[enigmisto]

The KK Relay book describes a series of shortcuts when doing Denial Cue Bidding. For example, after you've communicated the Aces and Kings in your hand, you skip any remaining suits and begin scanning the Queens. The assumption is that your partner can infer all your Aces and Kings from their own hand and your control point total, so no need to continue.

How are you supposed to alert these bids in a game?

You and your partner both (theoretically) know that you are now scanning for Queens, and know which suit you're on, but it's based on private information that the other partnership doesn't know. It's a very similar issue to other sorts of "encrypted bids".

I can only imagine how the discussion might go. "What does that bid mean?" "It's a denial cue bid which allows me to know whether my partner is holding or not holding certain honors." "Which suit?" "That depends on our holdings." "What honors?" "That depends on our holdings."

So what do you actually type into the alert box on BBO? How do other partnerships deal with that? Do they complain to the director?

[mycroft]

Please give a specific example. Whenever I have seen this in other DCB systems, relayer knows:

• Responder's complete shape
  
• Responder's AK, or AKQ count
  
• and the answers to the previous calls, and what they mean (even if what they mean is "0 or 2 AKQ in spades, but one in diamonds. 0 or 2 in hearts...")

And the possible explanation is significantly more than "depends on her holdings".

But I don't know enough of the system, or of DCB in general, to know what without an example.

[foobar]

 enigmisto, on 2022-January-28, 05:32, said:

The KK Relay book describes a series of shortcuts when doing Denial Cue Bidding. For example, after you've communicated the Aces and Kings in your hand, you skip any remaining suits and begin scanning the Queens. The assumption is that your partner can infer all your Aces and Kings from their own hand and your control point total, so no need to continue.

How are you supposed to alert these bids in a game?

You and your partner both (theoretically) know that you are now scanning for Queens, and know which suit you're on, but it's based on private information that the other partnership doesn't know. It's a very similar issue to other sorts of "encrypted bids".

I can only imagine how the discussion might go. "What does that bid mean?" "It's a denial cue bid which allows me to know whether my partner is holding or not holding certain honors." "Which suit?" "That depends on our holdings." "What honors?" "That depends on our holdings."

So what do you actually type into the alert box on BBO? How do other partnerships deal with that? Do they complain to the director?

As mycroft noted, specific examples will help, but in general, the opponents are entitled to complete disclosure about the methods, and not specific inferences that the relay captain (RC) can draw based on information about combined holdings. For example, the RC might know that a stop showing either nothing in the suit, or AK actually means the latter, but while alerting, it should be explained as "Either AK or nothing in the suit". Note that this isn't very different from a RKC response that shows say 0/3; in most cases the asker will know the actual holding, but it's still alerted as 0/3.

[enigmisto]

Example from the book, describing the following hand:
S: AQ764 H: 93 D: 72 C: KJ95

This hand has just bid 3S showing 3 control points and is now going to begin denial cue bidding in response to 4C.

Ordinary scan of suits would be spades, then clubs, then hearts (skipping our shortest suit - diamonds - for the AK scan).
The first denial cue bid with this hand is 4NT.
The scan of Spades is Yes (A)
The scan of Clubs is Yes (K)
But now, we take a shortcut and skip Hearts in our AK scan proceeding directly to Queens, because we assume our partner can deduce we've got the A of Spades and K of Clubs which equals all our control points. So the next step is a scan of spades for Queen (Yes), and so we bid 4NT to show a "No" for the scan of Queen of Clubs.

Our partner knows we skipped over hearts and the "No" represents a no for Queen of Clubs. But this is not public information. If our hand held the K of Spades and K of Clubs, our scan would have needed to include the hearts, so the 4NT bid would represent a denial of a completely different suit.

How does one alert this, when the actual suits scanned are dependent upon holdings?

The following truthful statement is way too complex: "This bid means A or K of Spades, A or K of Clubs, A or K of Hearts if not all the controls have been shown yet, otherwise Q of Spades. The final step means no Queen of Spades or no Queen of Clubs depending upon whether spades and clubs held both the A and K."

