[Free]

akhare, on 2011-April-21, 17:23, said:

I find the first reverse relay very intriguing because it reflects my own opinion that they should be used to describe minimum hands with no slam interest. However, there are other schools of thought which maintain that it's better the balanced hand to do the asking (regardless of strength).

What is your general perpsective on the use of reverse relays?

This is my personal view and I haven't used this in practice.

Imo it might be more efficient if opener always asked (for example with 1♣-1♦-1♥) about responder's shape whenever he's balanced. And whenever opener is unbalanced, he could show his shape (for example by bidding 1♣-1♦-1♠+). Note that this basic change would considerably improve the relay system, because 1♣-1♦-1NT should no longer be balanced (so you win 1 step for most unbalanced shapes!). However, this complicates matters a bit, and you might give away too much information when it's not necessary (now you can have a fast arrival 1♣-1♦-1NT-3NT if that's not shape showing). Not to mention the amount of wrongsiding contracts.

Since opener usually has a minimum balanced hand, it's nice to rightside the contract by rebidding 1NT immediately. I think it would be better to use the normal (preferably complex) 1NT structure over this, instead of 2♣ as relay and 2♦+ as reverse relays. This way responder can still show his shape when he's unbalanced.

This might make things very efficient:
1♣-1♦-?
1NT = any balanced hand (any range)
...The same structure used after 1NT openings
Other = unbalanced, showing shape (you may use the normal symmetric scheme, perhaps with reversed 1M rebids)
...Responder asks or reverses relays with immediate misfit

The main advantage is that with balanced vs unbalanced hands, the balanced hand always asks and the unbalanced hand always shows. This makes light distributional slams much easier. With 2 unbalanced hands, opener starts to show but responder may reverse immediately after 1♣-1♦-1M. The relaystructures are more efficient for sure. And you can easily have a lot of fast arrival auctions like 1♣-1♦-1NT-3NT.

The biggest disadvantage is that 1♣-1♦-1NT is unlimited. Opener will have to zoom from time to time, and it's not always clear how to show various strengths (you see, I don't have experience with it ). Also, bidding balanced vs balanced hands loses a lot of accuracy. This also loses flexibility: with this approach opener no longer has a choice between showing or asking about shape.

All things considered I think this is a better approach.

[hrothgar]

As I mentioned before, reverse relays are one of the great failings in relay literature.
These don't get discussed at all (aside from in an abbreviated fashion)

I reason for this is pretty obvious.

• This is a complicated topic with conflicting design goals
  
• It hasn't been studied extensively.
  
• There aren't many good proof's, just rules of thumb that people have developed

Its probably most useful to start the discussion by listing a number of different design goals for reverse relays. I'll kick things off with the following list

• Natural bidding is better than relay bidding for determining strain in game level contracts. (Making an informed decision between 3N, a Moysian, or 5m requires information about stoppers, suit quality, and the like which shape oriented relay methods don't describe)
  
• Identify when you have a pair of misfitting two suited hands ASAP
  
• Balanced hand asks, unbalanced hand shows
  
• Being able to limit strength quickly is a good thing

Once you've compiled your list, you generally try to identify different stages in the relay process. You'll (hopefully) be able to match a specific design goal to a given stage. For example, assume that the auction starts

1♣ - 1♦ where 1♦ is an artifical GF

This is a great opportunity to use a reverse relay to limit the strength of the strong club opener. (Responder hasn't provided any shape information which rules out a number of options)

Alternatively, suppose that the auction starts

1♣ - 1♦
1♠

Where 1♠ is a reverse relay (showing a variety of different 1 and 2 suited hand types that contain 4+ Spades)

At this point in time, a relay break is (arguably) best used to show hands with Spade shortage, trying to quickly identify a misfit.

[Free]

I guess akhare was more asking about which design goals should get the highest priority. Imo "balanced hand asking" should be pretty high priority, limiting strength is the lowest priority imo since we can zoom afterwards without much trouble. This is in contrast to what we play ofcourse.

I'm not sure how high I'd place "identifying a misfit early".

