[semeai]

David Bird and Taf Anthias have a nice article in the Thursday Euro Open bulletin on lead simulations for the auction 1NT-3NT. (They also say they have a book on leads against notrump coming out soon.)

One of the surprises for me: When leading from KQxx, the King is better than low by a fair margin at both imps and matchpoints when the suit is a major. In a minor suit, K or low are roughly equal. Moreover, when this is a major, it's often the best lead.

This is especially disturbing as I've played before that the King at notrump requests count or unblock while playing the rest of the honor leads standard. With those agreements, I wouldn't be able to lead the K or Q systemically. Have others played this or was I just crazy?

Here are their lead problems for you to try, with brief comments on their results in a hidden comment (see the bulletin article for discussion). The auction has always been 1NT-3NT, and 1NT is 15-17. You may give imp and mp answers.

I hope nobody minds my quoting them for discussion. Presumably it's fair use in any case. Several of them seem controversial (surely intended on their part).

Edited: Previously #0 was #1 and I left off #1 but the answer was to the current #1!

0. J86 95 KQ72 KJ72
1. KQ72 KJ72 J86 95
2. A8 KQ82 KJ53 853
3. Q965 104 K82 QJ102
4. Q6 K5 10873 J10432
5. K105 A5 1086532 84

Spoiler

0. Imps: S6 > H9 >> C2 > D2 > DK. MP: S6/H9 >> C2 > DK/D2
1. SK best at both. Others about equal.
2. HK >> C/H2/SA >> D3 at both
3. Imps: H10 > CQ/S5 > D2. MP: CQ > S5/H10 > D2
4. Imps: SQ/HK >> D3/C3. MP: SQ > D3/C3 > HK.
5. Imps: HA >> D5 > S5 >> C8. MP: Same, but put > instead of >>.
I'm not sure how much the "Aces are better double-dummy because you can always switch correctly" phenomenon affects the last result. They don't mention this.

If you disagree with their answers, some reasoning on why their lead is worse single dummy would be welcome.

[JLOGIC]

What a shock, when the opponents always guess correctly when you lead small, the king is better than small!

On a more serious note, when you have a ton of high card points and/or another good suit, the king is the right lead imo (KQxx KJxx Jxx xx is a perfect example of this, low might give them their 9th trick whereas if the SK is wrong and you can run hearts, they might have only 8 tricks). You effectively combine your chances better with both suits, similar to leading an ace with AQxx AQxx xxx xx.

[bluecalm]

Quote

What a shock, when the opponents always guess correctly when you lead small, the king is better than small!

+1
The only way for people who are not programmers is to fire up something like dm pro and manually go through the cases to see how often declarer guesses.It could be made automatic but it requires some coding which for now nobody did (or published). I think that if the book will be terrible enough it will be good incentive for me to actually do it

Quote

3. Imps: H10 > CQ/S5 > D2. MP: CQ > S5/H10 > D2

I can tell from experience that this is wrong and in fact after running the simul:
6♠ - 151
T♥ - 177
2♦ - 162
Q♣ - 194

And there is a factor that xx major leads are overrated by dd simuls because with semi-solid suit and side entry partner would often double 3NT (or as some play he would do taht only with spades then xx spade lead is overrated).
I suspect that they screw up the assumptions and forgot about some cases (like for 1NT - 3NT auction responder doesn't have stiff major (unless it's at least a Q generally) or they excluded 4M-3-3-3 hands which is also a mistake).

[semeai]

JLOGIC, on 2011-June-30, 13:56, said:

What a shock, when the opponents always guess correctly when you lead small, the king is better than small!

On a more serious note, when you have a ton of high card points and/or another good suit, the king is the right lead imo (KQxx KJxx Jxx xx is a perfect example of this, low might give them their 9th trick whereas if the SK is wrong and you can run hearts, they might have only 8 tricks). You effectively combine your chances better with both suits, similar to leading an ace with AQxx AQxx xxx xx.

