denial cue-bidding
Micha Keijzers
I've posted this question in a dutch newsgroup some time ago, but got no
satisfactory anwers. So, I'm posting the question here: does anyone of
you know how denial cue-bidding works and if there is any literature
(perhaps on the internet) about it? I believe that denial cue-bidding is
a system of cue-bidding that is a bit familiar with what is called
spiral scans. So, if anyone knows more about that, then I'd be happy to
hear about that also.
Best regards,
Micha Keijzers
Michael Arnowitt
>
> Hi all,
>
> I've posted this question in a dutch newsgroup some time ago, but got no
> satisfactory anwers. So, I'm posting the question here: does anyone of
If it's like spiral scans (I think I read about those in Rosenkranz and
Alder's Bidding on Target) a cuebid promises controls in all the suits
skipped, denies one in the step actually bid...a way to show several
controls at once.
So maybe if hearts agreed suit, 5D would show controls in spades and
clubs (because you skipped those steps) but not in diamonds? (never
played against anyone who used them, so I am really guessing here based
on the name)
Michael
Skippy
<key...@sci.kun.nl> wrote:
>Hi all,
>
>I've posted this question in a dutch newsgroup some time ago, but got no
>satisfactory anwers. So, I'm posting the question here: does anyone of
>you know how denial cue-bidding works and if there is any literature
>(perhaps on the internet) about it? I believe that denial cue-bidding is
>a system of cue-bidding that is a bit familiar with what is called
>spiral scans. So, if anyone knows more about that, then I'd be happy to
>hear about that also.
>
>Best regards,
>Micha Keijzers
>
I'm not sure as I don't play the method....but I think what your
discribing is commonly called an East Coast Cue. You might want to
try looking something up on that????
Skippy
Douglas Newlands
:I've posted this question in a dutch newsgroup some time ago, but got no
:satisfactory anwers. So, I'm posting the question here: does anyone of
:you know how denial cue-bidding works and if there is any literature
:(perhaps on the internet) about it? I believe that denial cue-bidding is
:a system of cue-bidding that is a bit familiar with what is called
:spiral scans. So, if anyone knows more about that, then I'd be happy to
:hear about that also.
Denial cues is a scheme for bidding several controls at once.
One needs to have an agreed order for the suits. If one is using a
relay system, typically one uses suit length for the ordering i.e. if
you have shown 3154 then the ordering is DCSH (remember in a relay
system, no suit is set as trumps when dcb's are initiated).
If one is using a natural system then a typical agreement is trumps
then the other suits in descending order e.g. if H is trumps then
the order is HSDC.
A supplementary rule is that singletons are scanned once, doubletons
twice. This clearly easy in a relay system and, if splinters have
been used, can be sensible in a natural system.
Typically in natural systems, dcb is used after rkc and is called
spiral scan.
Also there is a matter of what is shown; again relay systems have often
had a gross count of controls (A=2,K=1) or AKQ-controls (A=3,K=2,Q=1)
so that an upper limit is known. Then one shows As and Ks and after these
are denied one shows Qs or one can shows As, Ks and Qs.
A control is shown by skipping the step corresponding to that suit and
a control is denied by bidding the step corresponding to that suit.
Here are some examples using RKC as the start where the cards being
scanned are, in order, QS,KH,KD,KC,QH,QD,QC (can agree to stop here
or can scan for Js where JS>QH but all other Js<QC)
1S 3S
4N 5C rkc=0 or 3
5D 5N shows QS since skipped 5H
shows KH since skipped 5S
denies KD since bid 5N
6C 6S 6C requests more dcb/spiral scan
shows KC since skipped 6D
shows QH since skipped 6H
denies QD since bid 6S
now with a splinter
1S 4D spl
4N 5C
5D 6D shows QS since skipped 5H
shows KH since skipped 5S
shows KC since skipped 5N (no step for diamonds since spl)
shows QH since skipped 6C
denies QD since bid 6D
If you play a strong club, you can fit RKC in earlier (or Gerber :) )
and scan deeper. Drawback is that you lose room to determine if slam
is really there and end up using rkc+dcb as your slam investigation
which is really not much better than those who overuse rkc/Bwood!
