[shevek]

When I learned symmetric decades ago, we used positive cueing for singletons (& stiff K = 1 SP). I note that most relay pairs these days ignore singletons, since they can usually infer that holding.
Here's a fudged example where direct knowledge of the singleton is crucial:

Scoring: IMP

---

East manages to stop in 4NT or 5♦ opposite these 6 SPs.
However, likely for West is
♠Axxx   ♥Axxx   ♦xxxx   ♣x

Yes, East should have asked for kontrols. Too late.
Anyway, is the ability to sort out these rare cases more than compensated by the step saved if your method is to ignore singletons?

[rbforster]

shevek, on May 20 2009, 09:30 PM, said:

Yes, East should have asked for kontrols. Too late.

Certain hands are better suited for asking for aces; others for asking about general strength. For the former, RKC approaches (or AK controls) work well while for the latter AKQ points and denial cue bids work better. Your given hand is a perfect example of wanting to ask for aces - all you need is partner's number of aces to place the contract (in 4N/6♦/7♦).

I play methods where the cheapest relay step inquires about AKQ points. However, higher steps (aside from signoffs like 3N or 4♦ end relay) start RKC inquires in each suit. I think the combination of these methods is definitely worthwhile. While DCBs are a great tool, they are not always the right tool and it's worth learning to recognize when not to employ them.

shevek, on May 20 2009, 09:30 PM, said:

Anyway, is the ability to sort out these rare cases more than compensated by the step saved if your method is to ignore singletons?

I count all the honors (including stiff honors) for AKQ points but also skip singleton suits in DCB responses. It's usually possible to determine all of partner's honors through denial cues and process of elimination, although sometimes this isn't the case. Especially if you are missing too many honors yourself, there may still be ambiguity, such as below where the West hands both have the same 6 AKQ points (and others are possible too):

7N or 7♦?

---

7N or 7♦?

---

Sure the first hand will deny a 2nd honor in the majors eventually which will allow you to separate them, but having to stop repeatedly in clubs and diamonds may prevent full resolution at a safe level.

[dake50]

I think relays get it backwards. Show lo-mid-hi-no single/short before 2nd suit and residual. Then scan if 'right' single found. Wrong single signs off on middling slam tries.

[gwnn]

Usually infer? You mean infer 99.9% of the time

[awm]

My approach has been to include singletons in the scan when using AKQ points. A large number of simulations validated that this is substantially helpful in resolving partner's high cards. The issue is that it's often difficult to distinguish between AK+small singleton and AQ+singleton Q and these play very differently. Note that this difficulty is much rarer if you are scanning only controls since if you are considering slam you're usually not missing very many controls (whereas you may be missing several potentially non-relevant queens).

I disagree with gwnn that you can usually infer side queens (aces/kings he is right). For example:

♠xx
♥KJxx
♦Axx
♣AKQx

Partner has 4414 with seven AKQ points. These include two top hearts and one top spade. Is it:

♠Axxx
♥AQxx
♦x
♣xxxx

where six of a round suit is excellent? or is it:

♠Kxxx
♥AQxx
♦Q
♣xxxx

where six of a round suit requires the spade finesse and some suit breaks? or is it:

♠Qxxx
♥AQxx
♦K
♣xxxx

where we are off two top spade tricks and can't escape even if they fail to lead the suit?

[akhare]

awm, on May 21 2009, 02:37 PM, said:

♠Qxxx
♥AQxx
♦K
♣xxxx

where we are off two top spade tricks and can't escape even if they fail to lead the suit?

Isn't the last case easy to identify? Assuming that you stop on the first round scan with either AKQ or no AK, the sequence might go something like this (starting with say 3♠ = 4=4=1=4 with 7 QPs):

4♣ (1) - 4♦ (2).....................1: DCB.....2: AKQ of ♠ or no AK
4♥ (1) - 4N (3).....................1: DCB.....3: AKQ of ♣ or no AK

Given that this example assumes missing AK of ♠, relayer must hold AK of ♣, A♦ and K♥ for the auction to make any sense at all.