Thanks, two nice points. What are the main guesses?

dummy declarer
J9x opposite Axx
J8x opposite Axx
Jxx opposite A9x
Jxx opposite A8x

Others?

bluecalm, on 2011-June-30, 14:05, said:

I can tell from experience that this is wrong and in fact after running the simul:
6♠ - 151
T♥ - 177
2♦ - 162
Q♣ - 194

I assume this is # times set out of 1000? They have (scaled to 1000):
5♠ - 231
T♥ - 250
2♦ - 193
Q♣ - 232

Given this it seems their simulation raises to game more aggressively than yours. They show an example hand (not this one) in which dummy is QJx xx Jx K10xxxx, so they're not religious point counters.

I'm not sure how this affects things. I suppose it makes passive leads better, but that's not the issue on this hand.

Quote

I suspect that they screw up the assumptions and forgot about some cases (like for 1NT - 3NT auction responder doesn't have stiff major (unless it's at least a Q generally) or they excluded 4M-3-3-3 hands which is also a mistake).

I was wondering about the singleton question myself, as 3M or even 3x showing a singleton is popular.

[bluecalm]

Quote

I was wondering about the singleton question myself, as 3M or even 3x showing a singleton is popular.

Assumptions I usually use for responder are:
2-3 card in majors, 9-15hcp or any 4-3-3-3 9-15hcp.

Hands with 8hcp and 6 card minor usually don't raise to 3nt imo but I am not sure. I can check in my database if that's turns out to be important point but I highly doubt it.

[Cascade]

bluecalm, on 2011-June-30, 14:05, said:

+1
The only way for people who are not programmers is to fire up something like dm pro and manually go through the cases to see how often declarer guesses.It could be made automatic but it requires some coding which for now nobody did (or published). I think that if the book will be terrible enough it will be good incentive for me to actually do it

You can do this single dummy with GIB (rather than human analysis) if you are willing to accept the errors in GIBs play.

I have done 10000 or so hands comparing leads against 1NT 3NT.

[gordontd]

semeai, on 2011-June-30, 12:58, said:

This is especially disturbing as I've played before that the King at notrump requests count or unblock while playing the rest of the honor leads standard. With those agreements, I wouldn't be able to lead the K or Q systemically.

With those agreements, isn't the queen the systemic lead when you don't want partner to unblock?

[semeai]

gordontd, on 2011-July-01, 02:36, said:

With those agreements, isn't the queen the systemic lead when you don't want partner to unblock?

Well, the Q showed the J or shortness ("standard leads" other than the K), so it would have problems too. Maybe you're suggesting playing the Q shows the K or the J or shortness. Is this a reasonable way to play?

[gordontd]

semeai, on 2011-July-01, 04:53, said:

Well, the Q showed the J or shortness ("standard leads" other than the K), so it would have problems too. Maybe you're suggesting playing the Q shows the K or the J or shortness. Is this a reasonable way to play?

What I'm saying is that if you play that K asks for unblock, then with KQx(x) you need to lead the Q. So your K agreement has a knock-on effect that not all your other leads can be standard.

If I'm playing that way and I'm asked about Q leads, I say "we play that K asks for unblock or count so sometimes we have to lead the queen from KQ. Otherwise it's standard".

[jogs]

semeai, on 2011-June-30, 12:58, said:

David Bird and Taf Anthias have a nice article in the Thursday Euro Open bulletin on lead simulations for the auction 1NT-3NT. (They also say they have a book on leads against notrump coming out soon.)

One of the surprises for me: When leading from KQxx, the King is better than low by a fair margin at both imps and matchpoints when the suit is a major. In a minor suit, K or low are roughly equal. Moreover, when this is a major, it's often the best lead.

DD analysis cannot be trusted. Print hundred boards where the king and small produced different results. Inspect them to see if you agree with the DD results.

[semeai]

jogs, on 2011-July-01, 12:17, said:

DD analysis cannot be trusted. Print hundred boards where the king and small produced different results. Inspect them to see if you agree with the DD results.

Sounds good, though I'm not up for generating them myself. If someone wants to dump 100 hands I'd be happy to try my hand at at least some of them. If others want to analyze some portion too, we could distribute the work.

[jogs]

semeai, on 2011-June-30, 12:58, said:

0. J86 95 KQ72 KJ72
1. KQ72 KJ72 J86 95

Why are answers to 0. and 1 different?
Were there assumptions made on the length of the majors
in the dummy?

