[shevek]

In our relay method, asker has the choice of control asks after shape is out.
For us, first step = Slam Points (A=3, K=2, Q=1) second step = kontrols (A=2, K=1)
Sometimes asker makes what turns out to be the wrong ask.

Le's say you ask for Slam Points to check out 6♥ where you find out that opponents have 6 SPs, including ♥A.
You have a stiff club opposite either ♣Axx or ♣KQx.

Do you ou think "Damn. I should have asked for controls"?
Maybe but your SP ask allowed you to pick up the esssential ♠Q.
There is no time to do another lap to clarify the clubs.

Do you think "A method that equates A with KQ is no good" and sign off in 5♥?

Or do you shrug and bid 6♥ anyway?
After all, 9C2 = 36, while 9C1 = 9. Does that really make ♣Axx four times as likely?

[nige1]

delete please

[straube]

You're comparing the combinations of two small cards vs one small card (J or less)? I'm getting 45 vs 9. Anyway, I'd go with the odds then if I just didn't have the room.

If you used parity cue bidding you might have less of a guess on this hand, but you would have other guesses.

[The_Badger]

 shevek, on 2017-September-10, 20:44, said:

Do you think "A method that equates A with KQ is no good" and sign off in 5♥?

I think that one sentence has clarified how you feel about this aspect of your relay system. It is good most of the time but not good enough: it has a flaw.

How many times have we all said that about a particular aspect of our system?

[sfi]

Yes - I would go with the percentages and play partner for Axx absent any other information. Teammates understood the concept and you win more IMPs than you lose. (I agree with shevek that it's 36-9. You have 9 small cards in both situations, since you know partner can't hold AK or AQ.)

I played shevek's system for a few years and noticed something curious. This requires a bit of background:

On the second pass of denial cues (which is what happens after you find out about AK or AKQ points), you get the chance to show a second honour in a suit where you have already shown something. Since you can't have a really strong holding if you show an honour (AK or AQ after having shown slam points or AK after showing kontrol points - those deny anything in the suit and partner can usually work it out), this second honour can be the queen when asking about kontrols or the jack when asking about slam points. But only if partner is not going to be confused - you really don't want partner playing you for AJ when you have KQ in a suit.

Given that, there were a number of auctions where it was better to ask about AK points if I cared about a specific queen, or AKQ points if I didn't care about a queen that partner might hold. It was always slow to work out those positions though (there was almost always a pause after finding out about partner's shape) and I never managed to put together any general guidelines for doing so.

[dcrc2]

You need better methods. My preference is not to relay on unbalanced hands in the first place. If you insist on relaying with all shapes, OK but you're going to need to have a better variety of asking bids. If your only choice is controls vs QP, that isn't going to cut it.

A good method is to have a way to set trumps. This should certainly be possible here where your suit is a major. Then normally the first response is keycard, but after that you can find specific cards in other suits.

[hrothgar]

 dcrc2, on 2017-September-11, 02:34, said:

You need better methods. My preference is not to relay on unbalanced hands in the first place. If you insist on relaying with all shapes, OK but you're going to need to have a better variety of asking bids. If your only choice is controls vs QP, that isn't going to cut it.

A good method is to have a way to set trumps. This should certainly be possible here where your suit is a major. Then normally the first response is keycard, but after that you can find specific cards in other suits.

Assuming away the question really isn't particularly useful...

[Zelandakh]

 sfi, on 2017-September-11, 02:01, said:

Given that, there were a number of auctions where it was better to ask about AK points if I cared about a specific queen, or AKQ points if I didn't care about a queen that partner might hold. It was always slow to work out those positions though (there was almost always a pause after finding out about partner's shape) and I never managed to put together any general guidelines for doing so.

There are some methods you can use to offset some of these issues. An important one I have found from my methods (based on controls) is to use a relay break to NT as a queen ask. This gets around your issue of AK points being better but needing to know about a specific queen. For queen points, the most convincing methods I have seen are based on parity advances, which seems to offset some of the drawbacks on most hands. QPs tend to be designed around exchanging the maximum amount of information quickly rather than precise information if their were an unlimited amount of space. Because of this, it is not uncommon to have to go with the percentages on a few hands. Against that, it is rarer to run out of space before being in a position to make this decision than with controls.