Having a natural auction to game depends largely if you've both limited your strength. With 2 unlimited hands it causes problems from time to time if you don't immediately find a fit. I noticed that when starting a very basic precision system using unlimited natural GF responses with natural continuations (only frivolous 3NT could limit one's hand).

[wclass___]

Free, on 2011-April-22, 07:03, said:

I'm not sure how high I'd place "identifying a misfit early".

Quite a lot of IMPs can be gained if you can stop below game in misfit hands. As for immediate relay breaks this has the highest priority for me.

And as i know that i can stop low, i even slightly lowered requirements for some "GF" responses. (to me 1♣-1♦ is negative)

[Free]

wclass___, on 2011-April-22, 07:30, said:

Quite a lot of IMPs can be gained if you can stop below game in misfit hands. As for immediate relay breaks this has the highest priority for me.

And as i know that i can stop low, i even slightly lowered requirements for some "GF" responses. (to me 1♣-1♦ is negative)

Can you give an example auction of how you identify the misfit and how opener or responder limits his strength so that the GF auction gets passed out safely? I just wonder if it's really worth the trouble, and if it's possible to incorporate this in a 1♣-1♦ GF scheme.

[wclass___]

1♣-1♥ (=4-5♠ 0-3♥, ''GF'')
=>
...1♠ relay, really GF now.

...1N short ♠, minimum, NF.
...2♣ natural, misfitting, NF.
...2♦ natural, misfitting, NF.
...2♥ natural, misfitting, NF.
...2♠ 4333♠ minimum, NF
...2NT minimum balanced with 2♠, NF.
...etc.

Normally one wouldn't want 2♠;2N rebids as NF, but as i lowered requirements for ''GF'' they become important for me.

[straube]

Akhare and I are playing a structure called SCREAM which was developed by Mark Abrahams. We've previously experimented with the strong club portion of Moscito.

I would imagine that Moscito's 1D (hearts), 1H (spades) and 1S (diamonds) and 1N (12-14) are very strong in combination and I'd love to have 1D-2H auctions. Unfortunately, we're in ACBL-land where such things aren't permitted. Rightly or wrongly, we've used 1D as our weak NT and other assorted hands without a 5-cd major which gives us 1N as 14-16. Our club is then similar to Meckwell in strength...stronger than Moscito.

We ought then to be in a worse position than Moscito in our non-club auctions, but in a stronger position for our club auctions. A net loss no doubt.

I think Moscito gives preference to the semipositives because of the sheer frequency of them. I think that the breakdown is something like (and people are always underestimating the number of DN responses which are at least as follows)...

DN-20%
SP-50%
GF-30%

Faced with this, the designer of Moscito gave more relay preference to the semipositives, making them +1 (instead of +2) and GF as +1 (as opposed to +0). It's not ideal to give them parity. There are far more slams imo opposite a GF response than a SP response and that is where relays help the most. Not only that, but the SPs are limited to 3-5 QPs where the GF hands can be 6, 7, 8, 9, 10, etc. (and for us 5, 6, 7, 8, 9, etc) so they are not as well defined and could use more room. Also, Moscito has (had) such auctions as 1C-2S to show a SP single-suited spade suit. This is set up for relays at +1 but is something of a misfire when opener passes or signs off in a different suit.

Again, it's the high frequency of the SPs that lead to this design. Better to get suits shown on the more frequently occuring strength range was the thought. And yet, most of the SPs still are grouped in the 1H response.

With our stronger club, we have chosen to start GF auctions with as few as 5 QPs and made the SPs 2-4 QPs (with 2 being optional). My last tally was...

DN-19%
SP-42%
GF-39%

so we give less space to the SPs. True, they are slightly more common than the GF, but they are also less important in terms of game or slam. Also, we can relay them at +2 (1C-1H, 1S). In practice, I'd guess we only relay these hands out maybe 25% of the time; it takes opener having something like a 20 ct to do so. We have many 1C-1H, 1N auctions.

This means that our 1C-2S (for example) shows a GF hand with balanced single-suited diamonds. The sequence always leads somewhere as it seems it should.

Of course I would agree with Free about wanting a balanced hand to ask about an unbalanced hand. Even while playing Moscito, I strained to rebid 1H with a minimum balanced hand because it was so profitable when partner had an unbalanced hand that fit well. There is just no way to know with AJx xxxx KQxx AK whether I have slam interest or not until I know partner's pattern.