An additional rule which can be used (with a previous gross control
count only I think) is to by-pass suits where you hold the AKQ - this
should be deducible.
FWIW, i think that rkc with responder showing the lowest useful K
and opener being able to ask for another K is easier and just
as useful most of the time (unless you really like conventional
mechanisms - been there, done that!)
Douglas, Geelong
--
Dr. Douglas A. Newlands, School of Computing & Mathematics,
Deakin University, Geelong, Victoria 3217, Australia,
Tel +61(03)52271165 [Fax 52272028]
email: do...@deakin.edu.au http://www.deakin.edu.au/~doug
Micha Keijzers
Douglas Newlands schreef:
Thank's for your reaction, I have two questions though:
- When you want to investigate further, do you always bid the next suit or is it possible
to skip some suits to ask for a specific value in a specific suit?
- When you want to sign off in a slam, can you just break the relays and bid the slam
which you want to play or do you need some conventional bid to tell partner we're going
to sign off?
These two questions are off course related. When the answer to the first question is
positive, then I guess you need some conventional bid to tell partner we're signing off
somewhere. When the answer to the first question is negative, then the second question is
not necessary.
Another problem I don't quite understand is: what do you do when the answer to a
DCB-relay is the suit under which you want to play in. In the situation below you are
asking about partners hand. Partner does not know what suit you want to play in. He
therefore just gives answers to your questions and he does not know if some bid on the
six-level was to play or another DCB-relay. So how do you tell partner that a bid in this
case is a sign-off.
The situation of which I know DCB are used is after 1NT openings: (example)
1NT - 2C Stayman relay
2H - 2S relay
3S - 4C 3S: say 4-4-3-2, 4C: relay (ask number of controls A=2,K=1)
4H - 4S 4H: 5 controls (say), 4S: first DCB relay
etc.
etc.
Problems here you should agree upon is what to do with AKQ in a suit (as Douglas
mentioned earlier). So, can anyone tell me how this is resolved?
I think I read or heard somewhere that spiral scans come from the ROMEX-system. Can
anyone of you confirm this?
Best regards,
Micha Keijzers
richard e. willey
>- When you want to investigate further, do you always bid the next suit or is it possible
>to skip some suits to ask for a specific value in a specific suit?
>- When you want to sign off in a slam, can you just break the relays and bid the slam
>which you want to play or do you need some conventional bid to tell partner we're going
>to sign off?
There are a number of different schools of auction termination methods
for relay based systems.
One school which I typically associate with Polish Club uses a
comination of RKCB and a so-called "End-Signal" or Terminator Puppet.
Using this type of method, the "asker" in the relay sequence has the
option of simultaneously setting the trump suit and making a RKCB
asking bid. In this type of system, since the trump suit has been
set, asker can skip steps making the scan asking bid to indicate that
specific suits should be skipped in the response structure. (Bidding
the trump suit terminates the scan, even if this normally would be
used as a relay ask)
Alternatively, most of the systems that I have seen based off of
Symmetric Relay use a first step response following the final shape
descriptive bid as an AKQ control ask. (In many cases, the second
step is used as an AK control ask). Using this type of structure, no
trump suit is set. Any break of the relay chain is to play.
There is an obvious tradeoff with respect to available bidding space.
The terminator puppet often uses more bidding space in the early part
of the control scan, but has advantages both in signing off and also
asking for key controls.
The Denial Cue bid structure allows asker to get more information at a
relatively low level and allows more precision in deciding between
game and a small slam.
I have never seen a good mathematical simulation regarding which of
these two methods is "optimal". Every once and a while, I consider
trying to do one.
