[benlessard]

Quote

e.g. After p bid 3♥ showing 5-4-1-3, don't you want to ask for ♣ keycards with KX KQX AXX AXXXX? But what if p bids 5♦…

Quote

No, with such hand RKC is foolish. I ask for AKQ-points and let partner show his entire hand...

Yes you do and especially when you assume that we don't read your post.

Anyway its not really important, what is important is finding out why and when RKC and scan is better than QP & DCB and vice-versa.

A couple of years ago I did a couple of hands on both DCB and RKC and it was clear to me that RKC and scanning was superior. Maybe the sample was too small or some other important DCB gadget that i didnt know were invented. I know some now play DCB with parity, this is something that I know little about and didnt test. Anyway I see no reason why we dont compared these methods once for all. We just need to bid 50 random slammish hands to see wich "scanning" method is best and most importantly why. This will trump any discussion that we have here.

This is a hand that was posted on BW not so long ago.

AQJxx
A
xxx
Axxx

Kxxxx
Qxxxx
AKQ
void

1S-1Nt
2D-2Nt (H,art GF)
3C-3D (low short,ask)
4C-4S (5530) (RKC S) 4D would be PES
5K-6C (2noQ) (bypassed the K of H and ask for K of D)
6NT-7S (Kd,QH,QD no J of S)

Here there is 2 way to ask for the K of D, ask for K of H or bypassing K of H, we dont have any special rules here but it can be improved. Also 5NT is always PAS sure we can but some rules where 5NT is asking instead but we keep it simple. Both M & S+C are the worse hands to relay because the GF start at 2NT.

This is a very easy hand for QP because you can keycard at +1 with no cost here. I undestand that many will make a Jacoby 2nt instead or relaying also.

[hrothgar]

benlessard, on 2013-July-22, 17:27, said:

A couple of years ago I did a couple of hands on both DCB and RKC and it was clear to me that RKC and scanning was superior. Maybe the sample was too small or some other important DCB gadget that i didnt know were invented. I know some now play DCB with parity, this is something that I know little about and didnt test. Anyway I see no reason why we dont compared these methods once for all. We just need to bid 50 random slammish hands to see wich "scanning" method is best and most importantly why. This will trump any discussion that we have here.

Your sample size was two? Am I safe in assuming that this is some kind of pathetic attempt at humor?
Even 50 is pretty small. (I suspect that you'd want to simulate hundreds of hands opposite any "known" strong club opener.

FWIW, I've always thought that developing some kind of test bed for auction termination mechanisms would have some real value.
This is a topic that seems amenable to brute force methods...

[benlessard]

Iv used "couple" (its an french-canadian expression that I mistranslated) but it should be "some". As for sample size it depend on how quickly it diverged. I guess if you do 50 hands and its like 30-20 in favor of your method and you have a pretty good idea why its like that than your side probably dont need to do more than knowing that your method is at the very worse slighty inferior and that you dont need to change it. Anyway I dont think its too tough to check more hands if necessary.

We should start by checking with limited hands first because this is what most relayer play. I can adjust my range to match Moscito with no difficulty.

For me I know my gains will be mostly 64?? pattern or pattern ending at 3H, where the keycard at 3S in the primary suit will be a lot better than the one at 4C. Ill also do well on the single suiters. Ill do worse on the 4M+5m hands or when we bid 4NT quant and we have a lot of jacks and not enough QP.

A hand ive found http://www.bridgebas...-relay-systems/

Zel continuation is really mysterious but if you play RKC scan this hand is really easy.

QTx
AQJx
Jxx
AQx

---
Kxx
AKQxx
KTxxx

ill even be in 7H here.

3H showed 0355---3S (C kc)
4D (2noq) --- 4H (K of D ?)
5D (Kd+KH+QD,no QH)---5H (J of club ?)
5S (no)--

Here I would bid 7H.

5-1 or 6-0 H is 16% while Jxxx-x or Jxxxx-void & void-Jxxxx is 15%.

But without the T of clubs 7H is 10% better than 7D (unless black suit squeeze or first round ruff)

[akhare]

hrothgar, on 2013-July-22, 17:39, said:

Your sample size was two? Am I safe in assuming that this is some kind of pathetic attempt at humor?
Even 50 is pretty small. (I suspect that you'd want to simulate hundreds of hands opposite any "known" strong club opener.

