[DavidKok]

Prompted by the other ongoing DCB thread and a great exchange of ideas in private, I am considering running simulations on denial cue-bidding. The goal of these simulations would be to compare different choices people make in setting up their DCB, hoping to see if simulations can give some insight into which rules are clearly profitable and which are clear stinkers. As always all the caveats of computer simulations and double dummy analysis for scoring apply. Nevertheless I think this is a useful tool for gaining insight, as long as we are careful with interpreting the results.

My thought is to implement the DCB as faithfully as possible, programming the rules in the script and then having it tell me the final contract. The merits of this contract can then be evaluated using DD simulations - e.g., keeping our relay bidders hands fixed, running it 100 times or so randomising the other two hands to get an average and comparing this against other plausible contracts. It's easy enough to (later) include a number of variations on this idea, e.g. splitting the hands by relay points, by hand shape or feature length, by degree of fit and whatnot.

There are two questions that came up already which I'd love more input on before I start on this.

• What is the stopping rule? When do we anticipate a death response and sign off instead? I imagine this will have a major impact on the final contract reached, and typically relies on counting tricks and visualising the hand. What would be a good way to approach this programmatically?
  
• What is the interaction between shape, QPs and inferences on exact honours and trick-taking potential? Do I need to restrict initial approaches to only certain fixed hand shapes, or can this easily generalise?

If you think simulations like this might be interesting, please share any thoughts you have on what might be a good way to go about building such a simulation.

[foobar]

For #2 above, I think 9+ card fits with 18-21 combined QPs is likely the most interesting range. With 22+, it's just a question of small or grand, and those hands are relatively rare.

One interesting aspect is that control responses might rule out a slam early, but might conversely miss out on the ones where there are sufficient controls, but there's a risk of responder being thread bare.

[DavidKok]

I do want to point out that I'm always going to condition on opponents not having a suitable overcall of the 1♣ or the initial response, so many of the 9+ card fits will be less common by negative inference.

[foobar]

 DavidKok, on 2026-July-02, 05:14, said:

I do want to point out that I'm always going to condition on opponents not having a suitable overcall of the 1♣ or the initial response, so many of the 9+ card fits will be less common by negative inference.

Sure, it's just that 18 QPs are seldom interesting for slam purposes unless also accompanied by a 9-card trump fit.

[foobar]

While we await others, here's my understanding of KK-relay DCB:

- AK controls based. Stop shows AK or none
- First scan examines at most three suits and the shortest is always skipped
- For 54xx, 64xx, with a singleton or void at most two suits are scanned
- First round scanning stops as soon as all AK controls have been enumerated and immediately proceeds to Queen scan starting with the longest suit. So, for example on say a 6322 hand with three controls and say A...6th in the long suit, and K in the second longest, we will skip two steps for the A+K, then immediately scan for the Queen in the longest suit
- The Queen scan examines all 2+ card suits
- There is an exception for DCB scan if responder has shown 1/2 controls. In this case, we stop with one control and proceed with the DCB with 2

[pilun]

I was thinking about demoting stiff honours after this hand came up.

---

5♠ survived but might well have failed opposite ♠Qxxxxxx

The 4♣ ask seemed okay because 6 SPs will usually be useful. Most likely is
♠KQ plus red K & Q, where 5♠ is pretty solid and slam is probably on a finesse.

So this was a hand where demoting the stiff ace would have helped.
The issue is that treating a stiff ace as 2 SPs might make other DCB auctions more ambiguous.

In fact this was a good hand for our 4♦ Mild Slam Try. This asks describer to bid 4♥ on all minimums OR a max (Base +2 SPs) with bad hearts. Not Hx(x) or HHxx(x). 4♦ basically says "My next bid will be a natural slam try, requiring good trumps".

[foobar]

Here's the description of PCB to be best of my knowledge.Adam (awm) can correct as needed:

- QPs, but no singleton is discounted
- All suits are scanned once, starting with the shortest suit in case of a singleton
- The general rule is skip with odd parity in A/K/Q and stop with even, except that with a singleton honor, stop with (unexpected case), and skip without (usual case)
- After the initial scan, next ask scans Jacks
- 6N is never an ask

[awm]

 foobar, on 2026-July-05, 07:54, said:

Here's the description of PCB to be best of my knowledge.Adam (awm) can correct as needed:

- QPs, but no singleton is discounted
- All suits are scanned once, starting with the shortest suit in case of a singleton
- The general rule is skip with odd parity in A/K/Q and stop with even, except that with a singleton honor, stop with (unexpected case), and skip without (usual case)
- After the initial scan, next ask scans Jacks
- 6N is never an ask

Sam and I play this additional step after all suits are scanned, which works as follows:
1. If describer has a suit with king but no queen and another suit with queen but no king, stop if the king is in the longer (or earlier scan) suit and take a step if it’s in the shorter (or later scan) suit.
2. If #1 does not apply, but describer has a suit with KQ and another (non-singleton) suit with no honor, stop if the KQ is in the longer (or first scan) suit.

We also never zoom directly into jacks.