My current view is that systems in which the known hand is highly limited in terms of hcp are on average best described using controls and hands that are somewhat less well defined in terms of strength benefit much more from QPs. It is fair to say that this assessment has changed somewhat over time though and I think the advantages and disadvantages of each approach are close enough that there is no clear winner at present. One thing I will say about systems that allow for both types of ask is that a key feature to note is kings in partner's non-shortage suits. Because of the parity responses, missing kings are often the part that causes ambiguity in QP relay auctions. In this respect, holding the majority of these kings can be a good indicator towards using the QP relays. Having few or no such kings might suggest using controls instead, particularly if you can see that no more than one queen is going to be critical further down the line. With few kings and also multiple critical queens it is difficult to know which is best. It might be best to fall back on basics here - controls if partner is highly limited and QPs if their range is less well defined. Fortunately this latter scenario is fairly rare in slam auctions.

In any case, the above guidelines (essentially kings for QPs, queens/aces for CPs) are somewhat unintuitive so perhaps they might at least provoke some thought in re-analysing some hands where there was some difficulty and finding guidelines that work most of the time for your specific system. Despite dcrc2's comment, I am strongly of the belief that shape relays represent a much more efficient method of bidding than Standard. It is nice to be able to drop out into Standard methods sometimes and many systems do incorporate such sequences but in most of these systems the majority of hands will continue in relays until the final contract is reached and RKCB auctions will be the minority approach.

[straube]

 Zelandakh, on 2017-September-11, 06:35, said:

One thing I will say about systems that allow for both types of ask is that a key feature to note is kings in partner's non-shortage suits. Because of the parity responses, missing kings are often the part that causes ambiguity in QP relay auctions.

We have the K parity question after scanning each suit for odd/even. I've wondered, though, if there would be a useful question before even the first suit is scanned for odd/even. Like, do you have an odd or even number of kings? I.e. a question that would pertain to the whole hand, that might eliminate a lot of possibilities at the outset and as the scans progressed. Probably there is a better initial question.

Sorry for the hijack

[shevek]

 sfi, on 2017-September-11, 02:01, said:

Yes - I would go with the percentages and play partner for Axx absent any other information. Teammates understood the concept and you win more IMPs than you lose. (I agree with shevek that it's 36-9. You have 9 small cards in both situations, since you know partner can't hold AK or AQ.)

I played shevek's system for a few years and noticed something curious. This requires a bit of background:

On the second pass of denial cues (which is what happens after you find out about AK or AKQ points), you get the chance to show a second honour in a suit where you have already shown something. Since you can't have a really strong holding if you show an honour (AK or AQ after having shown slam points or AK after showing kontrol points - those deny anything in the suit and partner can usually work it out), this second honour can be the queen when asking about kontrols or the jack when asking about slam points. But only if partner is not going to be confused - you really don't want partner playing you for AJ when you have KQ in a suit.

Given that, there were a number of auctions where it was better to ask about AK points if I cared about a specific queen, or AKQ points if I didn't care about a queen that partner might hold. It was always slow to work out those positions though (there was almost always a pause after finding out about partner's shape) and I never managed to put together any general guidelines for doing so.

Stephen, notes said that 2nd pass picked up KJ, KQ, QJ but specifically not AJ.
Jacks are such big cards in trumps. I was amused by a recent hand where they had the kitchen sink but ♠Kxx opposite ♠Qxxxx.
Everyone except us steamed happily into 6♠

[DinDIP]

Having used the methods in question with both shevek and sfi it's no surprise I am strongly in favour of taking the 4-1 odds.

 straube, on 2017-September-11, 08:56, said:

We have the K parity question after scanning each suit for odd/even. I've wondered, though, if there would be a useful question before even the first suit is scanned for odd/even. Like, do you have an odd or even number of kings? I.e. a question that would pertain to the whole hand, that might eliminate a lot of possibilities at the outset and as the scans progressed. Probably there is a better initial question.