[semeai]

jogs, on 2011-July-01, 17:30, said:

Why are answers to 0. and 1 different?
Were there assumptions made on the length of the majors
in the dummy?

The auction was 1NT-3NT, so there's the assumption that responder didn't bid Stayman (or anything else besides 3NT).

[DavidBird]

The articles that Taf Anthias and I wrote for Bridge Magazine
(one of which was reprinted in a Poznan bulletin) are part of
a 200-page book to be published by Master Point. In that book
we address some of the subjects that BBO members have mentioned
in this discussion.
We are fully aware that double-dummy simulation has a few
side effects, such as saving declarer a guess when you lead
the king from K-Q-x-x-x. This is an excerpt from the book that
covers it:

How often does an honor lead save declarer a guess?

Our double-dummy simulations suggest that it is profitable to lead from touching honors more often than traditional text-books imply. Occasionally the lead of a king from such as sxK-Q-6-5-2 will save declarer a guess. Look at this single-suit position, for example:

..... sx J 9 4

sx K Q 6 5 2 ..... sx 10 8

..... sx A 7 3

If you lead low against a real-life declarer, he may play dummy’s sx9, which will create a second trick for him when your spades are headed by the K-10 or the Q-10. The defenders can then establish four tricks in the suit. (A low lead would not succeed at double-dummy because the all-seeing computer declarer would rise with dummy’s jack.)

If instead you lead the king, declarer can place you with the K-Q and will score two tricks in the suit. This sort of position may throw some doubt on the validity of using double-dummy simulation, so we have taken some trouble to evaluate how often these ‘save declarer a guess’ positions arise.

We have written software that analyzes the frequency of the various honor positions for North/South on one of our 5000-deal simulations. Look back at Hand 9, where the sxK lead was rated best from sxK-Q-6-5-2. How often would such a lead have saved declarer a guess in the suit? These are the figures from our profile:

North South Probability
A J 9 ..... x x (+) ..... 0.36%
A J x ..... 9 x (+) ..... 0.54%
J 9 x ..... A x (+) ..... 1.44%
J x x ..... A 9 (+) ..... 1.20%

Leading the king will save declarer a guess in the suit on 3.54% of the deals. Does this wipe out the 3.2% advantage awarded to the sxK lead in our table for on Hand 8, above? No, because on many of the deals the contract would be made, however declarer plays the spade suit. Also, on some deals declarer cannot afford to lose a trick in spades (because he fears a switch to some different suit); he might then play dummy’s jack even after a low lead.

So, we have ‘come clean’ on this aspect of double-dummy simulation. There is a similar effect when you lead a top card from A-K-x-x-x. This may save declarer a guess when dummy has Q-10-x, or the Q-x-x in dummy with the 10 in declarer’s hand. When you lead an honor from Q-J-x-x-x, you may find dummy with K-10-x and declarer with A-8-x. Whenever our simulation results recommend an honor lead from touching honors, you can mentally subtract a percentage point or two from the ratings, to allow for the effects against a living, breathing declarer.

END OF EXCERPT

Another member's comment referred to the fact that if you lead king for unblock-or-count against no-trump
contracts, you cannot do the same from K-Q-x-x. Yes, but all players who use this method are familiar
already with leading the queen when they do not want unblock-or-count. 'Ace for attitude, king for count'
embeds also 'queen for attitude'. The fact that players may now lead an honor from K-Q-x-x has no
bearing on the matter. (In the US they tend to lead the ace to pass the unblock-or-count message and
therefore lead the K from A-K when they do not want an unblock.)

[wclass___]

How should partner signal when dummy has Axx and ducks? Can you pick up xx, Jx and Jxx in dummy?

What if dummy has x(xx)?
(DD simulator won't make mistakes here, but players in RL? )

How often we will just blow our natural trick if declarer has A and partner has no help?
(Feels like there is quite a big difference between KQ8x and KQ7x, in case partner has 9xx)

If i hit Jx with partner it will be somewhat awkward, but if he has J9 or i have that 9 it might already be bad.

IMHO toying with percentages and/or DD simulators is not the best approach to these problems.

[wclass___]

On hand 3 (Q965 104 K82 QJ102), just like bluecalm simulation suggest i would prefer ♣.