FWIW, I've always thought that developing some kind of test bed for auction termination mechanisms would have some real value.
This is a topic that seems amenable to brute force methods...

Richard, good to see you posting around these parts again.

What do you have so say about the original question about whether the RKC ask should be based on the relative lengths of suits (instead of assigning fixed slots for specific suits)?

My conjecture is that the relay captain is likely to want to ask about the first (or second longest suit), but I don't have any data to back it up. One option might be deal hands with say a combined total of say 18 - 21 QPs (the slave hand has a one or two suited distribution). We can then compute the QPs in the two longest suits and the DD tricks and examine the hands that do make slam.

Comments or suggestions?

[Free]

Ben,

what do you mean with "RKC with scan"? How do you continue after your RKC response?

After RKC I've always played Suit Asking Bids, which basically asks for top honors in a specific suit (the suit is determined from long to short, again with the order H-S-C-D when they're equal). As a result of this, RKC wastes tons of space, but gives us the opportunity to ask very specific questions. Combined with the general approach QP ask offers, it's a very good package.

Some important remarks about QP ask:
- when you have an inefficient relay structure (= shows hand pattern quite high) then QP ask is less efficient. Losing 1 step in the relays can result in the loss of 2 steps for slam auctions. When 3♦ shows pattern, then 3♥=QP ask ; 3♠=RKC 1 ; 4♣=RKC 2 ; 4♥=RKC 3. When you change that to 3♥, then 3♠=QP ask (1 step higher) ; 4♣=RKC 1 (2 steps lost) ; 4♥ = RKC 2 (2 steps lost) ; 4♠=RKC 3 (1 step lost).
- when you're scanning a balanced hand you can't always get an accurate picture of the hand quickly because you're scanning 4 suits. Scanning 3 suits is usually no problem.
- when you need lots of QP's from partner, sometimes it's better to signoff and hope partner can zoom. When 3♠ is QP ask with base level 6, then 3NT shows 6, 4♣ shows 7, etc. However, when you need 9 or more and have no fit, you can bid 3NT after which partner will zoom to 4♣ showing 9 (gain 2 steps), 4♦ showing 10, etc.

In my experience QP ask with denial cuebids is extremely efficient, that's why most relay systems put it on step 1, immediately after showing exact distribution. The alternative Control ask with denial cuebids is also very good. Sometimes it's even better, sometimes it's a bit worse, depends a lot on the hands and the amount of Queens.

RKC obviously becomes less efficient since we have to start higher. But even if you could start quite low, you waste tons of space to show 1 thing at a time (possibly an entire level for keycards, possibly an entire level for suit 1,...). Hence my opinion that RKC is only useful when you need to check keycards and the top honors in 1 side suit.

Whatever method you want to use though, you need to prepare and calculate if you're going to have enough bidding space to ask everything you need to know.

[hrothgar]

benlessard, on 2013-July-23, 00:04, said:

As for sample size it depend on how quickly it diverged. I guess if you do 50 hands and its like 30-20 in favor of your method and you have a pretty good idea why its like that than your side probably dont need to do more than knowing that your method is at the very worse slighty inferior and that you dont need to change it. Anyway I dont think its too tough to check more hands if necessary.

FWIW, here's how I would start to examine this question

1. Start by assuming a specific hand for the relay asker, a specific shape for the relay responder, and sufficient strength such that its worth investigating slam.
2. Generate 50-100 hands consistent with these conditions.
3. Have a two pair of players bid out the hands starting with shape resolution. Evaluate whether one auction is more successful that the other.

Once you're (broadly) happy with your results, change the relay asker's hand and see whether this changes your results in any significant fashion.
Repeat this a few times.

Once you're done with this, you can start varying relay responder's shape...

[hrothgar]

akhare, on 2013-July-23, 01:17, said:

Richard, good to see you posting around these parts again.

What do you have so say about the original question about whether the RKC ask should be based on the relative lengths of suits (instead of assigning fixed slots for specific suits)?

