[awm]

---

IMPs, strong team match. Partner leads the ♦3 at trick one, ♦5 from dummy, ♦A from me, ♦2 from declarer.

Our agreement is 3rd from even and low from odd, so the ♦3 looks like it's fifth best from five (or perhaps third from three if declarer has six diamonds). 2NT is a standard 20-21 balanced (obviously LHO can upgrade/downgrade if he wants).

What should I play at trick two?

[whereagles]

Going with a club here.

Partner is unlikely to have led from ♦K or Q into a 2NT opener, so he rates to have a blank suit there. If we continue diamonds a club might be discarded on declarer's would-be ♦KQ. Of course, if pard did happen to underlead a diamond honor, the opposite could be true and a diamond back might be the right play, lest declarer cash AKQ♣ to throw a diamond from dummy.

In any case a club should be statistically better because a diamond is probably only good if pard has the king. Also, even if partner has a diamond honor, setting up a possible club trick might be preferable.

All in all, stuff definitely points at a club shift.

[wank]

it could be right to return a diamond, but only when partner has the k of d plus 2 tricks in the majors, which is quite a specific collection. better odds to try the clubs imo.

[shyams]

I think there's no hurry to switch to clubs. If partner has ♣AQ, our side cannot have any other honor card and we cannot defeat 4♥.
If West has ♣AQT9, a club switch would do more harm than good.

West's problem is entries to try all their finesses. I think a ♣ switch will help solve West's problem.

A diamond continuation looks less harmful.

[johnu]

shyams, on 2014-May-13, 13:03, said:

I think there's no hurry to switch to clubs. If partner has ♣AQ, our side cannot have any other honor card and we cannot defeat 4♥.
If West has ♣AQT9, a club switch would do more harm than good.

West's problem is entries to try all their finesses. I think a ♣ switch will help solve West's problem.

A diamond continuation looks less harmful.

Hmmm, dummy has 5 HCP, you have 5 HCP, declarer has ~20-22. Are you sure that if partner has ♣AQ that declarer has the rest of the honors? They must play really strong 2NT where you play

[mikeh]

It sure looks as if declarer has 4 diamonds, and there is a risk that a diamond return will establish a trick on which he could pitch a club loser. Partner's presumed major winner may be in trump

---

I don't buy into the notion that partner won't have a diamond honour: indeed, I think it would be odd to lead low from xxxxx. My example hand is but one illustration of the problem.

I think a club is clear: not as in it is the only switch that might work but in the sense that it seems the most likely to work, assuming that a set is possible.

[gszes]

p is slated t0 have around 9 possibly 10 hcp
and the underlead of a K is very risky since
it is known dummy will hit with 54 in the majors.

Given this info a lead from the dia K might be an
acceptable risk only if it will not create a disaster
if p has the "help" we need to keep the lead from being
bad in the first place. If p has underled the dia K they
will almost certainly have the J also. This means it
cannot really hurt to return a dia.

A dia may help if p has AJ8 of trumps (to go along with
KJ of diamonds since p can return a low dia at trick 3
and start an uppercut sequence against dummy possibly
getting our side 2d and 2h when we have no real hope
of getting tricks elsewhere. When we return the dia T
p will have a great idea of how to continue since it
will be very hard for declarer to be short in diamonds.

[mikeh]

I don't understand why anyone would think that partner wouldn't underlead the diamond K. We have one poster saying that he would never make that lead into a 2N bid and another saying that he would do so only with the KJ.

We can infer that partner holds about 9-10 hcp, maybe 11 if opener stretched as some openers are known to do on occasion. We can infer that partner has placed dummy with at least 4 hcp.

We can infer that partner has chosen his lead with the intention of trying to beat the contract, rather than holding the overtricks, and that to do so means that he is hoping we will hold 5-6 hcp or so.

He knows that dummy has 9 major suit cards, and that while he expects us to have spade length, he won't lead away from a short honour there....and since he is short in spades, rates anyway to have his honours in the minors. Given his likely weakness in spades, he will be concerned that given enough time, declarer can establish 10 winners, so he will tend to be aggressive, but not suicidal.

In my view, leading away from a K is often superior to leading away from a Q. For one thing, if we hit the A in partner's hand, we score our 2 tricks quickly. For another, when declarer is missing the AQ, he will often have a guess and may guess wrong, while leading the suit removes the guess. Picture our declarer with KJxx in diamonds and, say, no ability to pitch losers in dummy on club winners.

Thus far, this merely suggests being willing to lead from a King. However, it makes more sense to lead from a long suit than a short suit. When we hold Kxxxx, assuming some internal texture, we rate to lose at most one trick from a bad lead. When we hold Kxx, otoh, say we catch declarer with AQ10x. Now he can get an extra winner, beyond the free finesse at trick one, by ruffing the suit once. So Kxxxx and Kxx....all else being equal, the long suit should be led. While I am not for a moment suggesting that my earlier construction resembles the hand, the above thinking does underlie the thinking behind the construction.