Richard
Tony T. Warnock
asking bids (perhaps modified.) The main problem I have found is that
there are hands on which asking (or relaying) is most efficient and
hands on which telling (splinters, picture bids, void showing bids,
etc.) is better. Unfortunately, most of the time, which is better can be
determined only after responder's rebid.
Alex Martelli
39071e73...@News.CIS.DFN.DE...
>
> >- When you want to investigate further, do you always bid the next suit
or is it possible
> >to skip some suits to ask for a specific value in a specific suit?
> >- When you want to sign off in a slam, can you just break the relays and
bid the slam
> >which you want to play or do you need some conventional bid to tell
partner we're going
> >to sign off?
>
> for relay based systems.
>
> One school which I typically associate with Polish Club uses a
> comination of RKCB and a so-called "End-Signal" or Terminator Puppet.
I saw that in Klinger's "Power" system, but then he does admit
to having taken some ideas from Poland in forging it.
> There is an obvious tradeoff with respect to available bidding space.
> The terminator puppet often uses more bidding space in the early part
> of the control scan, but has advantages both in signing off and also
> asking for key controls.
>
> The Denial Cue bid structure allows asker to get more information at a
> relatively low level and allows more precision in deciding between
> game and a small slam.
There may be problems in the "terminator puppet" method if
relayer needs location-of-strength information to choose the
strain. Apart from that, I fail to see the "more precision"
as a general rule.
> I have never seen a good mathematical simulation regarding which of
> these two methods is "optimal". Every once and a while, I consider
> trying to do one.
That sounds like a quite interesting idea -- how would you go
about setting up such a simulation? I'm quite interested in the
subject -- it's been AGES since I last played a Relay system,
and back in those times the slam-asks available were pretty
rudimental; but now it seems David Morgan and I might try some
Moscito variant one of these days, and we were debating exactly
this subject recently... styles of slam-asks.
I'm pretty knowledgeable/experienced about simulating stuff, but
how to set up this one is not clear to me. I guess we could
assume one partner has described a certain specific shape and
HCP-range by a certain level (results may be affected, perhaps,
by the exact values chosen for these parameters); then, we fix
two possible slam-ask-styles, deal N hands for asker and
responder (such that responder has at least a certain strength,
presumably), and manually "simulate" the bidding in each case.
But that doesn't sound much like a "mathematical simulation" to
me. Did you have anything clever in mind...? I'm quite willing
to do the donkey-work of programming the simulation if you do
the clever-idea-provider!-)
Alex
Douglas Newlands
:Douglas Newlands schreef:
:> Denial cues is a scheme for bidding several controls at once.
:>[snip]
:
:Thank's for your reaction, I have two questions though:
:- When you want to investigate further, do you always bid the next suit or is
it possible
:to skip some suits to ask for a specific value in a specific suit?
:- When you want to sign off in a slam, can you just break the relays and bid the slam
:which you want to play or do you need some conventional bid to tell partner we're going
:to sign off?
If you are using dcb/spiral scan as an adjunct to rkc then you know what trumps
are and the idea of skipping steps to avoid info in which you are not interested
is quite viable e.g.
1S-4C
4N 5D
5N you have skipped 5H saying you are not interested in QS (you hold it!)
5S would have been sign off so 5N is next relay starting at KH
6H shows KH,KD,denies QH (or JS if you agree to show jacks), don't
stop for singletons so no club cards are shown/denied.
Not really a good example since it doesn't show a saving of space;
should have one with a minor agreed for this.
If you are playing a pure relay then trumps cannot be known so you just
have to relay using the next step and everything else is interpreted as
setting the final contract.
I see mention of end signals (ES) in other posts. These are usually used
before dcb is started and is typically 4D (puppet to 4H) to say the next
bid is the final contract. I have seen this in power, TRS and a number
of weak opening relay sets of notes (these systems were very popular in Oz
during the late 80s and early 90s since anything without a pass showing values
can normally be played just about any time here).