[foobar]

 enigmisto, on 2022-January-28, 16:13, said:

Example from the book, describing the following hand:
S: AQ764 H: 93 D: 72 C: KJ95

This hand has just bid 3S showing 3 control points and is now going to begin denial cue bidding in response to 4C.

Ordinary scan of suits would be spades, then clubs, then hearts (skipping our shortest suit - diamonds - for the AK scan).
The first denial cue bid with this hand is 4NT.
The scan of Spades is Yes (A)
The scan of Clubs is Yes (K)
But now, we take a shortcut and skip Hearts in our AK scan proceeding directly to Queens, because we assume our partner can deduce we've got the A of Spades and K of Clubs which equals all our control points. So the next step is a scan of spades for Queen (Yes), and so we bid 4NT to show a "No" for the scan of Queen of Clubs.

Our partner knows we skipped over hearts and the "No" represents a no for Queen of Clubs. But this is not public information. If our hand held the K of Spades and K of Clubs, our scan would have needed to include the hearts, so the 4NT bid would represent a denial of a completely different suit.

How does one alert this, when the actual suits scanned are dependent upon holdings?

The following truthful statement is way too complex: "This bid means A or K of Spades, A or K of Clubs, A or K of Hearts if not all the controls have been shown yet, otherwise Q of Spades. The final step means no Queen of Spades or no Queen of Clubs depending upon whether spades and clubs held both the A and K."

Seems like paraphrasing it as "A/K ♠, A/K ♣, A/K ♥ OR (Q♠ and no Q♣)" is a complete disclosure of the methods?

[enigmisto]

 foobar, on 2022-January-28, 17:28, said:

Seems like paraphrasing it as "A/K ♠, A/K ♣, A/K ♥ OR (Q♠ and no Q♣)" is a complete disclosure of the methods?

Probably more accurate in this particular case would be "A/K ♠, A/K ♣, ((A/K ♥ and no Q♠) or (no AK ♥, Q♠, and no Q♣)". It gets messier the more steps.

And if you have only two control points, there are three possible junctures at which you might skip ahead to scanning for Queens, depending upon whether your control points are in your first one, two, or three suits scanned. So with two control points, every step could be representing three possible scenarios. It's hard enough to think through the actual logic of the specific cards you are facing with your hand and your partner's, let alone try to describe all the possible scenarios to your opponents with other potential hands. How does anyone communicate this reasonably in real games when a bid could represent denial in three potential suits, each with their own history of what was affirmed by the intervening steps?

[foobar]

 enigmisto, on 2022-January-28, 18:04, said:

Probably more accurate in this particular case would be "A/K ♠, A/K ♣, ((A/K ♥ and no Q♠) or (no AK ♥, Q♠, and no Q♣)". It gets messier the more steps.

And if you have only two control points, there are three possible junctures at which you might skip ahead to scanning for Queens, depending upon whether your control points are in your first one, two, or three suits scanned. So with two control points, every step could be representing three possible scenarios. It's hard enough to think through the actual logic of the specific cards you are facing with your hand and your partner's, let alone try to describe all the possible scenarios to your opponents with other potential hands. How does anyone communicate this reasonably in real games when a bid could represent denial in three potential suits, each with their own history of what was affirmed by the intervening steps?

Frankly, in all these years of playing DCB, I can't recall any hand that caused too much consternation regarding disclosure . That said, my recommendation is to post this specific hand and the question to bridgewinners (use their hand editor tool so that it's easy to see the bidding, etc.). Kit is very good at replying to questions related to KK-relay.

[mycroft]

Sure it's complicated (esp. online in a small text box), and sure with 5 or 6 controls instead of 3, it's fuzzier yet, but it's not "depends on holdings".

Partner either has the Christmas hand and no spade Q, or AK offsuit in the blacks and the spade Q, but not the club Q. "If I can't figure out which it is, I shouldn't be looking for slam."

