[straube]

KQT4 5 QJ63 K942
V AKQ82 AT8 AQ865

South is captain. Let's say North shows pattern and 6 QPs at 3N.

1) With rules of short to long (and stop with odd for singleton or doubleton), king parity last

4C-4H small heart, even spade
4S-5D odd diamond, odd club, no DK
Cards placed

2) With rules of short to long (and stop with odd for singleton or doubleton), king parity first

4C-4D even K
4H-4N small heart, even spade
5C-5S
cards placed

3) With rules of long to short (and stop for singleton only), K parity last

4C-4D even spade
4H-5D odd diamond, odd club, small heart, no DK

4) With rules of long to short (and stop for singleton only), K parity first

4C-4D even K
4H-4S even spade
4N-5S
cards placed

[foobar]

straube, on 2017-October-18, 17:30, said:

KQT4 5 QJ63 K942
V AKQ82 AT8 AQ865

South is captain. Let's say North shows pattern and 6 QPs at 3N.

Here's a Byzantine method with the following rules:

1) K-parity first (with K=2 QPs always)
2) Singletons and doubleton suits are scanned first, with ties broken in rank order. Skip with nothing or stop with A/K
3) 4+ cards suits are scanned in length order with ties broken in rank order. Stop with even and continue with odd

4C...4D (relay; even K-parity)
4H...4N (relay; nothing in ♥, even ♠) -> this can't be AK since another K > 6QPs; if it's AQ, we can't satisfy even-K parity, if it's xx, it's impossible to satisfy 6 QPs, ergo KQxx)
5C...5♠ (relay; odd [diamond, club]

Can I perform all these computations at the table though ?

[yunling]

4♣-4♥(ask 1&2, 0/3/6)
4♠-5♥(ask 2&3, 2/5 not stiff honor even Kings)

Cards placed
I'm certainly at a disadvantage here since captain is void in suit 1 and I can't skip that.

[foobar]

straube, on 2017-October-18, 17:30, said:

KQT4 5 QJ63 K942
V AKQ82 AT8 AQ865

South is captain. Let's say North shows pattern and 6 QPs at 3N.

2) With rules of short to long (and stop with odd for singleton or doubleton), king parity first

4C-4D even K
4H-4N small heart, even spade
Cards placed

4) With rules of long to short (and stop for singleton only), K parity first

4C-4D even K
4H-4S even spade
Cards placed

Can you really place the cards? Inferring ♠KQ seems to follow, but how can one disambiguate between K♦/K♣ at this point?

[straube]

So let's see if I apply this right. The current IMprecision method goes...

4C-4D even spade
4H-5H odd diamond, odd club, small heart, CK

At 4N we know odd diamond
At 5C we know odd club (perforce CK)
At 5D we know small heart (so this would actually be an impossible response here because the only combination KQxx x Qxxx Kxxx requires a 5H bid.

Looks good

[straube]

foobar, on 2017-October-18, 23:32, said:

Can you really place the cards? Inferring ♠KQ seems to follow, but how can one disambiguate between K♦/K♣ at this point?

Thanks. Good catch

[awm]

straube, on 2017-October-20, 21:16, said:

So let's see if I apply this right. The current IMprecision method goes...

4C-4D even spade
4H-5H odd diamond, odd club, small heart, CK

At 4N we know odd diamond
At 5C we know odd club (perforce CK)
At 5D we know small heart (so this would actually be an impossible response here because the only combination KQxx x Qxxx Kxxx requires a 5H bid.

Looks good

We don’t actually invert the parity on singletons so would stop at 5C (odd diamond, odd club, small heart). 5S over the next relay shows club king/heart queen.

[straube]

awm, on 2017-October-21, 00:26, said:

We don’t actually invert the parity on singletons so would stop at 5C (odd diamond, odd club, small heart). 5S over the next relay shows club king/heart queen.

To save a step when you don't need that last question answered...which you must expect is very seldom? Thinking you must have looked at a lot of hands to get that precise.

So let's say partner has a stiff honor and has to take an extra step for that. For starters that would be bad news for me as I'm learning late that partner may have a less useful card, but I think he still needs to zoom on and answer that last question (or it costs two steps). But when he does not have that stiff honor, don't I have an opportunity to change the question I'm asking? On this example, he stops at 5C...