On hand 4 (Q6 K5 10873 J10432), although most of the time one of major leads will be good, we can't know which one. Splits and spot cards are good enough for us to expect that opponents will have hard time making anyway and i don't see why we would risk so much with major while we have JTxxx suit.

Also if partner has chance to bid before LHO opened 1N, major leads become reasonably worse.

[semeai]

DavidBird, on 2011-July-27, 04:24, said:

We have written software that analyzes the frequency of the various honor positions for North/South on one of our 5000-deal simulations. Look back at Hand 9, where the sxK lead was rated best from sxK-Q-6-5-2. How often would such a lead have saved declarer a guess in the suit? These are the figures from our profile:

North South Probability
A J 9 ..... x x (+) ..... 0.36%
A J x ..... 9 x (+) ..... 0.54%
J 9 x ..... A x (+) ..... 1.44%
J x x ..... A 9 (+) ..... 1.20%

Leading the king will save declarer a guess in the suit on 3.54% of the deals. Does this wipe out the 3.2% advantage awarded to the sxK lead in our table for on Hand 8, above? No, because on many of the deals the contract would be made, however declarer plays the spade suit. Also, on some deals declarer cannot afford to lose a trick in spades (because he fears a switch to some different suit); he might then play dummy's jack even after a low lead.

DavidBird, on 2011-July-27, 04:24, said:

Another member's comment referred to the fact that if you lead king for unblock-or-count against no-trump
contracts, you cannot do the same from K-Q-x-x. Yes, but all players who use this method are familiar
already with leading the queen when they do not want unblock-or-count. 'Ace for attitude, king for count'
embeds also 'queen for attitude'. The fact that players may now lead an honor from K-Q-x-x has no
bearing on the matter. (In the US they tend to lead the ace to pass the unblock-or-count message and
therefore lead the K from A-K when they do not want an unblock.)

Thanks for posting here, and for your helpful comments. I'm happy to know you took the care to worry about non-double-dummy play in the book.

[xxhong]

I think K from KQxx can be good in many ways, suppose partner holds xxxxx, that would be the only lead to be effective when opps are 2-2.
Also, it retains a tempo and allows you to switch without yielding a trick in that suit. These two advantages should weight more than the misguess situations.

semeai, on 2011-June-30, 12:58, said:

David Bird and Taf Anthias have a nice article in the Thursday Euro Open bulletin on lead simulations for the auction 1NT-3NT. (They also say they have a book on leads against notrump coming out soon.)

One of the surprises for me: When leading from KQxx, the King is better than low by a fair margin at both imps and matchpoints when the suit is a major. In a minor suit, K or low are roughly equal. Moreover, when this is a major, it's often the best lead.

This is especially disturbing as I've played before that the King at notrump requests count or unblock while playing the rest of the honor leads standard. With those agreements, I wouldn't be able to lead the K or Q systemically. Have others played this or was I just crazy?

Here are their lead problems for you to try, with brief comments on their results in a hidden comment (see the bulletin article for discussion). The auction has always been 1NT-3NT, and 1NT is 15-17. You may give imp and mp answers.

I hope nobody minds my quoting them for discussion. Presumably it's fair use in any case. Several of them seem controversial (surely intended on their part).

Edited: Previously #0 was #1 and I left off #1 but the answer was to the current #1!

0. J86 95 KQ72 KJ72
1. KQ72 KJ72 J86 95
2. A8 KQ82 KJ53 853
3. Q965 104 K82 QJ102
4. Q6 K5 10873 J10432
5. K105 A5 1086532 84

Spoiler

0. Imps: S6 > H9 >> C2 > D2 > DK. MP: S6/H9 >> C2 > DK/D2
1. SK best at both. Others about equal.
2. HK >> C/H2/SA >> D3 at both
3. Imps: H10 > CQ/S5 > D2. MP: CQ > S5/H10 > D2
4. Imps: SQ/HK >> D3/C3. MP: SQ > D3/C3 > HK.
5. Imps: HA >> D5 > S5 >> C8. MP: Same, but put > instead of >>.
I'm not sure how much the "Aces are better double-dummy because you can always switch correctly" phenomenon affects the last result. They don't mention this.

If you disagree with their answers, some reasoning on why their lead is worse single dummy would be welcome.