[pilun]

I'm too stingy to buy the book so most of my knowledge of KK is second hand, probably wrong.
Anyway, I THOUGHT that the 3rd suit was skipped first time when the 4th suit was singleton, but not with 4432, 5422, 4333, etc.
(Skipping with 5440 or 5431 seems strange too)

Anyway, I'm surprised that Kit reckons its usually of no value to split AK from KKK. If opener has

♠Qx ♥AQxxxx ♦AKQ ♣Ax,

he would want to know whether he is facing

♠Axxxx ♥xx ♦xx ♣KQxx, or

♠Kxxxx ♥Kx ♦xx ♣Kxxx

Of course Kit is a much better player than me and his book is much longer than ours, but we have have been playing symmetric much longer than he has.

[enigmisto]

 pilun, on 2022-January-29, 01:46, said:

Anyway, I'm surprised that Kit reckons its usually of no value to split AK from KKK. If opener has

♠Qx ♥AQxxxx ♦AKQ ♣Ax,

he would want to know whether he is facing

♠Axxxx ♥xx ♦xx ♣KQxx, or

♠Kxxxx ♥Kx ♦xx ♣Kxxx

His scheme does resolve this question. Usually you scan three suits for AK with 5-4-2-2, but once you've shown all your controls, you assume your partner knows it and skip ahead to scanning for Queens.

So, with the first hand:
Step 1 - Yes, I have Ace spades.
Step 2 - Yes, I have King clubs. [We've shown our controls, so skip hearts and start scanning for Queens].
Step 3 - No, I am missing Queen spades [so you bid this step].

With the second hand:
Step 1 - Yes, I have King spades.
Step 2 - Yes, I have King clubs.
Step 3 - Yes, I have King hearts.
Step 4 - No, I am missing Queen spades [so you bid this step].

Partner can infer in the second case that you have three kings because if you had A and K in spades and clubs, your third step would be a scan for Queen of spades, and he can see there's no way you can answer Yes to that question (since partner has the Queen of Spades). So if it takes you four steps to get to a "No", you must have scanned all three suits, therefore you must have three kings.

Kit's claim is that you can nearly always make these sorts of inferences, which is why you skip to scanning for Queens as soon as you've communicated your Aces and Kings, assuming your partner can work it out.

But these inferences are subtle, which is why I am wondering how you communicate this to the opposition. You can likely work out whether they scanned two or three suits before skipping to scanning for Queens (after hearing the bid, not before), but this is not knowledge available to the opposition.

[straube]

Anyone have a link to Kit's book (for purchase)?

Curious what Adam and Sieong think of KK relay.

[DavidKok]

It's available in the Bridgewinners store, https://bridgewinner...book/kk-relay/. It might also be for sale in other webshops, I haven't checked.

[foobar]

 straube, on 2022-January-29, 11:05, said:

Anyone have a link to Kit's book (for purchase)?

Curious what Adam and Sieong think of KK relay.

I would have mailed you my copy, except that going to the post office is such a chore these days .

Regarding the book itself, it's pretty much a vanilla symmetric relay response scheme over 1♣ for the most part, i.e., the rest of the system is left to the reader. It does talk about a 1M - 2♣ GFR that's a little too weird for my taste (resolves shape first), and ends up at symmetric+2. Per Kit, this is better for mnemonics; see BridgeWinnners discussion

The DCB uses controls instead of QPs, and then scans Qs.

[foobar]

 enigmisto, on 2022-January-29, 04:48, said:

But these inferences are subtle, which is why I am wondering how you communicate this to the opposition. You can likely work out whether they scanned two or three suits before skipping to scanning for Queens (after hearing the bid, not before), but this is not knowledge available to the opposition.

Couple of comments:

1) It might be just easier to type in the inferred holding, instead of tying oneself into knots over the various combinations, i.e., say divulging an inference about a Q is unlikely to matter
2) The KK-relay DCB isn't the only game in town. There are plenty of "Queen Points" based schemes that use A=3, K=2, Q=1 that might be better than just control based schemes. Naturally, there are tradeoffs involved, but choices include Adam's PCB (parity based cue) that scans suits only once. The SCAMP DCB scheme is very effective too (see pilun's posts).