Sorry for the hijack

I experimented with a king parity ask after ascertaining the number of QP but before scanning any suit, i.e. after QP were known the next relay asked for step one if holding an odd (or even depending on the number of QP) number of non-singleton kings. The point of this was that relayer could work out which permutations of honours were possible and, either after the parity response or after one or more DCB steps, would know whether partners three QP were Axx or KQx.

I played the method with Mark Abraham in one event (in which it made no positive or negative impact) but it became part of Mark's default methods (see the SCREAM or SPREAD notes here). Testing showed it to gain more than it lost but the sample size was small: the losses were caused by the answer to the parity ask preventing the necessary scan of a key suit below the safety threshold.

[straube]

That seems better because of course you want to skip more often than you stay. What's your specific king parity question? I.e. do you ask partner to show whether he holds the king in his first odd suit even though that odd suit is as yet unknown?

I'd wondered if the first question could be odd or even controls (but that's likely a 50/50 answer, too) or whether most of the QPs were located in the longest two suits (as would be expected and it should give something better than a 50/50 answer) but that would depend quite a lot on the total length of the two suits.

On a side note, have you looked at reversing the odd/even responses for doubletons (as is done for singletons)?

[DinDIP]

 straube, on 2017-October-13, 23:02, said:

That seems better because of course you want to skip more often than you stay. What's your specific king parity question? I.e. do you ask partner to show whether he holds the king in his first odd suit even though that odd suit is as yet unknown?

I'd wondered if the first question could be odd or even controls (but that's likely a 50/50 answer, too) or whether most of the QPs were located in the longest two suits (as would be expected and it should give something better than a 50/50 answer) but that would depend quite a lot on the total length of the two suits.

The parity ask is for the whole hand, i.e. the total number of non-singleton kings teller holds. After parity is known asker often knows which honour permutations teller holds. For example, if teller bids 3N showing a 5431 with 6QP then zooms to show odd parity of kings (significantly more frequent than even parity with this number of QP) then teller is known to hold AKQ or KKK or KQQQQ. Asker can frequently tell which option it is (for example, if she has one queen and two kings then neither KQQQQ nor KKK is possible). Conventional DCB can now locate/deny honours. Together, asker is frequently able to know the exact honour holding lower than just using DCB.

 straube, on 2017-October-13, 23:02, said:

On a side note, have you looked at reversing the odd/even responses for doubletons (as is done for singletons)?

No, as I haven't used parity scanning in individual suits but, using standard DCB, we reverse the scan if teller shows an honour in a doubleton because HH is much less likely than Hx and because HH is typically worse for slam so it's helpful to stop lower. The same principle applies if teller ever does a third scan of a three-card suit.

[straube]

I've sometimes wondered if suits should be scanned shortest to longest. For example, if I know partner has 5341 I'm statistically more likely to be maybe 2425 and have a hard time telling if partner's first "even" (in spades) is xx or HH. I also might take the view that partner's singleton is small until late in the scan and then I get a surprise when I find that it's an honor. Maybe that's wrong thinking but I'm throwing it out there.

I personally find it annoying when that singleton is an honor and learn about it late. Like if 3 of partner's QPs are in a stiff ace that might mean that the hand is not working as well as I would have liked. For that reason, I've taken to counting stiff Ace as 2, stiff King as 1 and stiff Q as 0 (but still stopping for the Q). Adam thinks this is wrong and assigns full value and I'm in doubt about it either way. Under-counting has problems, too.

I would be interested in starting a thread to compare various scanning methods. Anyone else interested? Adam?

[foobar]

 straube, on 2017-October-16, 08:43, said:

I've sometimes wondered if suits should be scanned shortest to longest. For example, if I know partner has 5341 I'm statistically more likely to be maybe 2425 and have a hard time telling if partner's first "even" (in spades) is xx or HH. I also might take the view that partner's singleton is small until late in the scan and then I get a surprise when I find that it's an honor. Maybe that's wrong thinking but I'm throwing it out there.