My conjecture is that the relay captain is likely to want to ask about the first (or second longest suit), but I don't have any data to back it up. One option might be deal hands with say a combined total of say 18 - 21 QPs (the slave hand has a one or two suited distribution). We can then compute the QPs in the two longest suits and the DD tricks and examine the hands that do make slam.

Comments or suggestions?

The RKC ordering is based on the assumption that you're most interested in Keycard information about responder's longest suit.

FWIW, I think that its pretty easy to demonstrate that QP ask + denial cues is going to work out to be significantly more efficient that RKCB.
As Free notes, your RKCB ask needs to be tied to a specific suit. This means that you need to allocate 4 separate asking bids.

Given that you're using RKCB as your primary asking mechanism, there's no way to intelligently arrange the order of the asking bid. (You're equally likely to want to ask in opener's long suit as responder's long suit). This means that on average, your RKCB ask is going to be 2.5 steps above shape resolution.

If you're typically resolving shape at 3♥+, this is deadly.
You lose one step for 3NT and a second for 4D, so on average, your initial ask is going to be a 4♥
In contrast, the QP ask+ DCB is able to use 3♠ for its first ask.

Even if RKCB is giving you much better information - and I don't believe that it is - it seems highly improbably that the methods can compensate for

1. Forcing you above 4M on a significant number of hands
2. Losing an entire level of bidding space

[straube]

hrothgar, on 2013-July-23, 05:05, said:

Given that you're using RKCB as your primary asking mechanism, there's no way to intelligently arrange the order of the asking bid. (You're equally likely to want to ask in opener's long suit as responder's long suit).

RKCB is ancillary. We're using S1 as QP ask and we follow this with Adam's Parity Cue Bidding. I think between having this and having the extra room that majors affords, that we do well to have a tool for exploring small minor suit slams...which is why we've assigned S2 as RKCB clubs and S3 as RKCB diamonds and so on. I've found that QP ask with PCB doesn't usually let us resolve which high cards pard has until somewhere around 5D or 5H on average...which is too high to avoid endplaying ourselves into 6m.

[akhare]

hrothgar, on 2013-July-23, 05:05, said:

The RKC ordering is based on the assumption that you're most interested in Keycard information about responder's longest suit.

Richard,

As straube noted, we use S1 as DCB and its efficacy isn't in question. The question here that whether the subsequent steps that are reserved for RKC should be fixed (in the ♣, ♦, ♥, ♠ order) or whether they should be dynamically based on length of responder's suits (a la Moscito).

My conjecture is that the latter approach is superior.

[hrothgar]

akhare, on 2013-July-23, 10:46, said:

Richard,

As straube noted, we use S1 as DCB and its efficacy isn't in question.

Sorry, I thought that Ben was asserting that RKCB was better than slam points and DCB...

From what I can tell, Straube seems to be asserting that RKCB is most useful when you hold clubs or diamonds.
I'm not sure that I buy this, but it seems like the sort of thing that one could test...

If it turns out to be true that RKCB hands are very much biased towards the minors than the extra step between 6C and 6D might be more important than suit length.
I don't know what the right answer its, but I'm unwilling to discount the idea without method information...

[benlessard]

Ive learn spiral scan quite long ago and its the only method ive ever seriously played.
http://www.bridge-wi...vent/SpScan.htm
http://www.bridgeguy...SpiralScan.html

Its a easier than QP & DCB but you still need concentration because the mistakes are costly.

The way Ive learned QP is 321 & DCB stop is 03 and bypass is 12. Second round is 0 -1 even if the next suit is a possible SOFF Ive play this as a relay therefore if youre in 4D for better or worse you cannot soff in 4H (unless having 0-1 spades where 4S is relay and 4H to play) I know some play 4Nt for ask when both M are possible) there is probably a lot of other gadget & agreements i dont know.


I believe that "scanning" for card is complex, important and likely to have a "mathematical based" solution.

Every slammish hand that I read I automatically check how we would have bid it and my opinion is that spiral scan do great and better than QP-DCB most of the times. The 2 main problems I have with Sscan are when you you are missing a keycard and you dont know if its the K of trumps, lacking a side K and the K of trump is often a 75% slam (same for lacking KQ of trumps). Imo considering the K of trumps as a keycard is possibly a wrong method and RKC Aces- ask for KQ of trumps- scanning side K is something I would really like to test.

