Auction Termination Methodologies

[straube]

Straube is testing....

 

S1-Parity Cue Bidding*

S2-End Signal-zooms with 9 QPs

S3-RKC Clubs

S4-RKC Diamonds

S5-RKC Hearts

S6-RKC Spades

 

I'm not actually using my own system because it won't be a good comparison for continuations. I'm presuming we are in a GF auction and responder has shown a base of 6 QPs and has shown pattern at standard symmetric +0.

 

*

1. All suits, including singletons, are scanned in order of length; equal length ties are broken in rank order, highest to lowest.

 

2. On the first pass, stop in the first suit with even honor (AKQ) parity and continue scanning with odd parity. This continues until all suits have been scanned. The exception to this rule is when singletons are scanned. Since most singletons won’t have an honor, it is efficient to skip with 0 but stop with either ace or king (not queen).

 

3. After all suits have been scanned, the first suit with odd parity is reexamined; RR will stop with the K or AKQ but will skip with the A or Q. If all suits have even parity, skip when the first scanned suit has the king.

 

4. After king parity (for the first odd suit) has been determined, jacks are scanned using the same suit order, skipping with and stopping to deny the jack in focus. Jacks are not scanned until captain specifically asks (no zooming). This explains why we skip with the king (it is a less common occurrence and thus uses less room on average.

[rbforster]

I'll bid hrothgars old slam methods from his Moscito notes (S1=QPs, with denial cues; 3N/4♦ end signals with 9+ QP zooms, and the rest RKCs by responders length). I'll use my own strong club methods, which almost always do relays by the strong hand.

[hrothgar]

I'm also going to be bidding based on my MOSCTIO notes

 

It will be interesting to see whether the auctions are at all similar.

[awm]

I'll use a modified version of IMPrecision, since we would not actually relay (or would reverse-relay) on some of the hands he's going to give. I'll assume:

 

1. Symmetric-style shape resolution (so 3♦ for 5431)

2. Step one (but not 3NT) = QP ask with range of 5-11 (our standard range for an opening 1♦ bid), followed by parity scans and "one honor suit parity"

3. 3NT = to play, 4♦ end signal

4. Relay breaks below 3NT = stopper ask (three card fragments, then two card fragments)

5. Relay breaks above 3NT = asking controls outside a specific suit starting with 0-1, suits scanned in length order (tie break S > H > D > C)

[akhare]

Just for fun, I will throw in:

 

S1: QP ask, followed by DCB

S2 - S5: RKC ask based on length with ties broken in symmetric order (except for the canonical 3N and 4♦ canonical terminators)

[rbforster]

Just for fun, I will throw in:

 

S1: QP ask, followed by DCB

S2 - S5: RKC ask based on length with ties broken in symmetric order (except for the canonical 3N and 4♦ canonical terminators)

This is going to be pretty similar to what I'm doing, up to a few small tweaks / preferences I use. For example, my first step after QPs is <=6 (we GF based on HCP and shape, not requiring strictly 6 QPs for a positive). Also, I count QPs for singleton honors, don't scan singleton suits, and don't scan suits in the 2nd pass when they showed 0/3 AKQs in the first pass.

[straube]

I feel like Adam and Rob have disadvantaged themselves with respect to Moscito and what I have modeled. Would they consider starting their gf at 6 QPs? If not, we are not at the same starting place and the results we get will be biased against their methods.

[hrothgar]

I feel like Adam and Rob have disadvantaged themselves with respect to Moscito and what I have modeled. Would they consider starting their gf at 6 QPs? If not, we are not at the same starting place and the results we get will be biased against their methods.

 

Aren't we trying to test whether stuff like "Start QPs at 6" is better?

[straube]

Aren't we trying to test whether stuff like "Start QPs at 6" is better?

 

That's not my interest. No. I want to compare DCB vs PCB vs RKC etc at the most similar starting conditions as possible. If you start at 6 compared to someone who starts at 5 you obviously have an advantage.