Ergo, after 4N it should be clear that responder is missing AK of ♠ (marked with A♥, AKQ of ♠ would give too many. Note that AQ♥, K♦, Q♣ with XXXX in ♠ is possible too).

On a side note, I have seen some people count only stiff A/K and exclude stiff Qs in the reported count. Is there an advantage one way or the other?

[keylime]

With Larry and I we use DCB involving singletons once the shape is known.

[awm]

It obviously depends on how you define your scans. Using akhare's methods, he will have problems with:

♠Kxxx
♥Axxx
♦K
♣Jxxx

versus

♠Axxx
♥AQxx
♦x
♣Jxxx

In either case he shows the same AKQ count, then shows "at least one of A/K but not AKQ" in each major. Yet slam opposite the first hand is awful and slam opposite the second hand is nearly cold.

In my methods where the parity of the honor holding (i.e. 0/2 or 1/3) is shown, it is easy to distinguish these cases regardless of whether we scan singletons, but harder to distinguish some of the other cases without scanning for singleton honors.

Also note that bidding keycard is only of mild help here with spades wide open; in particular we can consider:

♠Qxxx
♥AQxx
♦K
♣Jxxx

♠KQxx
♥AQxx
♦x
♣Jxxx

where the same number of keycards for hearts are held in either case, yet the first hand gives no play for slam and the second makes slam almost cold. This holds even if you get the AKQ count before bidding keycard (which can be difficult, from a space perspective).

[shevek]

akhare, on May 21 2009, 03:02 PM, said:

awm, on May 21 2009, 02:37 PM, said:

♠Qxxx
♥AQxx
♦K
♣xxxx

where we are off two top spade tricks and can't escape even if they fail to lead the suit?

Isn't the last case easy to identify? Assuming that you stop on the first round scan with either AKQ or no AK, the sequence might go something like this (starting with say 3♠ = 4=4=1=4 with 7 QPs):

4♣ (1) - 4♦ (2).....................1: DCB.....2: AKQ of ♠ or no AK
4♥ (1) - 4N (3).....................1: DCB.....3: AKQ of ♣ or no AK

Given that this example assumes missing AK of ♠, relayer must hold AK of ♣, A♦ and K♥ for the auction to make any sense at all.

Ergo, after 4N it should be clear that responder is missing AK of ♠ (marked with A♥, AKQ of ♠ would give too many. Note that AQ♥, K♦, Q♣ with XXXX in ♠ is possible too).

On a side note, I have seen some people count only stiff A/K and exclude stiff Qs in the reported count. Is there an advantage one way or the other?

David Morgan did analysis that suggested ignoring stiff queens and counting kingletons as 1 SP, so we go with that.

Of course you miss slams with stiff Q opp AKJxxx

[DinDIP]

shevek, on May 21 2009, 09:43 PM, said:

akhare, on May 21 2009, 03:02 PM, said:

On a side note, I have seen some people count only stiff A/K and exclude stiff Qs in the reported count. Is there an advantage one way or the other?

David Morgan did analysis that suggested ignoring stiff queens and counting kingletons as 1 SP, so we go with that.

Of course you miss slams with stiff Q opp AKJxxx

Some system designers (such as Bo-Yin Yang in Terrorist Moscito) argue that all stiff honours should be demoted one rank, i.e. counting a stiff A as 2 AKQP, kingleton as 1 AKQP and stiff Q as 0 AKQP. My testing of this showed that, while this was reasonable when measuring the trick-taking potential of the honours, it was inaccurate when measuring their utility as controls (which is what we are considering in DCB). Too many slams were missed when R assumed that RR had a non-singleton king rather than a stiff ace. DCB resolved some of those ambiguities but, too often, it was too dangerous to investigate.

And, yes, one does miss slams with stiff Q opposite a good suit but something has to give.

On shevek's original question of whether or not to scan singletons: I'm uncertain. My testing suggests it is necessary much less often if you use some parity scan before DCB (such as stopping with an even number of non-singleton kings and zooming with an odd number) that enables R to reduce, often significantly, the number of permutations of honours RR might have. Without such a scan, my testing suggests scanning is worthwhile but my data is limited here.

David