The other drawback is a stiff A when your lacking an ace.

KQx
KQJx
Ax
xxxxx

opener showned a

5341 with 2 keycards and the KD. Here opener is going to have the stiff club Ace like 30% often leading to a very poor slam while if your missing the club ace and the slam is great. Of course when your not missing an A its irrelevant but opener cannot know that your not misssing an ace, in short many of our bad slams are because of a stiff Ace.

Quote

I think between having this and having the extra room that majors affords, that we do well to have a tool for exploring small minor suit slams...which is why we've assigned S2 as RKCB clubs and S3 as RKCB diamonds and so on. I've found that QP ask with PCB doesn't usually let us resolve which high cards pard has until somewhere around 5D or 5H on average...which is too high to avoid endplaying ourselves into 6m.

Agree 100%. I also would like to test is m should come before M the cost is that its going to be tougher to stop in 4M but its easier to stop at 5m vs 5M. Hand like 5044 for example maybe the cost of sometimes stopping at 5S instead of 4S is smaller than the benefit of being significantly better placed for 5m or 6m, its really something that needed to be tested.

Quote

FWIW, I think that its pretty easy to demonstrate that QP ask + denial cues is going to work out to be significantly more efficient that RKCB.

could you provide some random examples ? Im willing to bid randoms hands against any1 who think DCB is so much better that Sscan. We just have to write down or post links to our methods and ask some1 to post some randoms hands.

[hrothgar]

benlessard, on 2013-July-23, 13:22, said:

Im willing to bid randoms hands against any1 who think DCB is so much better that Sscan. We just have to write down or post links to our methods and ask some1 to post some randoms hands.

Before we go through the trouble of bidding hundreds of hands, please address the core of my argument.

Assume that shape resolution terminates with a 3♥ bid.
A system that uses RKCB to explore slam will bid

3♠ ~25% of the time (+1 steps)
4♣ ~ 25% of the time (+3 steps)
4♥ ~25% of the time (+5 steps)
4♠ ~25% of the time (+6 steps)

On average, you are going to be bidding +3.75 steps

In contrast, a system based on Denial Cue Bids will be bidding +1 step the vast majority of the time.
On average are throwing away close to three steps in bidding space compared to DCB and forcing yourself past 4M much of the time that partner holds a minimum hand.

Admittedly, things a re a bit better if shape is resolved with 3♦

Here you bid

3♥ 25% of the time (+1 step)
3♠ 25% of the time (+2 steps)
4♣ 25% of the time (+4 steps)
4♥ 25% of the (+6 steps)

Here you only average +3.25 steps and half the time your initial ask is below 3NT. Even so, I think that you're sacrificing way too much bidding space.

[awm]

Say I hold KQxx Axx AKx Qxx and open strong club. Partner shows a game force 5431 by 3D. If I'm playing RKC then next is 3H as keycard in spades. Partner bids 3S (one). Slam is okay if partner has heart KQ and great if he has that plus diamond queen. I can skip SQ but I need HK so my next spiral bid is 4C. Partner bids 4H showing the hearr king but denying DK. Now I still do not know either red queen and must chance the five level. But partner could have AJxxx Kxxx xxx x and even 5S is poor. This despite the fact that I got lucky and my trump suit of choice was the cheapest ask!

Playing QP this hand is easy. You need 6 QP for slam to have a shot, and if partner has five or less you easily get out in 4S. If partner has 6 QP the five level is safe, and you can find out which queen he has before deciding on slam.

[benlessard]

When its both m or S+D, 2H will be GF relay the following 2NT is 22(45), 3C is (31)(54), 3D is (21)55, 3H is (21)64. So for those cases

holding a 21(64) -- bidding 3S(keycard in the 6) is like 75-80% bidding 4C(keyc in the 4) is like 15%-20 and bidding 4D(keycard in the 2 card suit) is less than 5%. This is quick estimation, so its another thing that really need checking.

Holding a (21)55 I think bidding 3H is 50% bidding 3S is like 45% bidding 4C is like 5%.

holding a 5431 is probably something like 45-30-25%. But of course 5431 are a lot more frequent than any other hand type.

--------------------------------------------------------------------

Holding H+C GF relay will be at 2D so its even cheaper and we rarely have problems here.