[akhare]

This is going to be pretty similar to what I'm doing, up to a few small tweaks / preferences I use. For example, my first step after QPs is <=6 (we GF based on HCP and shape, not requiring strictly 6 QPs for a positive). Also, I count QPs for singleton honors, don't scan singleton suits, and don't scan suits in the 2nd pass when they showed 0/3 AKQs in the first pass.

In that case, I won't post an auction unless it's different (subjectively or otherwise). BTW, I was planning on using 5+ QPs for the positive responses, with the first step in the QP ask showing 5/6.

[PrecisionL]

When we designed Ultra Club Relay and its upgrade Copious Canape Club (C3) we decided to avoid an artificial termination bid and just go with (1) distribution relays, (2) Beta confirming fit, and (3) TURBO with Denial Cue Bidding.

 

I will be following the posts very carefully to learn what seems best in the majority of cases.

[rbforster]

In that case, I won't post an auction unless it's different (subjectively or otherwise). BTW, I was planning on using 5+ QPs for the positive responses, with the first step in the QP ask showing 5/6.

Go ahead and post yours. I've already seen several differences in style and also system with hrothgars and we've only done 5 hands.

[Free]

I will use the following method since it looks quite good and I changed my opinion about suit order for RKC:

Step 1 = QP ask when it's not 3NT.

Step 2 (below 3NT) = puppet*, partner zooms with 3+QP more than promised

Step 3 = RKC

Step 4 = RKC

Step 5 = RKC

(Step 6 = RKC)

If step 2 = 3NT or higher, then I'll play the original method: 3NT signoff, 4♦ = Terminator, step 1 = QP ask, step 2-5 = RKC.

 

* not necessarily an end signal, bidding a non-game contract afterwards is used for RKC (=> step 6 may not be needed)

 

Remark about the order of RKC:

- order longest to shortest

- when equal suits, use the following priority: ♣>♦>♥>♠

- in case step 2 is lower than 3NT: RKC 1 = step 3 ; RKC 2 = step 4 ; RKC 3 = delayed step 4 ; RKC 4 = step 5

 

It's even possible to add exclusion QP asks as well (suit order identical to RKC), since we can stay relatively low when our puppet is lower than 3NT (direct and delayed steps keep us low). However I don't have a clue how the base QP level should be adjusted and how you should respond.

 

Rules for scanning:

- Step 1 asks to continue, every other bid is signoff (so 3NT can be a relay)

- Singleton K/Q don't count for QP's

- When holding 10+ cards in 2 suits the QP base is lowered by 1

- Scan suits in the same order RKC's are determined

[sieong]

By the way, does anyone know of the JVCB method (see http://dipbridge.wordpress.com/2008/12/ for a variation - I do not know of the original method). I have heard from some Swedish players (Peter Bertheau and Frederic Wrang; try talking to them at the nationals, they are usually very friendly) that this method is superior to anything else they have used. My small sample (< 10, inconclusive) suggests that it is worse than PCB in terms of average resolution of AKQ locations; PCB is on average 2-3 steps cheaper than JVCB described on the webpage above. The goal of DCB/PCB/RKCB is obviously not only to resolve the honor locations, but also to stop low when they are wrong, so just measuring the average resolution level is inadequate, but I do wonder if I am missing something here. Does anyone know more about JVCB? I wonder if Ulf (ulven@) is watching this thread and would like to comment about it?

 

Btw, PCB is not really designed for positive responses to 1C (although it does a pretty good job so far, I think). If I were to make a tweak, I probably would add a parity step in terms of controls when QP is guaranteed to be 19+. This can usually completely resolve the honor structure (but not location). Given that I would certainly relay with some 11QP hands (I presume with a weaker hand, there are options to break relays earlier), this would mean when responder has 8+RP.

 

So without testing this at all and at the risk of looking very foolish, here is a variation of PCB for positive responses to 1C that probably won't work well out of the box but maybe some tweaks of which will:

 

If <8RP, regular PCB (I will not post on these hands since the auction will be identical to Adam's).

 

If >=8RP, stop if even # controls, skip if odd # controls, proceed to PCB but shift honors down by one (so parity over KQJ).