5C (small stiff, the usual expected response)
.....5D-asks partner to bid 5H for p/c
.....5H-asks the last and seldom needed question
.....5S-asks jack from longest
.....5N-asks jack from second longest
.....etc

It's late here. Maybe this doesn't make sense.

[awm]

I do not think we have actually done such a deep analysis. In general my view has been that there are many exceptions that could be added (i.e. "count 3/2/1 points unless it's singleton honor in which case it's 2/1/0", "stop with even unless it's singleton or doubleton and then stop with odd"; actually there are a bunch of other treatments like this I've seen as well, like treating six-card suit differently, or reordering the suits in some auctions, etc). My analysis has been that these almost never make a significant difference, and that if they lead to even one forget/miscount that we wouldn't have had otherwise we are probably losing. So we don't adopt them!

It's not that we "know it is worse to swap odd/even with singleton" -- it's that we know it very rarely matters and even if it's better by one hand in a thousand, the risk of forget is higher than that so we don't bother.

[straube]

Trying not to be tiresome, but if system forgets were not an issue...

1) do you think 2/1/0 is better?
2) would you skip a small singleton?
3) would you skip a small doubleton?
4) would you consider scan the singleton first and then long to short?

What do you think of the idea of puppeting late in the auction in order to ask different questions (e.g. specific jacks)? Have you looked at that?

[awm]

We have not done the large set of simulations that would be needed to actually determine this. My intuition is:

1. 2/1/0 is slightly inferior.
2. Skip with small singleton is a slight improvement.
3. Doubleton may even depend on the overall shape, doubt it matters much either way.
4. Singleton first is worse (frequency issue mostly).

We almost never ask jacks so it won’t much matter, but I suspect the one step lost for jack of longest suit will be very costly.

[straube]

ok thanks a lot. At the moment I'm still wondering if there's potential to save space at the end by being able to ask different questions. Like if I don't need the marriage/K or Q question I could actually save a step (especially when partner would have bid step 2) and more if I don't need the longest suit jack (or hold it)

I'm also wondering about the stiff Q question...like it would be interesting to know how often it's useful to know this card compared to the longest jack. The stiff Q question could be another choice at the end.

[awm]

straube, on 2017-October-21, 03:20, said:

ok thanks a lot. At the moment I'm still wondering if there's potential to save space at the end by being able to ask different questions. Like if I don't need the marriage/K or Q question I could actually save a step (especially when partner would have bid step 2) and more if I don't need the longest suit jack (or hold it)

I'm also wondering about the stiff Q question...like it would be interesting to know how often it's useful to know this card compared to the longest jack. The stiff Q question could be another choice at the end.

Suppose after 5♣ you are looking for a jack. It seems like:

1. First jack in sequence, 5♦ ask wins a step vs. 5♥ ask.
2. Second jack in sequence. 5♦ ask wins a step if partner has the first jack, breaks even vs. 5♠ ask if he doesn't.
3. Third jack in sequence. 5♦ ask wins a step if partner has the first TWO jacks, breaks even if partner has one of the first two jacks, loses a step if he has neither.
4. Fourth jack in sequence, 5♦ wins a step if partner has the first THREE jacks, breaks even if partner has TWO of the first three, loses a step if he has ONE, loses two steps if NONE.

Unless you find yourself often looking for the jack in partner's singleton suit, this is probably behind.

[straube]

ok. I give up

[sieong]

EPCB.
4♣ - 4♦ (even ♠)
4♥ - 5♦ (even ♥, odd ♦, odd ♣)

Honor structure is resolved at this point, although 5♥ will still be asking KQ resolution in the minors.

FWIW, my intuition on the four questions are same as Adam. Regarding J asks, there are more efficient structures, at the cost of not being able to stop as low compared to classic PCB. There has been less than 50 J asks in our partnership if memory serves right, and we have probably played maybe 5k hands together and bid another 10k hands ... so I would probably avoid optimizing for that.

[nullve]

straube, on 2017-October-18, 17:30, said:

KQT4 5 QJ63 K942
V AKQ82 AT8 AQ865

Auction, with some ambiguity left out:

(...)
3N-4♣ (10-12 hcp (say); relay)
4♦-4♥ (even # of non-singleton Ks, no singleton K; relay)
4♠-4N (even # of non-singleton queens; relay)
5♦-5♥ (♠K, no♥K (hence ♣K); relay)
6♣-? (♠Q, ♥Q, no ♠J; ?).

This post has been edited by nullve: 2017-November-13, 10:22