:These two questions are off course related. When the answer to the first question is
:positive, then I guess you need some conventional bid to tell partner we're signing off
:somewhere. When the answer to the first question is negative, then the second
question is not necessary.
:Another problem I don't quite understand is: what do you do when the answer to a
:DCB-relay is the suit under which you want to play in. In the situation below you are
:asking about partners hand. Partner does not know what suit you want to play in. He
:therefore just gives answers to your questions and he does not know if some bid on the
:six-level was to play or another DCB-relay. So how do you tell partner that a bid in this
:case is a sign-off.
If the suit you wish to play in is the next step, you just cannot play there
in a pure relay system. You are committed to the next level. Most relayers
have a rather fatalistic (kamikaze?) approach - where one might consider
where all responses might leave you before using rkc or an asking bid, relayers
tend to just relay and hope the system lets them out at the right place, in
other words they don't try to avoid the above situation.
Partner has no way of interpreting that you actually want to play int he
step suit, the relayer is the captain, the absolute captain of the auction
if it has gotten this far (there is scope early in the auction for judgement
in some/many relay systems).
:The situation of which I know DCB are used is after 1NT openings: (example)
:1NT - 2C Stayman relay
:2H - 2S relay
:3S - 4C 3S: say 4-4-3-2, 4C: relay (ask number of controls A=2,K=1)
:4H - 4S 4H: 5 controls (say), 4S: first DCB relay
I have played such systems (still do with some partners) and I think that
it is often much more useful to have a range ask than to try to work out
partner's range from his controls. I prefer not to play relay systems because
all auctions tend to be on rails and all the judgement which one likes to
think one has built up over the years becomes an irrelevance sloghed off
like an old skin. For contracts up to and including game level, you have no
idea where the high cards are though you do know the shape. It's only when
you dcb that you identify where the high card are.
:I think I read or heard somewhere that spiral scans come from the ROMEX-system. Can
:anyone of you confirm this?
I knew of spiral scan, by word of mouth or exchange of notes, in the late 70s
but I have no idea about origins.
Julian Lighton
Alex Martelli <al...@magenta.com> wrote:
>richard e. willey <rwi...@isi.com> wrote in message
>39071e73...@News.CIS.DFN.DE...
>> these two methods is "optimal". Every once and a while, I consider
>> trying to do one.
>
>how to set up this one is not clear to me. I guess we could
>assume one partner has described a certain specific shape and
>HCP-range by a certain level (results may be affected, perhaps,
>by the exact values chosen for these parameters); then, we fix
>two possible slam-ask-styles, deal N hands for asker and
>responder (such that responder has at least a certain strength,
>presumably), and manually "simulate" the bidding in each case.
>
>But that doesn't sound much like a "mathematical simulation" to
>me. Did you have anything clever in mind...? I'm quite willing
>to do the donkey-work of programming the simulation if you do
>the clever-idea-provider!-)
Might it be possible to hack GIB's bidding database to do the job?
--
Julian Lighton jl...@fragment.com
"The people who have crippled you / You want to see them burn
The gates of life have closed on you / And there's just no return"
-- Black Sabbath
Douglas Newlands
:Has anyone checked if spiral scans are more efficient than Culbertson's
^^^^^^^^^
You first have to define what efficient means. One measure which
I find interesting is how much space sequences take up. I spent some
time (but unfortunately cannot find the document) looking at various schemes
of showing cards beyond rkc. For each scheme I counted how many steps
it takes to show all possible holdings and formed relative rankings.
One comparison was rkc+spiral scan compared with this scheme
1 use 6ab instead of rkc (so you know about aces and KQ of trumps)
2 next step asks for side Kings using this response scheme which
you can see has symmetric relay roots
step 1 2 kings
step 2 0 or 3 kings step bid over 2 kings
step 3 1K, lowest 2K missing lowest
step 4 1K middle 2K missing middle
step 5 1K highest 2K missing highest
3 then a similar scheme for side queens (can flow on from step 5 if
you wish).