------------------------------------------------------------------------------

Holding H+D GF relay is 2S so 3C = 2(54)2 , 3D = ?(54)1or3 . etc so your not going to keycard at 4m taht often.

----------------------------------------------------------------------------------------------------------------
IMO the critical cases for us is really both M or S+C where the GF relay start a 2Nt and 3D is (5422), 3H is (5431) these cases its likely that QP and DCB work better.

Anyway ive played relay system long enough to know how complex salm scanning can be, I believe any1 with my expereince understand that comparaison testing is probably the only way to know what work best and what doesnt work so well.


Over a strong club opening we tend to relay 1 step lower so its also rare that we keycard over 3Nt there.

[hrothgar]

benlessard, on 2013-July-23, 14:47, said:

When its both m or S+D, 2H will be GF relay the following 2NT is 22(45), 3C is (31)(54), 3D is (21)55, 3H is (21)64. So for those cases

holding a 21(64) -- bidding 3S(keycard in the 6) is like 75-80% bidding 4C(keyc in the 4) is like 15%-20 and bidding 4D(keycard in the 2 card suit) is less than 5%. This is quick estimation, so its another thing that really need checking.

Holding a (21)55 I think bidding 3H is 50% bidding 3S is like 45% bidding 4C is like 5%.

holding a 5431 is probably something like 45-30-25%. But of course 5431 are a lot more frequent than any other hand type.

--------------------------------------------------------------------

Holding H+C GF relay will be at 2D so its even cheaper and we rarely have problems here.

------------------------------------------------------------------------------

Holding H+D GF relay is 2S so 3C = 2(54)2 , 3D = ?(54)1or3 . etc so your not going to keycard at 4m taht often.

----------------------------------------------------------------------------------------------------------------
IMO the critical cases for us is really both M or S+C where the GF relay start a 2Nt and 3D is (5422), 3H is (5431) these cases its likely that QP and DCB work better.

Anyway ive played relay system long enough to know how complex salm scanning can be, I believe any1 with my expereince understand that comparaison testing is probably the only way to know what work best and what doesnt work so well.


Over a strong club opening we tend to relay 1 step lower so its also rare that we keycard over 3Nt there.

I thought that we were comparing auction termination mechanisms...
Why are you explaining your shape resolutions?
If you want to discuss shape resolution, please start a separate thread.
As a starter, please provide the resolution level for all shapes - not just some random collection - along with the conditional probability that you hold said hand opposite a strong club opening...

I made a very specific point: A RKCB based system is going to burn significantly more bidding space than a DCB system.

Your primary counter seems to be based on the asserting that your key card asks will typically be in relay responder's long suit, therefore things aren't quit so bad..
I suspect that you'll get some efficiencies through this, however, I doubt that this ends up being all that significant.

For every hand that RR has a long suit there's going to be another where the relay asker wants to set his long suit as trump.
Moreover, on those hands where the relay responder happens to balanced, you're pretty much shooting craps...

[awm]

My view here is that you fairly often need a quantitative ask. There are a lot of hands where you need describer to have "a bit more" than he could have, and it is fine if this takes the form of queens and jacks.

If you're playing RKC (or controls) as a primary method you don' t get that info until way late. You will basically need a separate "strength ask" which burns steps and/or relay breaks. The nice thing about QP is that you get this info automatically because QP is a good proxy for overall strength.

[benlessard]

Quote

3♠ ~25% of the time (+1 steps)
4♣ ~ 25% of the time (+3 steps)

Quote

Why are you explaining your shape resolutions?

Just to show that hands that end in 3H usually have a fair amount of shape meaning that keycard in the longest suit is more like 75-80% not 25%. 5431 than ends in 3C can keycard in all 3 suits at the 3 level so why would I need QP here ?

Quote

We should start by checking with limited hands first because this is what most relayers play. I can adjust my range to match Moscito with no difficulty.

Ive assumed limited hands that the pts range is known that is why I think QP can be a bit pointless. In a system where its 11-18 than of course QP or any other way to check for min or max make more sense.

