Computer Says No

[lamford]

[hv=pc=n&w=sk32hq32dj432ca32&d=s&v=b&b=7&a=1n(12-14)p3nppp]133|200[/hv]

A run-of-the-mill auction, and run-of-the-mill hand.

 

Computers are able to help us a lot now. In chess, backgammon and go, they are already better than us. Assume South is 4432, 5332, or 4333, with 12-14. Assume North has 12-16 without a five-card major, and also no singleton but can have a six-card minor but no other assumptions.

 

What does the computer say is the best lead to defeat the contract at IMPs? And what is the best lead at matchpoints?

[mikeh]

I have no idea what the computer (the software) says but at imps I lead a low spade and at mps a low heart.

 

I need less in spades to beat 3N but a lead away from Kxx on this auction seems more likely to blow a trick than a lead away from a queen

 

No way am I leading a minor on this hand and auction.

[DavidKok]

At matchpoints I would lead a low diamond without question. At IMPs I would be tempted by the alternatives, in particular a small heart, but I would probably still stick with the diamond. I also have no idea what the computer will say except that the game is a huge favourite to make.

[AL78]

I'd be tempted with a heart lead at imps or maybe a spade lead. The opponents haven't shown any interest in the majors, so it is not asking for the world that partner has something useful in one of them.

 

At MPS, I have no idea, it is the hand of death where any lead could blow a trick, there is probably one good lead and two bad ones. Can't see it is anything more than a guess, although I'd likely lead a major for the reason given above.

[Douglas43]

Small question but why are the criteria only that North not have a 5 card major? It's rare to skip stayman with a 4cM.

[mikeh]

Let me add a comment about computers.

 

Bird and Anthias wrote two popular books analyzing opening leads, one against notrump and the other against suits. I have both and they make for interesting reading, and in one of my two partnerships we apply at least some of the lessons from the notrump book....including a spectacular result from the 2019 GNT playoffs in Las Vegas, where after 1N 3N my partner had a weak hand with 2=2 majors, and the only card he had of any use was Kx in hearts.

 

He led the King, which pleased me immensely. I was vulnerable at imps and had passed 3N with AJxx A1098xxx xx. Dummy had QJ tight in hearts!

 

Declarer called the TD once he saw that I had 7 hearts, claiming that I had taken so long over 3N that it conveyed UI to partner. Fortunately, one side effect of the cheating scandals was that all playoff matches were on video, so the TD was able to confirm that, while I had hesitated over the jump to 3N, my hesitation was the prescribed 10 seconds (I always hesitate in these auctions, and commend that practice to everyone...I did have a problem on that hand, and were I in the habit of passing quickly with no problem, then the declarer's point would have merit)

 

As it is, partner had just finished reading the book on notrump, where this would undoubtedly have been identified as the lead most likely to beat 3N (although usually the contract is cold and the lead will rate to cost a trick more often than to gain).

 

Back to my point: the books are based on double dummy analysis, and that's where they have problems. We don't defend double-dummy.

 

In particular the analyses assume that both defenders will 'work out' what it going on. So if one leads low from a 3 card holding, playing 4th best, partner will 'know' that you only have 3 and so will defend accordingly...double dummy. And if the best lead is, say, the Ace, partner will play you for the holding you have rather than the holding that your leading methods say the Ace shows, and so on.

 

In other words, as far as I can tell, computers don't have 'agreements' with partner as to what leads mean...every time it leads, it runs simulations without regard to agreements, which is why (I think) GIB is so hopeless on defence.

 

I wouldn't cede bridge superiority to computers yet:) But I do think that the B&A books are very interesting, provided one is alive to the double dummy issues.

[smerriman]

I'd expect the computer to find a low diamond a clear favorite at MPs.

 

At IMPs, probably still a diamond, but it's a close call between that and the club ace.

 

But as Mike mentioned, this is the best lead under the assumption that your partner can see your hand after you lead it, not necessarily the best lead.

[pilowsky]

I defer to Mikeh on this, obviously, but having just read the Bird Anthias books it seems that the same logic applies to a weak NT opening as a strong NT opening.

The B&A simulations were based on the best outcomes with simulations after 15-17 NT openings.

 

Still, in both cases, there is an implication that the Declarer will be looking to make tricks in minor, not major suits.

 

They suggest that after an Auction of this type (1NT-3NT) the Declaring side will have 'on average' only 5.8 major suit cards between them BUT defenders will have 7.2.

This means that their analysis favours a major suit lead since that is the likeliest source of tricks for the defenders.

[Cyberyeti]

Whenever my partner or I lead anything other than a diamond, it feels like the hands are 3325/4423 and partner has enough in the diamond suit to ensure that it goes off if they're led while it makes with overtricks if you lead anything else. We have a bad record trying to get clever on this type of hand.

[helene_t]

Our hand is quite strong, partner has a maximum of 6 HCPs and his expectation is of course less than that. That means that even if we hit a long and strong suit of his, declarer can still hold it up if he has Axx, as partner is unlikely to have an entry.

 

So maybe we should aim at establishing our diamonds, but that's a bit of a fat chance.

 

But the strength of our hand also suggests that their game is tight and a passive lead may be called for. But I don't know which lead is the most passive. Maybe even a club.

[nige1]

[hv=pc=n&w=sk32hq32dj432ca32&d=s&v=b&b=7&a=1n(12-14)p3nppp]133|200|Lamford 'A run-of-the-mill auction, and run-of-the-mill hand.Computers are able to help us a lot now. In chess, backgammon and go, they are already better than us. Assume South is 4432, 5332, or 4333, with 12-14. Assume North has 12-16 without a five-card major, and also no singleton but can have a six-card minor but no other assumptions.What does the computer say is the best lead to defeat the contract at IMPs? And what is the best lead at matchpoints?'

