[straube]

Rob F, on Mar 26 2010, 11:01 AM, said:

Free, on Mar 26 2010, 06:02 AM, said:

Rob F, on Mar 26 2010, 10:53 AM, said:

You could even do stuff where the steps ordered which suit to start with, when shape is only partially known, i.e. step = shape relay, step+1 = start with S, step +2 = start with H, etc.

Yes, you have many shapes where this is applicable. Many so called "12 or 14 card hands". Here are some examples:

Balanced:
4333 with unknown 4-card = 3333
4432 with unknown 3-2 (4-4 known) = 4433
5332 with unknown 3-3-2 (only 5 card known) = 5333
(If you combine 5332/4333 in 1 bid (with known 4 card suit), you can also consider this 5333)

...This saves a lot of space (which is most useful with balanced hands).!

I agree that this is most useful for balanced hands. For example, to facilitate this, I designed my own version of balanced symmetric relays to replace the default "Crash"-style ones. In mine, for example, you know before the final ask information like this:

4M333
4m333
43=(42)
42=(43)

with the obvious priority that knowing exact major shape (even only 3 cards) is more important for strain decisions than resolving the minor shape. Of course you can still ask for shape, but you have most of the information you want in case you want to break relay early, something that the TOSR Crash relays for 4432's are pretty bad about (you must always ask with any 4cM since they might fit you). You can even reprioritize the shape asking bid to be higher than the next step...

1C-1S GF bal, among others...
1N-2H bal, no 4cM
2S-2N 4m333

3C QP ask
3D asks for the minor
3HS RKC for majors, or stopper asks, to taste
etc

Some really good stuff here. Can you write out the structure? Our balanced hands start to unwind at 2D. We might have a complication, however, because our balanced hands include all 5332s (no 4441s).

[rbforster]

straube, on Mar 26 2010, 01:06 PM, said:

Some really good stuff here. Can you write out the structure? Our balanced hands start to unwind at 2D. We might have a complication, however, because our balanced hands include all 5332s (no 4441s).

I'll write it up later and maybe post a link. It does only cater to 4333/4432 shapes, following TOSR's approach of including 5332 as single-suited. While I don't really agree with that style (I think 5332 is "balanced" generally), putting an extra 12 hands into your 2D+ balanced relays will result in higher shape resolution on all these common hands while the single-suited hands (which are much rarer) will resolve lower. So on a frequency basis, it's probably better to exclude 5332 from balanced for practical relay purposes, even if it's not ideal from a theoretical perspective.

[akhare]

Rob F, on Mar 26 2010, 01:23 PM, said:

straube, on Mar 26 2010, 01:06 PM, said:

Some really good stuff here.  Can you write out the structure?  Our balanced hands start to unwind at 2D.  We might have a complication, however, because our balanced hands include all 5332s (no 4441s).

I'll write it up later and maybe post a link. It does only cater to 4333/4432 shapes, following TOSR's approach of including 5332 as single-suited. While I don't really agree with that style (I think 5332 is "balanced" generally), putting an extra 12 hands into your 2D+ balanced relays will result in higher shape resolution on all these common hands while the single-suited hands (which are much rarer) will resolve lower. So on a frequency basis, it's probably better to exclude 5332 from balanced for practical relay purposes, even if it's not ideal from a theoretical perspective.

To provide more context, the balanced scheme is:

2D: No 4CM / any 4333 -> 2S=any 4333, 2N = 5332 with C, 3C=D+C
2H: 4H, not 4S -> 2N = 5332 with H, 3C=H+C, 3D
2S: S+m
2N: 5332 with S
3C: S+H

Note that there's another simplified balanced scheme in a different context includes all balanced hands (including 5332 that starts 2D):

...2S: Any 4432
...2N: Any 4333
...3C+: 5332 hands

[awm]

straube, on Mar 25 2010, 02:13 PM, said:

Adam, I know it's a little off topic, but what is your dcb structure after you show QPs? We use QPs but sometimes it seems like we're running out of room.

We show the exact number of QP first. This is helped by the fact that our initial response divides into 2-6 QP (responder describes) and 7+ QP (responder bids 1♦ and reverse relays). With this done, we scan through all suits in order of length (including singletons, but not voids of course). We show even or odd parity of the top three honors (so stop with 0 or 2, scan on with 1 or 3). After this, we describe the nature of the honor in the first one-honor suit (stop if it's a king, scan on if it's an ace or a queen). Extensive simulation has shown that this disambiguates almost all cases.

[straube]

awm, on Mar 26 2010, 01:35 PM, said:

straube, on Mar 25 2010, 02:13 PM, said:

Adam, I know it's a little off topic, but what is your dcb structure after you show QPs?  We use QPs but sometimes it seems like we're running out of room.

We show the exact number of QP first. This is helped by the fact that our initial response divides into 2-6 QP (responder describes) and 7+ QP (responder bids 1♦ and reverse relays). With this done, we scan through all suits in order of length (including singletons, but not voids of course). We show even or odd parity of the top three honors (so stop with 0 or 2, scan on with 1 or 3). After this, we describe the nature of the honor in the first one-honor suit (stop if it's a king, scan on if it's an ace or a queen). Extensive simulation has shown that this disambiguates almost all cases.

Say I've shown 6 QPs and 5-4-3-1 shape at the bid of 3S.

Assuming 4C is dcb relay, how would the bidding continue for

Kxxxx xxxx KQx K

assuming the singleton K is counted as 1 QP

[awm]

Well we do count 2 QP for singleton kings. But anyway:

4♥ = 1 or 3 top spades, 0 or 2 top hearts
---- 4♠ relay
4N = 0 or 2 top diamonds
---- 5♣ relay
5♥ = 1 top club, spades headed by the king (stopping to show king, run on to show A/Q)
---- 5♠ relay
5N = no spade jack
.......