Another comparison was with a suit bid shows that K or the other 2
side kings and 5 trumps/5N show 0 or 3 side K.
I seem to recall that there are marked differences in various schemes
but finding partners to play one's personal brainchild is not always
easy!
:asking bids (perhaps modified.) The main problem I have found is that
:there are hands on which asking (or relaying) is most efficient and
:hands on which telling (splinters, picture bids, void showing bids,
:etc.) is better. Unfortunately, most of the time, which is better can be
:determined only after responder's rebid.
My feeling is that relays are most useful when the relayer has a flat hand.
Once you have found a fit, it is necessary to identify the shortage
control (or lack of one) but one can save space by not having to relay
out fully. I have played precision with relays for flat hands starting
1C-1M-1N, 1C-2C-2D, 1C-2D-2H; with asks starting in the normal ways
but with the repeat of the ask being an instruction to go natural and
probably start cue-bidding e.g. 1C-1H-2H-2N-3H. It gives a certain flexibility.
Alex Martelli
news:zFMN4.3068$4O2.2...@newshog.newsread.com...
> In article <8e7q9...@news1.newsguy.com>,
> Alex Martelli <al...@magenta.com> wrote:
> >richard e. willey <rwi...@isi.com> wrote in message
> >39071e73...@News.CIS.DFN.DE...
> >> I have never seen a good mathematical simulation regarding which of
> >> these two methods is "optimal". Every once and a while, I consider
> >> trying to do one.
> >
> >I'm pretty knowledgeable/experienced about simulating stuff, but
> >how to set up this one is not clear to me. I guess we could
> >assume one partner has described a certain specific shape and
> >HCP-range by a certain level (results may be affected, perhaps,
> >by the exact values chosen for these parameters); then, we fix
> >two possible slam-ask-styles, deal N hands for asker and
> >responder (such that responder has at least a certain strength,
> >presumably), and manually "simulate" the bidding in each case.
> >
> >But that doesn't sound much like a "mathematical simulation" to
> >me. Did you have anything clever in mind...? I'm quite willing
> >to do the donkey-work of programming the simulation if you do
> >the clever-idea-provider!-)
>
> Might it be possible to hack GIB's bidding database to do the job?
Just see what contracts GIB ends up with, playing, say, Moscito
with either of the above slam-search set-ups? Yes, I think it
would be theoretically possible -- half of the work is already
done (I believe Moscito Byte includes 'classic' denial QB). But
the other half sounds very messy (GIB's database not being known
for ease of manipulation...:-) and the results might be heavily
affected by the specific way GIB chooses contracts -- although I
suspect the latter effect would be slight, since projecting the
possible hands for partner and 'mentally playing' with them is
a good approximation of the 'right' way to bid, still the fact
that GIB's 'mental play' is double-dummy may give some specific
'warping' (systematically bidding all grand slams in which he
knows trumps are AJxxx-KTxx, for example, since in his DD 'mental
play' he ALWAYS catches the trump Queen...:-).
Alex
Tony T. Warnock
> My feeling is that relays are most useful when the relayer has a flat hand.
> Once you have found a fit, it is necessary to identify the shortage
> control (or lack of one) but one can save space by not having to relay
> out fully.
I would add that relays are most useful when the relayee has less to show than the
relayer, relative to the start of the auction. If the auction begins with a natural
NT bid, opener may have already shown enough for responder to use relays.
Alex Martelli
3906ECAB...@sci.kun.nl...
[snip]
> I think I read or heard somewhere that spiral scans come from the
ROMEX-system. Can
> anyone of you confirm this?
They were first published by Rosenkranz, so probably first
tested in a Romex context, but his Bridge World articles on
them presented them in a non-Romex context -- I don't know
if those articles pre-date a Romex-context publication.
Alex
Micha Keijzers
Alex Martelli schreef:
> Micha Keijzers <key...@sci.kun.nl> wrote in message
> 3906ECAB...@sci.kun.nl...
