Lead testing request for simmers

[Jinksy]

After the auction 1♠* P 1N** P P*** P, would someone mind testing the leads from this hand:

 

AJT KT4 K976 654

 

* 5+♠s

** 6-10 HCP

*** may include a four-card minor, may not include 4+♥s. May include a 6-card ♠ suit with only 1 of AKQJ unless also 15+ HCP.

 

This hand is very slightly modified from a Bird and Anthias one, which is being annoyingly inconsistent with the algorithm I'm trying to develop based on their data - I need to test a hypothesis.

 

Any bets on how it will work out, while we wait? I'm hoping for a marginal win for a small ♦ over a small ♣.

 

Thanks,

 

J

[lamford]

Does anyone know how to get a copy of this lead-testing program? I am happy to pay for it!

[Cthulhu D]

It is free: http://bridge.thomasoandrews.com/deal/downloading.html

 

It just requires some scripting skills.

[gszes]

I will place my wager on a small dia being over the 5% better mark above a

club just bear in mind I have never cashed a lottery ticket sigh.

[Siegmund]

You've not provided enough information about the system to properly constrain either opener's or responder's hand.

 

For instance:

Does responder have any other ways to show a weakish hand with a 6- or 7-card minor (intermediate jump shifts, and 3-level weak jump shifts, both intrude on "6-10".)

Does responder ever have 3 spades on this auction?

Does opener open 1S or 1NT on 5(332) hands? If 1NT, what notrump range?

You say "may contain a 4-card minor" ... but I am sure you don't mean all hands with a 4-card minor. Only 5224 and 5242? Only if the 4-card suit is weak?

 

I did take a first stab at it using simple (too simple) conditions. 1000 hands, which makes the percentages good to ±3%.

 

Responder: 6-10, 0-2 spades, 0-6 hearts, 0-6 diamonds, 0-6 clubs.

Opener: 12-16, 5(332), 5242, or 5224.

 

Lead, P(par possible after this lead), P(1NT can be set after this lead), avg. # tricks worst than best double-dummy lead. Didn't report average IMP loss because I accidentally left my script calculating IMPs in 3NT rather than 1NT.

 

♠A 20% 18% 1.30

♠J 52% 31% 0.64

♥K 50% 29% 0.83

♥T 55% 33% 0.60

♥4 57% 34% 0.56

♦K 45% 28% 0.90

♦9 60% 35% 0.50

♦7 63% 36% 0.46

♣6 67% 35% 0.42

 

Repeated with opener always 5(332), but still allowing 12-16. We expect this situation to show a stronger contrast, because it magnifies the importance of our fourth diamond -- a diamond lead never is into a 4-card suit -- and a "passive" club never runs into a running 4-card side suit.

 

♠A 18% 18% 1.32

♠J 50$ 33% 0.66

♥K 43% 25% 0.96

♥T 51% 29% 0.69

♥4 54% 32% 0.62

♦K 47% 27% 0.90

♦9 62% 36% 0.50

♦7 66% 37% 0.45

♣6 72% 38% 0.36

 

Sorry, but the club wins, at least under these conditions. I would have led a diamond at the table, too. Apparently I am strong enough that I should expect my partner not to be much help and I need to seek a passive lead.

[Jinksy]

Thanks Siegmund. I took the details from the book, so can't add much. They say 'opener would not rebid a 4-card minor suit, preferring to pass 1N unless he was strong enough to see a chance at game'! Sounds odd to me too, but I'd guess they're including at least 51(34s) and 53(14)s in the pass, unless say 15+HCP.

 

If you're feeling really industrious, their results might help you calibrate:

 

With a hand identical to the one posted, except with T96 in ♣s, they had

 

J♠ 27.9% 5.65

4♥ 37.8% 6.00

6♦ 40.5% 6.15

T♣ 42.0% 6.21

 

Where the first stat is equivalent to your second %age (1N can be beaten after this lead) and the second is average number of tricks possible to take after this lead.

[Cthulhu D]

I will place my wager on a small dia being over the 5% better mark above a

club just bear in mind I have never cashed a lottery ticket sigh.

 

I'd note that the fact that it's a K and not a Q is probable reasonably sized black mark against the lead - P is much more likely to hold the K or J vs the A or Q.

[Siegmund]

Well, that's one more thing I can add to my list of reasons not to like the book.

[Jinksy]

I'd note that the fact that it's a K and not a Q is probable reasonably sized black mark against the lead - P is much more likely to hold the K or J vs the A or Q.