I personally find it annoying when that singleton is an honor and learn about it late. Like if 3 of partner's QPs are in a stiff ace that might mean that the hand is not working as well as I would have liked. For that reason, I've taken to counting stiff Ace as 2, stiff King as 1 and stiff Q as 0 (but still stopping for the Q). Adam thinks this is wrong and assigns full value and I'm in doubt about it either way. Under-counting has problems, too.

I would be interested in starting a thread to compare various scanning methods. Anyone else interested? Adam?

Off the top of my head, think that discounting stiff A is a mistake. It would certainly be interesting to evaluate awm's method of according full value to stiff-Ks vs. the more common stiff-K=1. Regarding long vs. short, the approach is probably driven by the simple statistical probability, and my guess would be that you will likely draw several initial blanks in the PCB (evens), but it would be an interesting experiment none-the-less.

[straube]

RR bids Step 1 or higher according to the following scheme:
King Parity..... AKQ controls ............................................ Action
Even ..............2, 3, 6, 9, 10, 12, 13, 16, 18, 19, 20 ......... Bid Step 1
Even ............. 4, 5, 7, 8, 11, 14, 15, 17, 21+......................Zoom to Step 2 or higher
Odd............... 2, 3, 6, 9, 10, 12, 13, 16, 18, 19, 20........... Zoom to Step 2 or higher
Odd............... 4, 5, 7, 8, 11, 14, 15, 17, 21+..................... Bid Step 1

Seems a lot to remember. I wonder if there's an easier (if not as good) rule.

[straube]

 foobar, on 2017-October-16, 15:14, said:

Off the top of my head, think that discounting stiff A is a mistake. It would certainly be interesting to evaluate awm's method of according full value to stiff-Ks vs. the more common stiff-K=1. Regarding long vs. short, the approach is probably driven by the simple statistical probability, and my guess would be that you will likely draw several initial blanks in the PCB (evens), but it would be an interesting experiment none-the-less.

No one has ever liked my idea of discounting a stiff ace. I would argue that a stiff ace is worth less than a working king and if one is on the verge of asking for PCB one would say "Well, partner has X QPs but he has potential to have a singleton ace (because I don't have it in that suit) and if he has that ace then that is 2 of his X QPs and we still have a shot at slam.

Plus if we scanned shortest to longest, we'd know early.

Maybe short to long is a bad idea, but I just sampled a very few hands and it can be useful when captain has a long suit opposite shortness.

For the doubletons I was thinking skip with even and stop with odd.

[sieong]

I thought the topic of comparing different scanning methods came up before, and Richard (hrothgar) has a series of threads where folks compared different scanning methods. Going through the archive, it seemed like there was no consensus?

Fwiw, I have been considering a different approach to simulate the outcome, so as to remove the subjectivity of such exercise. I will try to post the idea once it is fleshed out some more.

[awm]

At some point I actually did a series of sims to determine the level/likelihood that relayer can locate all honors under different methods. Mostly I was looking to find the most useful single bit of information after QP count and honor parity. Anyway it seems possible to do the same here.

[DinDIP]

 straube, on 2017-October-16, 20:14, said:

RR bids Step 1 or higher according to the following scheme:
King Parity..... AKQ controls ............................................ Action
Even ..............2, 3, 6, 9, 10, 12, 13, 16, 18, 19, 20 ......... Bid Step 1
Even ............. 4, 5, 7, 8, 11, 14, 15, 17, 21+......................Zoom to Step 2 or higher
Odd............... 2, 3, 6, 9, 10, 12, 13, 16, 18, 19, 20........... Zoom to Step 2 or higher
Odd............... 4, 5, 7, 8, 11, 14, 15, 17, 21+..................... Bid Step 1

Seems a lot to remember. I wonder if there's an easier (if not as good) rule.

Given that this was being used in the context of a strong hand asking, the simple mnemonic was even parity for multiples of three and for two and ten QP.