In many other cases it depend on how your system is structured, for example we play transfers responses that are weak or GF so opener accept the transfer with 15-20 and jump raise with 20-22 and 4 card support so if responder is holding 8-10 he can just blast to game or lie about the number of keycard to signify no slam interest. That why in AWM example we wont be in 5S with minimum values and if slam is there we will probably bid it.

Another example
4351 and 9 pts.

1C--1H
1S--?? (1S is 15-20 with 3/4S) but with 20 pts and 4S OP will bid 2S.

here I know slam is too unlikely so im just showing my shape to play the best game and when partner RKC in S or H I do -1 with my keycard and we play game (in my system the best hand opener can have is 19 pts bal and 5H+3S or bal with 5D). With a 5341 and a 1C-1H-1S start I would just bid 4S over 1S to play and disclose information and avoid lead directing X even if ill be dummy. The corollary is that when we show 2 keycards or we show some shapes a slight slam interest is promised.

We do have some auction where pts are unlimited. 1C--2NT (2245) or 1C--3C (1345) where responder can be 9+ pts and opener is unlimited, I agree that in those case a QP solution is probably better. However the case that I would like to test and the most pertinent for my system is the 10-14 range with some shape for opener and hands with no voids nor self sufficient suit (relay break hand) for responder.

[hrothgar]

You seem to be including so many additional restrictions and constraints that the results aren't going to generalize
This will probably reduce other folks interest in your results.

One interesting starting point would be an analysis that examined what fraction of the time responder is showing a long suit, but the relay captain wants to set some other suit as trump...

[hrothgar]

♠ AQJxx
♥ A
♦ xxx
♣ Axxx

♠ Kxxxx
♥ Qxxxx
♦ AKQ
♣ void

FWIW, I am attaching the MOSCITO auction because its a pretty amusing one
(Lots of different options in bidding the hand, but all roads lead to 7S)

1♣ - 1♦ (With a 5-5 shape and 9 slam points South's hand is a bit strong for a limited opening)
(1♦ = art GF, denies certain hand types)
1♥ - 2♣ (1♥ = Relay: Even with a minimum, opener prefers not to RR with a 5530 shape)
(2♣ = SS with Clubs or 2 suited with Blacks)
2♥ - 2♠ (2♥ = Reverse Relay, 0-1 Clubs and 4+ Spades)
(2♠ = Relay)
2N - 3♣ (2N = Spades and Hearts, 3♣ = Relay)
3♦ - 3H (3♦ = 5/5 in the majors, 3♥ = relay
3N - 4♣ (3N = 5=5=3=0, 4♣ = QP ask)
4♦ - 4♠ (4♥ = 9 slam points, 4♠ = DCB)
5♦ - 7♠ (5♦ = 1-2 Controls in ♠ + ♥, 0/3 Controls in Diamonds

Alternatively, I could use RKCB in Spades, in which case I'd bid 4S rather than 4C

Here, the auction would continue

5D - 5H (5D = 2 Keycards no queen, 5H = CAB in Hearts)
5N - 6C (5N = AK or Q in Hearts, 6C = CAB in Diamonds)
6N - 7S (6N = AKQ in Diamonds)

[benlessard]

I think relaying a 10-14 pts opening hand is the most pertinent for everybody. What will work for 10-14 is likely to work for any limited range.

Ive started to scout for all unbal 10-15 on a slam or near slam hands on the WBF web page of the last tournament.

On the close hand (9 or 15 pts) we can switch a pts to the other side.

http://www.zgqpw.com...anza=0&board=19

http://www.zgqpw.com...tanza=0&board=5

http://www.hkcba.org...t/hand_r1_6.pdf
hand 26 & 31

http://www.hkcba.org...t/hand_r1_8.pdf
31

http://www.hkcba.org...t/hand_r1_9.pdf
8,
http://www.hkcba.org.../hand_r1_10.pdf
21, 26 is a relay break, 30 is close

http://www.hkcba.org.../hand_r1_11.pdf
11 is close

http://www.hkcba.org...t/hand_r2_1.pdf
22,24

http://www.hkcba.org...t/hand_r2_2.pdf
7,8

http://www.hkcba.org...t/hand_r2_4.pdf
2,9

http://www.hkcba.org...t/hand_r2_5.pdf
17 (preempt for me),18

I dont claim these will represent a perfect sample but I dont believe these are biased.