++++++++++++++++++++

I rank

1. ♦x -- 4th highest -- Granny would approve :)

2. ♥x -- Majors are always popular :)

3. ♠x -- But not if you're ever lucky enough to partner Tony Forrester :(

4. [CL}A -- Computer's favourite -- because, at trick 2, it finds the DD switch :)[/hv]

[mcphee]

in years of playing I have fund that leading from J empty 4th rarely wins and I hate it. I have preferred what I call the stand bye lead, hearts against NT when other options look dismal and enjoyed good things more often than not.

[Tramticket]

(1) The results of a double dummy simulation using Lead Captain and Lamford's exact parameters (Sample Size 10,000):

 

- At IMPs (Set %): 2♦ 10.35%, A♣ 10.06%, 2♥ 9.79%, 2♣ 9.64%, 2♠ 9.02%, Other honours < 8.65%

- At MPs (Number of Tricks): 2♦ 3.11, 2♣ 3.05, A♣ 3.01, 2♠ 2.99, 2♥ 2.99, Other honours < 2.93

 

(2) The opponents might use Stayman. Here is a Lead Captain simulation using Lamford's parameters, but only allowing a four-card major if exactly 4333:

 

- At IMPs (Set %): 2♥ 9.96%, 2♠ 9.80%, 2♦ 9.13%, Q♥ 8.90%, A♣ 8.75%, 2♣ 8.20%, Other honours < 7.84%

- At MPs (Number of Tricks): 2♦ 3.07, 2♠ 3.02, 2♣ 3.02, 2♥ 3.01, A♣ 2.99, Other honours < 2.94

 

(3) General Observations:

 

We are not defeating this very often (about 10%) and the opponents expect to make about 10 tricks. The results are very close which is unsurprising since no lead looks attractive. A diamond lead is most passive and does slightly better at MPs. As Pilowsky points out, our partnership is likely to hold more cards in the majors (particularly if we allow Stayman) and a major suit lead is more likely to defeat the contract.

 

(4) Some comments on the Double Dummy simulation

 

Firstly, DD simulations tend to favour ace leads, because the simulation will always make the optimum trick 2 switch. The A♣ lead didn't perform particularly well here because we will usually want to play passively rather than find an optimum switch. But I would always discount results for aces in DD simulations.

 

DD simulations always find the queen. In real life leading from ♥Q32 is likely to be more dangerous because humans make wrong guesses looking for the queen. For this reason, I would lead a low spade at IMPs.

[lamford]

(2) The opponents might use Stayman. Here is a Lead Captain simulation using Lamford's parameters, but only allowing a four-card major if exactly 4333:

 

- At IMPs (Set %): 2♥ 9.96%, 2♠ 9.80%, 2♦ 9.13%, Q♥ 8.90%, A♣ 8.75%, 2♣ 8.20%, Other honours < 7.84%

- At MPs (Number of Tricks): 2♦ 3.07, 2♠ 3.02, 2♣ 3.02, 2♥ 3.01, A♣ 2.99, Other honours < 2.94

 

Yes, my thoughts pretty much. The diamond lead is best at matchpoints. I used different software, and had the heart slightly ahead at IMPs. If you think North won't have a four-card major, then it was ahead, which suggests North shouldn't be bidding Stayman with KJTx of hearts!

[mikeh]

Yes, my thoughts pretty much. The diamond lead is best at matchpoints. I used different software, and had the heart slightly ahead at IMPs. If you think North won't have a four-card major, then it was ahead, which suggests North shouldn't be bidding Stayman with KJTx of hearts!

I suspect the reason dd analysis says a diamond is the best at match points is that it gives away the least, on a dd basis. IOW, given that declarer is playing literally as if it can see all 52 cards, it will pick off your Jxxx whenever it is possible. In real life, this is unlikely unless you lead it.

 

Picture dummy with Q10xx and declarer AKx

 

Unless there is some good bridge reason for finessing, most declarer’s will/should play for the drop, but never double dummy. In the real world, a diamond blows the suit. Of course, any lead could blow a trick but not as frequently as from Jxxx.

 

IMO, this just goes to show how useless dd lead analyses are, if put forward as ‘the’ answer.

 

There was a study, admittedly many years ago, of leads, taken from very high level competition. As I recall, the worst lead from length v notrump was Jxxx where the highest x was smaller than a 9. Apparently J9xx was quite a bit more successful, which makes sense to me.

[lamford]

I suspect the reason dd analysis says a diamond is the best at match points is that it gives away the least, on a dd basis. IOW, given that declarer is playing literally as if it can see all 52 cards, it will pick off your Jxxx whenever it is possible. In real life, this is unlikely unless you lead it.

 

Picture dummy with Q10xx and declarer AKx

 

Unless there is some good bridge reason for finessing, most declarer’s will/should play for the drop, but never double dummy. In the real world, a diamond blows the suit. Of course, any lead could blow a trick but not as frequently as from Jxxx.

 

IMO, this just goes to show how useless dd lead analyses are, if put forward as ‘the’ answer.

 

There was a study, admittedly many years ago, of leads, taken from very high level competition. As I recall, the worst lead from length v notrump was Jxxx where the highest x was smaller than a 9. Apparently J9xx was quite a bit more successful, which makes sense to me.

Future generations of lead analysis programs will have both sides playing perfect "single dummy", although inferences from the auction will be difficult to program accurately.