[Fluffy]

a club, partner has 3 kings, lets just hope one of them is clubs.

[whereagles]

mikeh, on 2014-May-13, 14:46, said:

I don't buy into the notion that partner won't have a diamond honour: indeed, I think it would be odd to lead low from xxxxx.

I don't want to get into an arguing with you, but recent simulations (Bird/Anthias) are questioning some old leads concepts. In particular there are hints that, in abstract, a lead from xxxxx is more attractive than from Hxxxx. This situation isn't really abstract (we know opener is balanced and responder has 5-4), but even then I would guess that a sim would not prefer the Hxxxx lead.

From the hand you shown, xx Kxx Kxxxx Kxx, the sim would probably call for a spade lead. Odd, I know, but I believe that would be the outcome.

By the way, sims also suggest underleading a K is considerably more risky than underleading a Q. In fact, the more your holding resembles xxx, the safer it gets. An old book by Borel/Cheron (from the 1940s) even calculated odds, but for some reason this ancient knowledge went forgotten until recently. Quite aggravating in fact, as Borel was one of the 20th century greatest mathematicians lol.

[Phil]

Club and I honestly wouldn't consider anything else except for the fact AWM posted this.

[mikeh]

whereagles, on 2014-May-14, 02:46, said:

I don't want to get into an arguing with you, but recent simulations (Bird/Anthias) are questioning some old leads concepts. In particular there are hints that, in abstract, a lead from xxxxx is more attractive than from Hxxxx. This situation isn't really abstract (we know opener is balanced and responder has 5-4), but even then I would guess that a sim would not prefer the Hxxxx lead.

From the hand you shown, xx Kxx Kxxxx Kxx, the sim would probably call for a spade lead. Odd, I know, but I believe that would be the outcome.

By the way, sims also suggest underleading a K is considerably more risky than underleading a Q. In fact, the more your holding resembles xxx, the safer it gets. An old book by Borel/Cheron (from the 1940s) even calculated odds, but for some reason this ancient knowledge went forgotten until recently. Quite aggravating in fact, as Borel was one of the 20th century greatest mathematicians lol.

I hope this is more of a debate than an argument.

I agree that there is risk underleading a K rather than xxx, and even with the K as opposed to Q, tho I suspect the difference is minimal in the latter case. However, to focus only on risk is not a profitable way to think at the table. Heck take it to an extreme and we see we should pass all balanced 25 counts in 1st seat....if we open, we may end up with a minus score The question must encompass an assessment of the benefits that the lead may generate as well as the risk.

As for Bird/Anthias, I confess I haven't read the book but iirc it was heavily criticized for relying on double dummy simulations. I have often expressed my reservations about the utility of double dummy simulations, and I am far from alone in this regard. I really don't think that they are persuasive methods for this purpose. I may be wrong there, of course, but I seem to have some pretty good company.

[neilkaz]

Phil, on 2014-May-14, 04:39, said:

Club and I honestly wouldn't consider anything else except for the fact AWM posted this.

LOL and we agree.

[whereagles]

@mikeh: it's true that double dummy probably introduces a bias. Whether or not this is enough to change sim opening leads rank is another thing. My gut feeling is that in most cases it is not.

By the way, if I recall correctly from Borel/Cheron, the odds for blowing a trick are something like this: (1 suit viewpoint only)

Ace, unsupported lead: ~8%
Ace underlead: ~15%
King underlead: ~30%
Queen underlead: ~20%
Jack underlead: ~5%

I might be a few percents off, but it's more or less this. I can check the exact figures if you want (got the book here somewhere).

What I find interesting is that Bird/Anthias sims actually agree with these figures from before the age of calculators.

[mikeh]

whereagles, on 2014-May-14, 14:44, said:

@mikeh: it's true that double dummy probably introduces a bias. Whether or not this is enough to change sim opening leads rank is another thing. My gut feeling is that in most cases it is not.

By the way, if I recall correctly from Borel/Cheron, the odds for blowing a trick are something like this: (1 suit viewpoint only)

Ace, unsupported lead: ~8%
Ace underlead: ~15%
King underlead: ~30%
Queen underlead: ~20%
Jack underlead: ~5%

I might be a few percents off, but it's more or less this. I can check the exact figures if you want (got the book here somewhere).

What I find interesting is that Bird/Anthias sims actually agree with these figures from before the age of calculators.

I am arguing from a position of ignorance in terms of how B/A and B/C did their analyses, but my take is that their simulations were naïve to the bidding. IOW, whether looking at a single suit or all 52 cards, they simply worked out the risks associated with various leads, given the opening leader's holding.

In real life, of course, we listen to the auction. As Reese wrote, before bidding boxes, there is no such thing as a blind lead, only a deaf leader.

If RHO opened 1♥, promising 5+, and ended up in 3N, I wouldn't consider leading a heart from, say, Kxx or Kxxx. However, if the opps bid all the other suits and RHO reluctantly stumbled into 3N, I would give serious thought to a heart lead. I am morally certain that double dummy analysis would show that leading a heart in the first example was less effective than on the second.