> [snip]
> > I think I read or heard somewhere that spiral scans come from the
> ROMEX-system. Can
> > anyone of you confirm this?
>
> They were first published by Rosenkranz, so probably first
> tested in a Romex context, but his Bridge World articles on
> them presented them in a non-Romex context -- I don't know
> if those articles pre-date a Romex-context publication.
>
> Alex
And these articles are not public I suppose. That's a pity, because I
would like to read some literature about it. But there is a book about
the Romex-system I believe?
Greetings, Micha Keijzers
Ed Reppert
<key...@sci.kun.nl> wrote:
> But there is a book about the Romex-system I believe?
There are several. Most recently _Godfrey's Stairway to the Stars_
(GSS), which describes the Romex Forcing Club (Rosencranz and his
regular partner are now using this in, IIRC, 1st and 2nd seat NV, and in
3rd at favorable, otherwise they're using Romex. Last month I picked up,
from Baron-Barclay (http://www.baronbarclay.com) a package of three
books: the aforementioned, _Godfrey's Bridge Challenge_, a simplified
presentation of Romex, and _Bid to Win, Play for Pleasure_, which
describes the full Romex system. GSS has an appendix listing in which
book various features of the two systems are to be found. The package of
three books cost me, I think, US$14.95, which is a pretty good deal,
considering _Bid to Win_ lists for twelve bucks. :-)
I also have _Bridge, the Bidder's Game_, published in 1985, and I think
there was at least one earlier book. Rosencranz said in GSS that the
system has been around for, I think, forty years.
Barry Rigal and Sue Picus
cuebidding in the Symmetric section.
the principle is as follows. Say North has shown a 4-3-4-2 shape and 4
controls.
South relays, and North now mentally sorts his hands into by length:
spades diamonds hearts clubs (higher of two equal length suits first)
the relay is say 4C.
4D denies S top hon *** AKQ counts as no honour - you hope partner can
work it out!***
4H shows S denies D
4S shows S+D denies H
4NT shows S+D+H denies C [two-card suits now do not get looked at again)
5C shows an hon in all 4 suits and a second honor in S
etc etc
the rules are far more complex than this --the issue of when a queen is
a top honor and when to shoiw a queen behind an A/K is not
straightforward. the genberal idea is easy enough though.
Barry Rigal
Michael Arnowitt wrote:
>
> Micha Keijzers wrote:
> >
> > Hi all,
> >
> > I've posted this question in a dutch newsgroup some time ago, but got no
> > satisfactory anwers. So, I'm posting the question here: does anyone of
Barry Rigal and Sue Picus
> I have never seen a good mathematical simulation regarding which of
> these two methods is "optimal". Every once and a while, I consider
> trying to do one.
>
I'd be very interested to know the answer here.
as indicated there are various philosophical approaches.
Ultimate, Ice relay and Viking [on which a book by Glenn Groetheim and
translated /editred by me is due out very soon from Baron] have
different relatys in different places.
by contrast Symmetric focuses on ease of use --perhaps at the expense of
efficiency.
from my trawl through all of these I see plus points for each method...
not very decisive eh?
Barry
Martin RJ Carpenter
> richard e. willey <rwi...@isi.com> wrote in message
> 39071e73...@News.CIS.DFN.DE...
> >
> > >- When you want to investigate further, do you always bid the next suit
> or is it possible
> > >to skip some suits to ask for a specific value in a specific suit?
> > >- When you want to sign off in a slam, can you just break the relays and
> bid the slam
> > >which you want to play or do you need some conventional bid to tell
> partner we're going
> > >to sign off?
> >
> > There are a number of different schools of auction termination methods
> > for relay based systems.
> >
> > One school which I typically associate with Polish Club uses a
> > comination of RKCB and a so-called "End-Signal" or Terminator Puppet.
>
> I saw that in Klinger's "Power" system, but then he does admit
> to having taken some ideas from Poland in forging it.
>
Gets a mention in Matula's polish club book (They seem to use RKB then
some kind of asking bids).