 

Qxxx leads certainly work out much better, but Kxxx seems to come out better than you'd expect against 1N (though I don't think they sim any hands where it's the top lead).

 

Keep in mind, leading from Kxxx towards nothing won't necessarily cost if declarer still has to play the suit later, and leading from Qxxx towards P's J can still carve the suit.

[gszes]

Stupid question time why was the dia 6 (4th best) not included in the parameters.

and is it possible to include a twitch in declarer's LOP taking into consideration

the possibility of the opening leader having 5 diamonds?

 

I ask this because I could not use the program (possibly due to old dumpy computers

and operating systems ummm make that probably hmmm ok definitely). I mention this

because all of these hands where the computer knows everything are not anywhere close

to real life simulation where so much is unknown.

[the_clown]

I would lead a diamond. It looks like it wont be hard for them to set up spades quickly so I would try to be aggressive. If I were sitting over the 1♠ bidder I may well lead ♣

[Siegmund]

Stupid question time why was the dia 6 (4th best) not included in the parameters.

 

Every possible lead is considered, but D7 and D6 are equals, as are C654 and SJT. There are no signals to give partner in double-dummy analysis.

 

The program will run under DOS on an ancient machine. The learning curve can be steep, however.

[1eyedjack]

It is free: http://bridge.thomas...ownloading.html

 

It just requires some scripting skills.

 

Here is a more simple tool. Does not require scripting, but probably a bit less powerful as a result. I prefer it as I am not and don't need to pretend to be a programmer.

 

it does cost $25, mind.

 

http://www.bridge-captain.com/LeadCaptain.html

 

 

 

 

 

[avonw]

Also: http://www.lajollabridge.com/Software/Lead-Solver/Lead-Solver-About.htm

 

Free. Recommended.

[Cthulhu D]

Also: http://www.lajollabridge.com/Software/Lead-Solver/Lead-Solver-About.htm

 

Free. Recommended.

 

That amazingly simplifies many, many things.

 

This problem is really cool, slapped this together in minutes:

 

                        Frequency of Tricks Taken
Ld   Avg  %Set      0   1   2   3   4   5   6   7   8   9  10  11  12  13
C6  6.18  40.45* [  0   2  10  53 176 386 564 453 240  98  18   0   0   0 ]
D7  6.15  38.70  [  0   1  12  41 175 407 590 437 238  82  17   0   0   0 ]
D9  6.06  36.20  [  0   1  13  49 190 428 595 421 213  75  15   0   0   0 ]
H4  5.95  35.70  [  0   9  17  80 234 416 530 402 223  73  16   0   0   0 ]
HT  5.86  33.60  [  0   9  17  80 264 438 520 401 188  69  14   0   0   0 ]
SA  5.72  28.90  [  0   5  30 105 257 500 525 350 139  71  18   0   0   0 ]
SJ  5.72  28.60  [  1   5  20 101 273 487 541 352 140  66  14   0   0   0 ]
DK  5.69  30.40  [  1   6  29 107 302 454 493 375 169  55   9   0   0   0 ]
HK  5.53  28.70  [  1  19  41 152 342 399 472 348 166  50  10   0   0   0 ]

[WesleyC]

A couple of years ago I did a bunch of DD lead simulations. Many of my simulations showed a strong bias towards passive leads (especially at NT), however I'm very skeptical that they work as well in real life.

 

Although it's hard to quantify this phenomenon some strong logical justifications are:

 

1) Single dummy: A passive lead can solve a suit for declarer that would otherwise be a guess.

Double dummy: Declarer was never going to get the guess wrong.

 

2) Single dummy: Breaking a suit unfavourably can lead to a situation where declarer still needs a correct guess to gain a trick.

Double dummy: Declarer will always punish an unfavourable lead as much as possible.

 

3) Single dummy: Passive leads are sometimes difficult for partner to read and leave more room for defensive errors later in the hand. This is especially true in auctions where declarer's shape is mostly unknown. When it's close, leading your longest, strongest suit is a powerful signal.

Double dummy: Partner will always defend accurately, so no need to signal.

 

So although I have no mathematical justification, I'm pretty sure at the table a diamond lead would be a clear winner.

 

WesC

[rhm]

I simulated this with dealmaster assuming matchpoint conditions:

 

Result (1000 deals)

 

♠A(426) ♠J(428) ♠T(428) ♥K(419) ♥T(492) ♥(575) ♦K(403) ♦9(599) ♦7(665) ♦(665) ♣6(719) ♣(719) ♣4(719)

 

The number in brackets show on how many deals out of 1000 this lead will give the defense the chance of getting the maximum number of tricks available to them.