1C-1D, 1N as balanced just kills responder's ability to pattern out. Wouldn't we be suddenly at +3 compared to standard symmetric? 1C-1D, 1N and we'd have all of responder's patterns to sort out starting with 2C as opposed to 1H for standard symmetric. I agree with Free that if this sequence showed a balanced hand, that responder would desire to describe his shapely hand, but there just isn't room for it. Using a standard response structure (stayman and transfers)can never show the exact shape (by approximately 3N I mean) and creates a co-captaincy.

For Moscito players, why not use...

1C-1D GF
.....1H-bal or....
..........1S-bal or....
...............1N-bal
...............2C+ various unbal
..........1N+ various unbal
.....1S+ various unbalanced

You would have your choice of letting opener reverse relay (at +1 approximately) immediately or hiding certain unbalanced patterns with 1C-1D, 1H and then spinning them off after 1C-1D, 1H-1S...which would let you be +0 more often. You'll never get to +0 in total because you've assigned 1C-1N+ for mostly semipositives.

Btw, if 1C-1D, 1H-1N as an unbalanced GF hand were bad, then 1C-1N as an unbalanced SP hand ought to be worse. I think neither is attractive but is a necessary evil. We use 1C-1D, 1H-1N to show hearts and at least opener gets to declare heart contracts.

[hrothgar]

Let's consider relay breaks after 1♣ - 1♦
where 1♦ is an artifical game force

The way things work right now

After 1♣ - 1♦

1H = Relay
1S = Unbalanced hands with black suit
1N = Balanced or 3 suited
2C+ = Unbalanced hands with a red suit

If the strong club opener decides to reverse the relay and show rather than ask after 1♣ - 1♦

1S = Unbalanced hand with a black suit
1N = Balanced or 3 suited
2C+ = Unbalanced with a red suit

In general, the decision to ask rather than show indicates range. (The strong club opener is limiting his hand)

• If the strong club opener decides to show a balanced hand, he is very weak
  
• The strong club opener MUST ask with a 5440 pattern regardless of range

I think that this is an extremely elegant structure

You provide actionable information
You have complete symmetry regardless of whether the strong club opener asks or shows
You have complete symmetry with between the 1S response and the 2C+ responses
You have a very nice relay structure available after 1NT rebid (showing a balanced hand or a three suiter)

After an auction like

1C - 1D
1N

Responder can either

Ask with a 2♣ response OR show a minimum balanced hand by bidding 2D+

If responder shows, Opener is then well positioned to transition to natural bidding if appropriate...

[straube]

You emphasize very weak. Can you give me a maximum and a minimum hand for 1C-1D, 1N?

[akhare]

straube, on 2011-April-22, 11:54, said:

You emphasize very weak. Can you give me a maximum and a minimum hand for 1C-1D, 1N?

Don't want to steal Richard's thunder here, but I would imagine balanced hands in the 9-10(11) QP range.

Following 1♣ - 1♦ - 1N Responder still has 2♣ available to pattern out opener's hand using the standard balanced module. 2♦ / 2♥ can are the usual transfers and the remaining can be used to do things like size ask, stopper shows, etc.

The loss over the original 1C - 1D - 1N (reverse relay) is that opener can't reverse relay with the two/three suited suited with majors hand, but IMO, the tradeoff is well worth it.

[hrothgar]

straube, on 2011-April-22, 11:54, said:

You emphasize very weak. Can you give me a maximum and a minimum hand for 1C-1D, 1N?

For me, a representative maximum would be something like the following

Maximum
♠ KT94
♥ QJ8
♦ AQJ6
♣ QJ3

[straube]

hrothgar, on 2011-April-22, 13:11, said:

For me, a representative maximum would be something like the following

Maximum
♠ KT94
♥ QJ8
♦ AQJ6
♣ QJ3

Wow! An 8 QP hand with no opportunity for upgrading opposite shortness! We open that 1N 14-16 to begin with. We're at different starting points/considerations for sure.