One maybe unobvious thing about the style of termiantor puppets&RKB is
that it gets improved a lot if responder is forced to fall into a say 4
point range. 'Pure' relays don't mind so much coz the control info gives
some kind of range anyway.
You can get the range stuff over after a strong club but not after a 5
card major and a 2C game force relay. (And the range on a precision style
1Maj opener is a little uncomfortably wide if the lower end goes down as
far it can :)). You can do this after a 5crd major and a 1NT relay
however.... But that loses the nice 1Maj - 1NT sequences. That's yet
another hard question to answer :) Lindquist - Fredin go as far as using
step 1 after RKB in these auction as showing a minimum opener.
You can have great fun with relay sequences which finish in 3H if you
use 3S to puppet 3NT (then 4D as the puppet terminator),then 3NT/4C/4C/4D
as the RKB's with natural invites etc etc. The problem here is getting to
3H with neither player having bid NT and wrong siding 3NT :)
>
> > There is an obvious tradeoff with respect to available bidding space.
> > The terminator puppet often uses more bidding space in the early part
> > of the control scan, but has advantages both in signing off and also
> > asking for key controls.
> >
> > The Denial Cue bid structure allows asker to get more information at a
> > relatively low level and allows more precision in deciding between
> > game and a small slam.
>
> There may be problems in the "terminator puppet" method if
> relayer needs location-of-strength information to choose the
> strain. Apart from that, I fail to see the "more precision"
This problem is interesting: if you allow asking bids at the 6 level below
the set trump suit yes strain location can be tricky. But if you don't do
this then grand slams could maybe get tricky. Another problem to sort
out....
> as a general rule.
>
>
> > I have never seen a good mathematical simulation regarding which of
> > these two methods is "optimal". Every once and a while, I consider
> > trying to do one.
>
> That sounds like a quite interesting idea -- how would you go
> about setting up such a simulation? I'm quite interested in the
> subject -- it's been AGES since I last played a Relay system,
> and back in those times the slam-asks available were pretty
> rudimental; but now it seems David Morgan and I might try some
> Moscito variant one of these days, and we were debating exactly
> this subject recently... styles of slam-asks.
>
> I'm pretty knowledgeable/experienced about simulating stuff, but
> how to set up this one is not clear to me. I guess we could
> assume one partner has described a certain specific shape and
> HCP-range by a certain level (results may be affected, perhaps,
> by the exact values chosen for these parameters); then, we fix
> two possible slam-ask-styles, deal N hands for asker and
> responder (such that responder has at least a certain strength,
> presumably), and manually "simulate" the bidding in each case.
>
> But that doesn't sound much like a "mathematical simulation" to
> me. Did you have anything clever in mind...? I'm quite willing
> to do the donkey-work of programming the simulation if you do
> the clever-idea-provider!-)
>
(Assuming opener is relaying)
I think one way might be to first fix responders shape&range(if
done). Then opposite this do a load of possible openers and work
out which cards they need to be able to find in responders hand to bid
slam accurately. Then do this for all responders possible shapes. I think
you get a major simplification as you only have to treat the shapes not
the order of the suits at this stage.
(One problem could be that the systems may use differing ways to split
responders range.... Ideally I guess you do the first bit without any
range info at all then look to see which range splittings most benifit the
relays)
This sounds most likely to be possible for small/grand decisions. Having
done this it should be possible to use that info to work out which of the
methods is most efficent and even to design something wierd and definetly
non memorable but efficent.... :) Or to agree to play different relay
styles opposite different possible
shapes for responder ;)
(Basic idea is something like one style able to find trump J for grand
slam, other can do 2 fixed jacks in longest suits..... Or to help decide
how to order your spiral scans: different spiral scan orders opposite
different shapes!).
No good ideas about how to do the first bit though :) And it seems a bit
ugly....
