=4-2-1 5.5 3 2.5
4=5-4-0 6 4 2
4=6-3-0 6.5 4.5 2
5=4-4-0 7 4 3
5=5-3-0 7 4.5 2.5
The results show what we've all known for years, but it's nice to be
able to put some quantitative figures to it. Hands with support for
partner go up in value and those without support go down. Hands with
shortness go up even more in value the longer their support for
partner is. Again, nothing new here but with some real data behind
it, maybe we can use this to find some good rules of thumb to use at
the table for how much to actually adjust your values. These tables
could easily be used by computer programs for a more accurate hand
evaluation method as the auction progresses.
Note that all hands with 3+ support go up in value, even 3=4-3-3.
3433 increases in value by 0.25 Binky or over half a point in "normal"
valuation. This is because hands like these increase in value because
we're helping partner realize *his* full potential, since he's still
using regular Binky.
But it doesn't stop here. We can do another iteration to make our
evaluation even better. Let's say through the bidding, Opener knows
that Responder has 3+ support for his suit. He also knows that
Responder is using the Evolved Binky count. How does his Binky
evaluation change? We can do one more iteration and find out:
Shape Count Evol Orig Diff
5-3-3-2 120645 3.83 3.94 -0.11
5-4-2-2 82517 4.08 4.19 -0.11
5-4-3-1 100751 4.38 4.46 -0.08
6-3-2-2 34573 4.48 4.26 0.22
5-5-2-1 49693 4.72 4.78 -0.06
6-3-3-1 21133 4.79 4.52 0.28
7-2-2-2 2214 4.98 4.60 0.38
6-4-2-1 28945 5.03 4.75 0.28
5-4-4-0 9759 5.09 5.12 -0.03
5-5-3-0 13994 5.16 5.24 -0.08
7-3-2-1 8436 5.29 4.81 0.49
6-4-3-0 8042 5.55 5.22 0.33
6-5-1-1 4307 5.61 5.30 0.31
7-4-1-1 1700 5.84 5.31 0.54
7-3-3-0 1147 5.91 5.29 0.61
6-5-2-0 3926 5.91 5.56 0.35
7-4-2-0 1632 6.09 5.52 0.57
These adjustments aren't as large, but some of them are significant.
Note that all hands that only had a 5-card suit lose value, because
partner was averaging over all our possible hands, and we could have
had better. It also proves that Responder was trying to improve his
count for both of us at once (and doing a pretty good job) so we have
to cut back when we only have a modest fit. The point is that if we
both improve our hands too much we'll double-count the bonuses.
We can do it again assuming partner has promised 4+ support:
Shape Count Evol Orig Diff
5-3-3-2 52803 3.84 3.94 -0.09
5-4-2-2 35809 4.11 4.19 -0.08
6-3-2-2 12131 4.32 4.26 0.06
5-4-3-1 44206 4.46 4.46 0.00
7-2-2-2 573 4.59 4.60 -0.01
6-3-3-1 7335 4.67 4.52 0.16
5-5-2-1 21645 4.78 4.78 0.00
6-4-2-1 10162 4.96 4.75 0.21
7-3-2-1 2289 5.01 4.81 0.20
5-4-4-0 4333 5.22 5.12 0.10
5-5-3-0 6116 5.31 5.24 0.07
6-4-3-0 2880 5.51 5.22 0.29
6-5-1-1 1530 5.58 5.30 0.28
7-4-1-1 455 5.58 5.31 0.27
7-3-3-0 314 5.69 5.29 0.39
7-4-2-0 438 5.86 5.52 0.34
6-5-2-0 1367 5.88 5.56 0.32
Partner has promised more support, so we have different changes to our
valuations. One odd result is the 7-2-2-2 valuation. Why is it so
low? With a count of 573, the standard error is less than 0.03, so
that's not likely it. The numbers say that when partner has 4+
trumps, we'd rather have a 5521 or 6331 shape than 7222, even with the
11-card trump fit. This anomaly doesn't appear with the 3+ support.
I don't have an explanation for this yet except perhaps partner is
expecting his 4+ trumps to really help us with ruffs in our own hand.