It may be that on some auctions, a particular lead is incredibly bad, and that the auction would warn any bridge player from making the lead. However, the analysis would include those hands, and those leads that no-one would ever make, in the evaluation of the lead. This would distort the analysis.

Indeed, my thinking is that any analysis of opening leads that ignores the inferences available from the auction has no utility in bridge. It is maybe of considerable use in whist

I don't think it plausible to do any detailed, reliable analysis incorporating bidding. There are just too many systems, too many variants and styles within systems, and too many auctions for that to be possible. Consider a current thread here about advancing a t.o. double of 1♥ with a 3=3=4=3, and Jxx in hearts. Some happily bid 1N, others say one shouldn't do it without a stopper. How does one simulate leading from AQxxx in hearts? If RHO has a stopper, it is probably a very good lead: if the suit is Jxx xx Kxx around the table and partner has an entry, happy days lie ahead. If the suit is Kxx xx Jxx around the table, leading low blows a trick in the suit and may or may not blow the defence, depending on the rest of the hand. How does one simulate that? Note that making it AQ10xx changes but doesn't simplify the situation.

So to me, these double dummy analyses are maybe of some intellectual interest to those so inclined but they don't help us at the table. In particular, knowing that underleading a Q is on average less expensive than underleading a K is of no value to anyone. What I want to know is whether given all 13 cards I hold, and all of the bidding, and what the bidding means, it is better to underlead a Q in one suit or a K in another. I don't give a damn about how that decision plays out on other hands, with different auctions.

[Mbodell]

mikeh, on 2014-May-14, 16:36, said:

I am arguing from a position of ignorance in terms of how B/A and B/C did their analyses, but my take is that their simulations were naïve to the bidding. IOW, whether looking at a single suit or all 52 cards, they simply worked out the risks associated with various leads, given the opening leader's holding.

B/A were not naive to the bidding. They take the bidding very seriously and that is a large part of the point of their book. On an auction like 1nt-2nt-3nt a lot of somewhat dubious holdings in a major are better than more "standard" good leads from minors. Although it also depends on if you or partner expect to have the points. Obviously it also depends on the assumptions about the bidding (some people suppress 4 card majors, especially if 4333; some people open 1nt with a 5M, some don't).

Edited to add: that doesn't mean I'm saying that they are right or that there aren't other flaws in their methods. Just not the ultra naive bidding flaw.

[rhm]

Partner has between 9-11 HCP.
To beat this contract this must translate into 3 tricks. Let us look at some possible honor combinations for partner, which might accomplish our goal:

1) partner has 3 kings

If partner has the ♣K a club switch is required, since dummy's club is likely to go away on a diamond eventually.
If partner does not have ♣K (nor the ♣Q), a club switch is likely to be fatal. However chances for that are around 25% since the missing king could be in any suit.

2) Partner has 2 kings and an ace.

Again it is unlikely that partner has AK in a minor suit, because he might have led it. He could have AK in a major suit and the ♦K, but most would still prefer to have a look at dummy and lead from AK before switching to a highly risky diamond.
If partner has his honors all in different suits chances that he has nothing in clubs remains close to 25%.

3) Partner could have ♣AQ and any king

Obviously a club switch is necessary.

All in all it seems much more likely that partner has the ace or king in clubs and a club switch is necessary, than partner having nothing in clubs and the ♦K where a club switch would be fatal.
To see this you do not need complex mathematics or double dummy simulations.

Rainer Herrmann

[whereagles]

@mikeh: the figures of Borel/Cheron are very theoretical. They refer to suit contracts, where the opening lead is made from a side suit and there's no shortage of trumps for declarer. No bidding involved and results revolve around the side suit alone. As for Bird/Anthias, as Mbodell said, bidding is present and a full hand is generated and played, compliant with bidding constraints. The striking similarity between B/C and B/A figures is what amazes me the most, especially considering the 60+ years gap between the two.

[whereagles]

rhm, on 2014-May-15, 03:17, said:

To see this you do not need complex mathematics or double dummy simulations.

What? You mean I can't bring my laptop to the table??

[Zelandakh]

whereagles, on 2014-May-14, 14:44, said:

I might be a few percents off, but it's more or less this. I can check the exact figures if you want (got the book here somewhere).

I found an old BW post that has what I assume are accurate figures:-

Axx (x less than 8)
Lead x: 10.8 tricks lost per 100 hands
Lead A: 5.6 tricks

Axxx…………….x……………..13.3 tricks
Axxx…………….A………………4.6 tricks

Kxx (x less than 8)
Lead x: 11.3 tricks

Qxx (x less than 7)
Lead x: 7.0 tricks

KQ7
Lead K: 13.6 tricks

KQ10
Lead K: 7.7 tricks

No results for Jxx(x) and xxx(x).

I do not have the book though, so might be worth checking these.