[awm]

 straube, on 2022-January-29, 11:05, said:

Anyone have a link to Kit's book (for purchase)?

Curious what Adam and Sieong think of KK relay.

I'm not in the habit of buying random system books so I haven't read it.

Nothing I've seen from the discussion here or on Bridgewinners looks particularly different or better than what we played over 1♣ in TOSR in the early 2000s though. This set of relays is "not terrible" but we obviously think we can do a lot better (thus IMPrecision). Probably the biggest issue in this style is the inability to judge additional queens; there are many opening 1♣ hands opposite which Axxxx Axxx xxx x should play in 4M while AQxxx AQxx xxx x is a very good 6M, and if you can't distinguish these hands below the five-level you are not doing very well. They have the same number of controls of course.

[foobar]

 awm, on 2022-January-29, 13:27, said:

I'm not in the habit of buying random system books so I haven't read it.

Nothing I've seen from the discussion here or on Bridgewinners looks particularly different or better than what we played over 1♣ in TOSR in the early 2000s though. This set of relays is "not terrible" but we obviously think we can do a lot better (thus IMPrecision). Probably the biggest issue in this style is the inability to judge additional queens; there are many opening 1♣ hands opposite which Axxxx Axxx xxx x should play in 4M while AQxxx AQxx xxx x is a very good 6M, and if you can't distinguish these hands below the five-level you are not doing very well. They have the same number of controls of course.

Actually, they don't even the TOSR 1♠ multi-bids for balanced / reds etc. so, it's pretty much a transfer into ♠ / minors scheme, with 2♦+ for the balanced hands.

enigmisto,

Since you are starting anew, a strong +1 for QP based schemes for the reasons Adam outlined above. I can't speak for why Kit chose a control based scheme, but believe that a vast majority of symmetric players on this forum will vote for QP based schemes. Another possibility is that after shape resolution, relay can ask for QPs and relay+1 can ask for controls (à la SCAMP).

[sieong]

Most denial cue-bidding methods that I know of involves some kind of assumptions regarding relayer being able to tell apart holdings that responder may have (a classic one is AKQ vs none). When Adam and I were playing TOSR in the 2000's, we used the version documented in http://www.bantha.or...hk/bridge/tosr/, for which the primary post-shape relay is based on controls and it makes a number of assumptions as well. My experience is that it was reasonably efficient, but not as efficient as QP + PCB. I believe that methods based on controls followed by DCB tend to rely more heavily on assumptions as otherwise it would be too difficult to get to the critical queens before the auction goes past 5♦ (by which stage most of the time we cannot stop below slams).

I have no experience playing KK relay, but from the posts that Kit wrote on bridge winners, I found that for many of the hands he showed, a small change of the relayer hand would have resulted in difficult guesses. All methods (as long as it uses assumptions) will result in some guesses some of the time (and without assumptions it may go too high too often). My subjective belief is that the assumptions made in KK relay were on the overly-loose side, and I think this is a question that could be objectively evaluated through simulation.

Taking the example Nick wrote, say if opener hand was instead:

♠x ♥AQxxxx ♦AKQ ♣Axx

and responder's possible hands are:

♠AQxxx ♥xx ♦xx ♣KQxx vs
♠Kxxxx ♥Kx ♦xx ♣KQxx

I think opener would end up with a guess under KK relay while most other methods I know of would be able to tell them apart. As a side note, I think most methods would have a hard time separating ♠Axxxx ♥Jx ♦xx ♣KQxx from ♠Axxxx ♥xx ♦xx ♣KQxx so it is really down to the fractions of auctions that the methods could get right.

[DinDIP]

I reviewed the KK book for The Bridge World. It's an excellent guide for relay novices in how to learn a system and how to use it. The hundreds of examples show what to think about when, especially how to envisage a possible slam and how to investigate it, using Kit's and Kate's methods.

It's not flawless -- what book is? I was irked by a number of example auctions where asker's bidding was DD, especially when searching for necessary queens. For a few of the deals I found the claim that there was safety at the five level (so that relaying further was right) to be dubious -- I'm not convinced it's a sound strategy to reach five-level contracts that are 50% at best in search of slam.