Here's Milton Work points again:
Shape Orig 3+ 4+
5-3-3-2 1 0.5 1
5-4-2-2 1.5 1.5 1.5
6-3-2-2 1.5 2.5 2
5-4-3-1 2.5 2 2.5
7-2-2-2 2.5 3.5 2.5
6-3-3-1 2.5 3 3
5-5-2-1 3 3 3
6-4-2-1 3 4 3.5
7-3-2-1 3 4.5 4
5-4-4-0 4 4 4.5
5-5-3-0 4.5 4 4.5
6-4-3-0 4.5 5 5
6-5-1-1 4.5 5.5 5
7-4-1-1 4.5 6 5
7-3-3-0 4.5 6 5.5
6-5-2-0 5 6 6
7-4-2-0 5 6.5 6
For future work, I'd really like to see how honor valuations change as
the auction progresses. If partner shows a 5+ suit, how does the
worth of Qxx in his suit compare to Qxx in a side suit? I'm working
on this problem now, but I'm not sure if I can solve it.
Tysen
John Crinnion
> Thomas Andrews has done a considerable amount of research into hand
> evaluation techniques using double-dummy results. Anyone who is
> interested in hand evaluation and has never visited his web page
> should do so:
>
> http://thomaso.best.vwh.net/bridge/valuations/
>
> One of Thomas's creations is Binky points,
Binkie (note correct spelling) points were created by the late great
Brian O'Callaghan, who rejoiced in the nickname of 'Binkie'.
Stephen Pickett
> If you are used to using 4321 points for HCP and 321 for
> distribution, each Binky point is worth about 2.7 points.
Hi Tysen,
This is essentially a repeat of a comment I made in another thread (srry
folks)
You should most certainly look at:
http://public.aci.on.ca/~zpetkov/
The reason the method given there looks good is that you can do it easily in
your head, at the table.
Warning: this site doesnt work with Netscape 4.7x.
--
Stephen Pickett, PO Box 44538, Vancouver BC Canada V5M 4R8
Telephone: (604) 874-7327, Fax: (604) 874-7326, ICQ UIN#212132
Go see BRidgeBRowser at http://www.microtopia.net/bridge/
Lex Spoon
> For future work, I'd really like to see how honor valuations change as
> the auction progresses. If partner shows a 5+ suit, how does the
> worth of Qxx in his suit compare to Qxx in a side suit? I'm working
> on this problem now, but I'm not sure if I can solve it.
Another nice analysis would be to factor in tidbits about the
opponents' holdings. If RHO bids a suit, showing 4+ or 5+ in the
suit, then K's in that suit become more valuable, while if LHO bids a
suit the opposite happens.
There should be distributional implications as well. If you have a
singleton in one opp's suit, then your side should has more expected
tricks if you play in your best trump suit.
As you say, it's stuff that everyone knows, but it is nice to see some
numbers and to get a better idea of the *magnitude* of the importance.
-Lex
Thomas Andrews
Huh?
If there is something called "Binkie points," that's not what I named
my "Binky Points" after. Binky was a fake nickname I added to my name
when we were doing a show at MIT - the program listed me as Thomas
"Binky" Andrews. It was a joke, but it sort-of stuck.
=thomas andrews
=tho...@best.com
Josprung
>evaluation techniques using double-dummy results.
I have a major problem with this technique. It will clearly overvalue a
holding of say AKJTxx as it will get the suit right whenever it can.
Therefore these methods overvalue hands where a key card needs to be guessed,
and subsequently undervalue hands with solid, or semi solid suits like AKQxxx,
or KQJTxx.
Danny
tys...@yahoo.com
This is always a limitation when using DD data for evaluations. But
it is something you can keep in your mind and human judgement should
always be used. But note that specific honor holdings don't matter in
the case of a shape distribution analysis, so the findings presented
here should still stand.
Tysen
Gordon Rainsford
John Crinnion
Very glad to hear it (that he is not X - comparatively indifferent re
O') - I just thought I had recently seen a memorial event announced at
Earl's Court's Premier Bridge Club.
It is something of an 'alternate universe' experience, though, is it
not? Binkie points and - through the looking glass - Binky points,
and both of them stemming from dissatisfaction at the comparative
crudity of the Milton Work point count. I wonder if the two models
could be synthesised into - what? - Binkee points? Bridge Theory
could only be the richer.
And, as with LTC and TNT, silly old partner need never know what you
are up to!
Ian Payn
"John Crinnion" <jcri...@yahoo.com> wrote in message
news:2f389bab.04021...@posting.google.com...
> r...@gordonrainsford.co.uk (Gordon Rainsford) wrote in message
news:<1g93wh5.1vnqslz1cd60eN%r...@gordonrainsford.co.uk>...