--
****** _.
*maujv*@/ csv.warwick.ac.uk/highroans.demon.co.uk
********
" "
richard e. willey
<sue-...@worldnet.att.net> wrote:
>I'd be very interested to know the answer here.
>as indicated there are various philosophical approaches.
A couple people have expressed some degree of interest in trying to
evaluate the relative efficiencies of auction termination techniques
following a relay sequence. Here are my thoughts about how such a
study might be conducted. One necessary caveat. I am a simple
economist who has migrated into the computer industry. I am not a
serious programmer and do not claim that this type of approach is
necessarily easy or practical to implement.
With that said and done:
The first order of business to specifically define the question that
we wish to examine.
In this case, we want to compare the relative efficiency of the denial
cue bidding style of auction termination used in symmetric relay with
the "End Signal" style used by other systems.
(The precise definitions for the different relay methods to be
compared can wait until we agree on the testing methodology)
We will address this question using a Monte Carlo simulation.
Some simplifying assumptions will be used to allow the process to be
automated.
A dealer program will be used to generate pairs of bridge hands
containing 27+ HCP.
Both hands A and B will always be dealt a minimum of 8 HCP (we can
use whatever other definition people prefer to show a positive
response to a strong club opening)
If Hand A has between 8 - 9 HCP, hand A will "Pass" and then "Show"
during the relay sequence.
If hand A has between 10 - 15 HCP, hand A will "Open-Limited" and then
"Show" during the relay sequence during the relay sequence.
If Hand A has 16+ HCP, hand A will ask during the relay sequence.
Assume that two pairs are using a "Relay-to-Death" bidding style such
as symmetric relay.
Each pair will relay until the complete hand shape is determined.
Each pair will resolve final shape with the same bid.
basic symmetric relay resolution tables will be used for the sake of
simplicity)
Each pair will be allowed to declare the hand in its "best"
denomination.
[I would suggest that the best way to determine this would be to use a
double dummy engine like GIB.
Use the fixed hands as input and generate "N" complete bridge deals.
The double dummy engine can then be used to determine an appropriate
trump suit and safety level for each deal. (As I recall, GIB supports
the concept of a "par" contract").]
Once the appropriate trump suit has been set, it should be possible to
define an array that describes the key pieces of information that must
be located to determine whether or not slam is a reasonable
proposition.
If we can define a set of rules describing when the asker should
terminate the relay chain, then we can compare the number of bits of
information that the two relay methods are able to exchange. For the
purpose of making this determination, we will make two assumptions.
Scoring is IMPs.
Defenders are perfect.
At any given point during the scan, there exist sets of controls that
could be possibly be held by the hand that is showing.
At any given point in time during the auction, asker has two choices:
He can terminate the auction or he can choose to ask for the next
missing control in the array.
[I recognize the existence of cases where a player using terminator
puppet methods might chose to skip over an missing control. I am
chosing to ignore them to simplify the search tree]
We can define a probability density function describing the possible
contracts following a relay ask.
Asker will continue to relay so long as making an asking bid leads to
a positive expected value at IMP scoring.
[This is clearly the trickiest part of the entire process, since the
expected value at step "N" of the process depends on the bidding
choices available at step N+1. I'd be very curious to see how GIB
implemented this decision tree]
Suppose that the two hands are
Hand A Hand B
xxx x
AKxxx Qxxx
AQ Kxxxx
Axx KQx
In this case, hand A asks and hand B shows.
Hand A started with a strong club opening. Hand B is unlimited.
Hand B is a 5431 and resolves final shape with a bid of 3D.
From the perspective of hand A, the Key controls that need to be
identified are
[SA, HA, HK, AQ, DA, AK, DQ, CA, CK, CQ + CJ]
Several of those controls are located in asker's hand, which collapses
the actual set of cards that must be identified down to
[SA, HQ, DK, CK, CQ + CJ]
As I said, the now the tricky part starts.
As I mentioned, we might be able to derive a solution technique using
the GIB source. I'll see if I can figure anything out over the
weekend.
Richard