Even when I played Moscito and had such a hand (stronger but not upgradeable) , I was happy enough to rebid 1N.

I think you're not getting full value out of your 1N rebid. Consider the opportunity cost. I also think you hobble partner when he has a distributional hand with extra values. If I understand it correctly, Moscito uses 1C-1D, 1N-2D+ for responder to show balanced hands and every distributional hand has to use the 2C relay and ask about opener's hand. That's a lot of work for partner to do! He has to find opener's pattern (much of which he will not care a whit), he has to ask for QPs and then place cards for all four suits.

[hrothgar]

straube, on 2011-April-22, 14:08, said:

Wow! An 8 QP hand with no opportunity for upgrading opposite shortness! We open that 1N 14-16 to begin with. We're at different starting points/considerations for sure.
Even when I played Moscito and had such a hand (stronger but not upgradeable) , I was happy to rebid 1N.

I think you're not getting full value out of your 1N rebid. Consider the opportunity cost. I also think you hobble partner when he has a distributional hand with extra values. If I understand it correctly, Moscito uses 1C-1D, 1N-2D+ for responder to show balanced hands and every distributional hand has to use the 2C relay and ask about opener's hand. That's a lot of work for partner to do! He has to find opener's pattern (much of which he will not care a whit), he has to ask for QPs and then place cards for all four suits.

Comment 1
My 1NT opening shows 11+ - 14 HCP, not 14-16

Comment 2

I am a firm believer balanced hands should ask rather than show.
I like that the auction

1♣ - 1♦
1N

is as rare as possible. I pretty much use this to show balanced hands with less than 9 QPs such that the relay asker won't need to try for slam that often.

I could be convinced that ALL balanced hands should ask rather than show (bid 1♥ over 1♦ with a balanced hand regardless of strength). However, in this case the 1NT relay break is (essentially) empty.

I'm welcome to suggestions what other bid should get slotted in here...

[straube]

hrothgar, on 2011-April-22, 14:19, said:

Comment 1
My 1NT opening shows 11+ - 14 HCP, not 14-16

Comment 2

I am a firm believer balanced hands should ask rather than show.
I like that the auction

1♣ - 1♦
1N

is as rare as possible. I pretty much use this to show balanced hands with less than 9 QPs such that the relay asker won't need to try for slam that often.

I could be convinced that ALL balanced hands should ask rather than show (bid 1♥ over 1♦ with a balanced hand regardless of strength). However, in this case the 1NT relay break is (essentially) empty.

I'm welcome to suggestions what other bid should get slotted in here...

Well, if you (as we do) have opener rebid 1H with all balanced hands, you have to accept that 1C-1D, 1N shows a hand with shortness and that this is not ideal.

OTOH, it can't be ideal for Moscito users to rebid 1H with all of their 5440s for want of room. Whenever partner has a balanced hand, you can't be pleased at not knowing whether his hand is working for you or not.

No, I wouldn't use 1N for 5440s because 1C-1D, 1N is much more valuable than just that. We use it for both majors or 3-suited with both majors (including some 5440s). I would think Moscito players would use it differently.

I think certainly 1C-1D, 1H-1S should show that responder has (likely) a balanced hand and then opener has two chances to reverse relay instead of the one chance he has now. This means that you will frequently save an important step.

If you like, I'll play with it and see but it would be helpful to know what your current 1C structure is.

[wclass___]

Can someone please explain me what is the reasoning why balanced hands should ask?

If one shows balanced hand, he limits himself really well shape wise. There are basically only 3 shapes 5332;4432 and 4333 with permutations 12;12;4. Exact shape can be gotten across in relatively small place. But partner can have hands like 5431;6421;6430... 24+24+24 permutations.
My logic tells me that it is more about permutations of my hand, if it is low ->show, if it is high->ask. So with 4333;4441;7222 i would prefer to show, but with hands with permutation of 24 ask rather than show.

[akhare]

straube, on 2011-April-22, 14:50, said:

No, I wouldn't use 1N for 5440s because 1C-1D, 1N is much more valuable than just that. We use it for both majors or 3-suited with both majors (including some 5440s). I would think Moscito players would use it differently.

Once again this goes back the the fundamental choice of what hands should reverse relay.