> > John Crinnion <jcri...@yahoo.com> wrote:
> >
> > > tys...@yahoo.com wrote in message
> > news:<a1bce703.04021...@posting.google.com>...
> > > > Thomas Andrews has done a considerable amount of research into hand
> > > > evaluation techniques using double-dummy results. Anyone who is
> > > > interested in hand evaluation and has never visited his web page
> > > > should do so:
> > > >
> > > > http://thomaso.best.vwh.net/bridge/valuations/
> > > >
> > > > One of Thomas's creations is Binky points,
> > >
> > > Binkie (note correct spelling) points were created by the late great
> > > Brian O'Callaghan, who rejoiced in the nickname of 'Binkie'.
> >
> > He's neither dead nor O'
>
> Very glad to hear it (that he is not X - comparatively indifferent re
> O') - I just thought I had recently seen a memorial event announced at
> Earl's Court's Premier Bridge Club.
++++That was The Keith Loveys Memorial Individual. I donated the trophy for
that, an engraved tankard. I didn't intend to, it's just that no-one paid me
for it, and at this stage of proceedings I doubt anyone will. I was, by the
way, going to have it engraved with The Keith Loveys Memorial Individual
Trophy January 11th 2004, but the engraver charged by the letter, so I
settled on The Loveys Cup, saving myself a fortune.
As far as "Binkie Points" are concerned, the method has been known and
referred to as such for some years now. Big battle between Binkie Points and
Binky Points as to who uses the name? Doubt anyone cares enough. I can't
remember how many times I've had Binkie Points explained to me, but my eyes
always glaze over...
tys...@yahoo.com
interested they are in this type of analysis. Several people have
made suggestions for the next situation to examine or other things
they would like to see. I'm currently working on my honor value
problem that I stated in my original article, and I think I'm close to
coming up with an solution. After that I have a couple things I want
to look at.
1. Since the Binky count is essentially only usable by computers, I'm
hoping to come up with some rules of thumb that are simple enough to
remember that can capture 90% of what is discovered here.
2. Other situations are worth examining. How does your evaluation
change when RHO bids a suit, or when LHO bids one?
Again thanks to everyone who sent me praise for my original article.
Tysen
John Crinnion
> "John Crinnion" <jcri...@yahoo.com> wrote in message
> news:2f389bab.04021...@posting.google.com...
> > r...@gordonrainsford.co.uk (Gordon Rainsford) wrote in message
> news:<1g93wh5.1vnqslz1cd60eN%r...@gordonrainsford.co.uk>...
> > > John Crinnion <jcri...@yahoo.com> wrote:
> > >
> > > > tys...@yahoo.com wrote in message
> news:<a1bce703.04021...@posting.google.com>...
> > > > > Thomas Andrews has done a considerable amount of research into hand
> > > > > evaluation techniques using double-dummy results. Anyone who is
> > > > > interested in hand evaluation and has never visited his web page
> > > > > should do so:
> > > > >
> > > > > http://thomaso.best.vwh.net/bridge/valuations/
> > > > >
> > > > > One of Thomas's creations is Binky points,
> > > >
> > > > Binkie (note correct spelling) points were created by the late great
> > > > Brian O'Callaghan, who rejoiced in the nickname of 'Binkie'.
> > >
> > > He's neither dead nor O'
> >
> > Very glad to hear it (that he is not X - comparatively indifferent re
> > O') - I just thought I had recently seen a memorial event announced at
> > Earl's Court's Premier Bridge Club.
>
> ++++That was The Keith Loveys Memorial Individual.
I really must stop posting when hungover.
<SNIP>
> I was, by the
> way, going to have it engraved with The Keith Loveys Memorial Individual
> Trophy January 11th 2004, but the engraver charged by the letter, so I
> settled on The Loveys Cup, saving myself a fortune.
Yes, I make that two pounds ten reckoning at a notional rate of
half-a-crown per letter.
> As far as "Binkie Points" are concerned, the method has been known and
> referred to as such for some years now. Big battle between Binkie Points and
> Binky Points as to who uses the name? Doubt anyone cares enough. I can't
> remember how many times I've had Binkie Points explained to me, but my eyes
> always glaze over...
Can't you visualise an English Bridge cover: an image of Binky and
Binkie shaking hands superimposed on a background of the Atlantic
Ocean, with a caption such as " 'Hands' across the Sea"?