I noted in the review that the decision to use controls rather than QP was a surprising one and that it was a pity K&K didn't explain their choice. I did ask Kit separately but he didn't respond.

 sieong, on 2022-January-29, 16:41, said:

My subjective belief is that the assumptions made in KK relay were on the overly-loose side, and I think this is a question that could be objectively evaluated through simulation.

It would be good to have a tool that allowed such analyses, even if it would reduce some of the subjective discussion about which method(s) is (are) better.

IIRC, Michael Shuster reported on BridgeWinners that he and Sam Dinkin had bid hundreds of deals comparing the KK DCB method where n-1 non-singleton suits are scanned with their own where all n non-singleton suits are scanned and found the results to be virtually identical overall: there were wins and losses for both methods. They now use:

• Scan answerer’s longest suit first, tie to highest (twice if 6 or longer) then next longest suit and so on.
  
• Scan doubletons only once, singletons and voids not at all.
  
• For 9+ minimum responder with 3+CTRLs, do not scan queens until the Nth scan where N is the number of control cards (note well, control cards not control points).
  
• If you have super 1st, say no. Super 1st is AKQ for positives in a 3+ suit, or AK, AQ or KQ in a known doubleton.
  
• For semi-positives and hands with 0-2 controls, scan queens immediately and super 1st is also KQJ or AQJ.
  
• Don’t super first with double negatives.
  
• Scan queens immediately for double negatives.

Exception to Nth scan queens rule:

• if you have denied both the A and the K on the first scan, proceed to queens in that suit
  
• If you have shown AKQ or denied having any of AKQ, jacks ok the second scan in that suit

It's worth noting that they (for reasons I don't fully understand) use a shape-showing structure that is mostly symmetric+2 (+1 with minor one-suiters) so they have much less room for DCB than most symmetric pairs.

[Cascade]

 DinDIP, on 2022-January-29, 22:05, said:

I'm not convinced it's a sound strategy to reach five-level contracts that are 50% at best in search of slam.

The odds are surprisingly good.

Your scenario might be a little unusual - that the five level is 50% (your most optimistic view) but if the right cards are found we have slam.

One needs to do the numbers for each individual case. Here if we assume that slam is making (100% - unrealistic I know) with probability p and we will bid it, and the five level is 50% with probability 1-p, that is we either make five or go down one. Then we can calculate the IMP expectation versus the conservative route of stopping in four (assuming a major game/slam).

Vulnerable with probability p we score 1430 v 680 for +13 IMPs and with probability 1-p we score 650 v 650 half of the time and -100 v 620 for -12 IMPs half of the time.

13p - 12(1-p)/2 = 0 is our breakeven point. 19p=6 or p=6/19 which is a bit less than 1/3.

Not vulnerable with probability p we score 980 v 480 for +11 IMPs and with probability 1-p we score 450 v 450 half of the time and -50 v 420 for -10 IMPs half of the time.

11p - 10(1-p)/2 = 0 implies p = 5/16 which is again a little less than 1/3.

So it is worth the risk if the chance of slam is around 1/3 or better in this particular scenario.

[Cascade]

 enigmisto, on 2022-January-28, 05:32, said:

So what do you actually type into the alert box on BBO? How do other partnerships deal with that? Do they complain to the director?

Why not just list the possibilities.

Say partner has shown three controls (A=2 and K=1) and then goes four steps in the first denial cue-bid. Say for this example partner is 4=3=3=3. Just explain ♠A ♥K ♠Q and no ♥Q or ♠K ♥A ♠Q and no ♥Q or ♠K ♥K ♦K and no ♠Q. On the second round if partner goes two steps then the first two options are the same ♦Q and no ♣Q and the third option above leads to ♥Q and no ♦Q.

"sAQ hK no hQ, sKQ hA no hQ, or sK hK dK no sQ" isn't too onerous to type.

On the second round it is a little more complex: "With sA or hA then dQ no cQ OR with no ace hQ no dQ".