1) Richard prefers *most* that balanced ask rather than show

2) Wclass is skeptical about 1) and questions the rationale

3) IMO, 1C - 1D - 1N it's better to use 1♣ - 1D - 1N to show *min* balanced hands, at the cost of sacrificing the 1♣ - 1♦ - 1N reverse relay to show the two/three suited major hand.

While it's true that 1♣ - 1D - 1♥ - 1♠ (mostly balanced) - 1N will lead to right siding notrump, it's really silly to have to go through that auction and have to divulge so much information when it's possible to quickly discover 3N or 4M via the shorter router.

[straube]

wclass___, on 2011-April-22, 14:54, said:

Can someone please explain me what is the reasoning why balanced hands should ask?

If one shows balanced hand, he limits himself really well shape wise. There are basically only 3 shapes 5332;4432 and 4333 with permutations 12;12;4. Exact shape can be gotten across in relatively small place. But partner can have hands like 5431;6421;6430... 24+24+24 permutations.
My logic tells me that it is more about permutations of my hand, if it is low ->show, if it is high->ask. So with 4333;4441;7222 i would prefer to show, but with hands with permutation of 24 ask rather than show.

It's about whether the cards in the balanced hand are working (useful) to the unbalanced hand. I'm quoting myself from the other thread now...

straube said:

I'm referring to what has been called "the Balanced Hand Principle". Here's a link to mention of it by Danny Kleinman

http://books.google....iple%22&f=false

and apparently its name originated in a BridgeWorld article. In recent memory, we used a reverse relay to get to slam with only 17 QPs. Our hands were something like AQxx x AKx KQxxx opposite KJxxx xxx xx Axx. Partner knew with xxx of hearts opposite my singleton that slam was likely. Had I been captain and learning that partner had 5 QPs, I might have feared KJxxx KQx xx xxx. Whether we would have gotten to slam anyway is not my point. My point is that it was much easier to get to slam with our method.

[Echognome]

wclass___, on 2011-April-22, 14:54, said:

Can someone please explain me what is the reasoning why balanced hands should ask?

If one shows balanced hand, he limits himself really well shape wise. There are basically only 3 shapes 5332;4432 and 4333 with permutations 12;12;4. Exact shape can be gotten across in relatively small place. But partner can have hands like 5431;6421;6430... 24+24+24 permutations.
My logic tells me that it is more about permutations of my hand, if it is low ->show, if it is high->ask. So with 4333;4441;7222 i would prefer to show, but with hands with permutation of 24 ask rather than show.

The main reason is that when looking for slam, honor location will be important. That is much easier to judge when the balanced hand asks rather than tells. It is also more efficient when scanning (fewer suits to cover).

[akhare]

Echognome, on 2011-April-22, 15:26, said:

The main reason is that when looking for slam, honor location will be important. That is much easier to judge when the balanced hand asks rather than tells. It is also more efficient when scanning (fewer suits to cover).

On the flip side, isn't it true that balanced hands resolve lower? Most balanced hands should be able to complete their QP / PCB / DCB scan by 4N or so.

Contrast this with a 5530 long legged shape resolution that resolves at say 3N. The QP ask / DCB / PCB scan likely won't end before the 5-level..

[Echognome]

Obviously where a relay ends is entirely by choice. Most designs resolve more common shapes lower, so it's an unfair comparison. The fact that a 7420 is resolve at a higher level than a 5332 is due to frequency rather than something innate in the design of the relays. Maybe another way to say it is that I can swap around my relays so that 5530's are resolved lower than 5332's, without losing any shapes in my relay.

I do take your point, however, that we tend to resolve balanced shapes lower and that is perhaps a reason why saying the loss in the efficiency of the scan is made up for by the lower level of shape resolve. That is, however, entirely up to the design of the relay.

Of course there are other considerations as well, such as memory. I wonder what the average level of full shape resolution would be for the most efficient structure (measured precisely by the mean of level of when full shape is known over the distribution of all shapes that are shown) versus the more common symmetric structures that are easier on the memory. One could then have a measure of the efficiency of the relay. I am sure this can be done mathematically (if it hasn't been done already).