[akhare]

straube, on Mar 26 2010, 02:16 PM, said:

Say I've shown 6 QPs and 5-4-3-1 shape at the bid of 3S.

Assuming 4C is dcb relay, how would the bidding continue for

Kxxxx xxxx KQx K

assuming the singleton K is counted as 1 QP

Here's the same hand based on another DCB scheme called "Go Fish" starting with 4♣ = ask for queens:

...4♦: 1 Q
........4♥ ask for kings
...5♥ -- No K♥, perforce only K♠ or 3 Ks, including the other two

Here's an unedited post by John Sheehan (of Prism Signals fame) about his GOFISH alterative to DCB when using RPs (aka AKQ points).

Hi all,

Dlr: W
Vul: EW
Auction: Moscito Variation

West East
A863 K5
A652 K8
Q A4
KQJ5 A876432

1C 2D
2H 2S
2N 3C
3D 3H 2=2=2=7.
3S (1) 4S 10 relay points.
4N (2) 5S (No queens / all 4)
5N (3) 6C Same rank.
7C

(1) 'Number of Relay points?' (A=3 K=2 Q=1)
(2) 'Do you have any queens?'
(3) 'Tell me about your two kings.'

Note: The first step after the relay points response
asks,'Do you have any queens?'

In response to the queen ask:
Step 1 = 1 Q.
Step 2 = 2 Q's.
Step 3 = 3 Q's.
Step 4 = 0 Q's. Go Fish! (or 4 Q's)

If The number of relay points is known; and,
The number of queens is known; then,
The exact number of aces and kings is known.

Note: 'Tell me about your kings.'

The first step after the number of queens response
asks about location of kings.

(If you are more interested in queens
than kings, skip to the second step to
ask about the location of queens.)

In response to, 'Tell me about your kings.'

To show one or three kings, stop at the bid that
corresponds in the scanning order
to the one suit with (or without) the king.


If two kings are held; then
Step 1 = 2 K's of the same Rank
Step 2 = 2 K's of the same Color
Step 3 = 2 K's of the same Shape (C/H D/S)



Dlr: W
Vul: EW
Auction: Moscito Variation

West East
A7 KQ85
AQ3 K852
AQJ874 K6
J4 A96

1C 2D
2H 2S
2N 3C
3D 3H 4=4=3=2
3S(1) 4S 10 relay points.
4N(2) 5C 1 Queen.
5D(3) 5N No club king.
6C(4) 6D Spade queen.
7N

(1) 'Number of Relay points?' (A=3 K=2 Q=1)
(2) 'Do you have any queens?'
(3) 'Tell me about your kings.'
(4) 'What about that queen?'

Is Go Fish! for relay points a practical approach to solving the problem
of discoverying and pinpointing honors in slam auctions?

Johnny Sheehan
www.prismsignals.com

[straube]

Adam, am I right here? It seems like your structure gives a poor picture of responder's hand early on and then a very good picture at a higher level...especially when the nature of the first single honor is shown. So the relayer has to be willing to play a high level (like 5S or 5N) before contemplating an ask.

The vanilla structure that Atul and I have been using lets us abort early but often with an incomplete picture of partner's hand.

Would your structure work for us if we counted the singleton king as 1? (thinking it would). I assume you count a singleton queen as 1 and that you try to show whether you have it or not. Perhaps we could count it as 0 but show whether we have it or not (like a bonus). Or maybe that would throw confusion into things.

Anyway, I think I like your structure. I'm not quite getting the GoFish structure.

[awm]

The thing is that many inferences are available from opener's honor holdings. Usually if opener can't tell what's going on fairly early, it actually doesn't matter because too many honors are missing for slam to be practical. It's actually not unusual once QP are shown for opener to figure that there are only a small number of possible honor patterns, and to be able to sort them out fairly early.

Another observation is that we do play RKC asks as well. The hand types where opener is often unable to tell early in the sequence what's happening often correspond well to the RKC ask hands. I don't really know how to explain this further except to recommend trying it.

[straube]

Atul and I tried (your) Adam's structure last night and it worked really well. Thanks a lot.

[shevek]

DinDIP, on Feb 26 2010, 07:08 AM, said:

jukmoi, on Feb 25 2010, 05:54 AM, said:

Even though I have never tried it, I would expect AKQ321 is much harder to read. I mean that with AK21 you often can divine the exact location of honour cards or for example that they must be 2 aces (you hold 3 kings).

It is obviously harder to determine teller's honour combination from just the response (showing some given number of relay points) because, unless that number is very small, there are many more permutations that are possible than for a comparable number of AK21 controls.

But there are ways to reduce that number of permutations significantly: a cheap way of doing so is to have teller show his (non-singleton) king parity* before entering into DCB. One could do as the Swedes did (still do?): stop with even and zoom with odd. Better, however, is to stop with even when holding 2, 3, 6, 9, 12 or 15 RP and stop with odd with other holdings. This is easy to remember (stop with even when 2 or a multiple of 3) and much closer to the actual frequency. (Of course, the parity answer should also depend on how many RP asker has but I don't now of a system that can show that.)

I've done a lot of testing that shows that this method is better than not asking for king parity or asking for ace (or queen) parity; it's still not clear to me whether this is the best method of honour-showing/DCB. (Matt Ginsberg tells me it is clearly not for a computer program but we're interested in what humans can remember at the table.)

David

*This assumes that you wisely count a kingleton as 1RP not 2; again, extensive testing shows this to be much better.

Well done in ITS. Have you inflicted stuff like this on Justin?