As can be seen the club lead is clearly best.

I do not share the prejudice that passive leads fare better double dummy than single dummy.

 

Rainer Herrmann

[gnasher]

I do not share the prejudice that passive leads fare better double dummy than single dummy.

 

You may disagree with WesleyC's views, but it seems unreasonable to describe them as "prejuduce", which my dictionary defines as "Preconceived opinion that is not based on reason or actual experience".

 

He describes obtaining evidence, presents his reasoning, and implies that his opinion is partly based upon his own experience. That is almost the exact opposite of prejudice.

 

FWIW, I share his doubts about using double-dummy analysis to model real-life bridge, but I'd lead a club anyway. A diamond might give declarer the extra entry he needs to lead up to dummy's spades twice, and when it works it's unlikely to set up enough tricks to beat the contract. A club is less likely to give something away and more likely to hit a five-card suit in partner's hand.

[WesleyC]

FWIW my 'prejudice' was based on manual analysis of something like 100 deals where a DD simulation had a passive lead significantly ahead of an active lead on pretty random hand after the opps bid 1NT - 3NT.

 

Maybe if R. Herrmann has a couple of hours free sometime he could do the same and report findings?

 

WesC

[rhm]

You may disagree with WesleyC's views, but it seems unreasonable to describe them as "prejuduce", which my dictionary defines as "Preconceived opinion that is not based on reason or actual experience".

 

He describes obtaining evidence, presents his reasoning, and implies that his opinion is partly based upon his own experience. That is almost the exact opposite of prejudice.

 

FWIW, I share his doubts about using double-dummy analysis to model real-life bridge, but I'd lead a club anyway. A diamond might give declarer the extra entry he needs to lead up to dummy's spades twice, and when it works it's unlikely to set up enough tricks to beat the contract. A club is less likely to give something away and more likely to hit a five-card suit in partner's hand.

We are running in circles with regard to the validity of simulations.

 

But let me say, if someone has a prejudice it does not mean such a person does not reason or has no arguments.

Almost always the opposite is true.

 

The cause and problem with prejudice is a different one.

We tend to believe we form our opinion by reasoning when in fact it is usually the other way round.

We first form our opinion and then our brain looks for arguments and reasons, which can justify our opinion.

That's the way our mind works, not the other way round.

 

I wanted to avoid that, but let's go (again) through Wesley's arguments:

 

A couple of years ago I did a bunch of DD lead simulations. Many of my simulations showed a strong bias towards passive leads (especially at NT), however I'm very skeptical that they work as well in real life.

 

Although it's hard to quantify this phenomenon some strong logical justifications are:

 

1) Single dummy: A passive lead can solve a suit for declarer that would otherwise be a guess.

Double dummy: Declarer was never going to get the guess wrong.

Two-way finesses exist but are not that frequent. But do they exist predominantly only for passive leads?

Give me a break.

 

2) Single dummy: Breaking a suit unfavourably can lead to a situation where declarer still needs a correct guess to gain a trick.

Double dummy: Declarer will always punish an unfavourable lead as much as possible.

Same argument as before. Applies to passive as well as aggressive leads.

 

3) Single dummy: Passive leads are sometimes difficult for partner to read and leave more room for defensive errors later in the hand. This is especially true in auctions where declarer's shape is mostly unknown. When it's close, leading your longest, strongest suit is a powerful signal.

Double dummy: Partner will always defend accurately, so no need to signal.

That last sentence is again true for passive as for active leads.

 

With regard to signalling:

Whenever you change your style it initially creates new problems and uncertainties until you get familiar with all the ramifications.

But I readily admit:

If you should lead more often passive than what common wisdom suggests, maybe your signaling style should accommodate to that.

In my humble opinion fourth best leads help declarer more and more often than partner.

Attitude leads are ideal if you do not always lead from longest and strongest.

(Note: I am not claiming you should always lead passive, only more frequently than common wisdom)

Attitude leads are quite easy to interpret for partner and tend to give more headaches to declarer. He does not really know how the suit led breaks.

I know from experience.

 

So although I have no mathematical justification...,

Neither do I.

 

I'm pretty sure at the table a diamond lead would be a clear winner.

My experience at the table is quite different.

 

Rainer Herrmann

[awm]

One might be able to look at hands where two tables had the same auction but different opening leads, and compare the results from the leads which are "better double dummy given the auction" versus the leads which are "worse double dummy given the auction" and see how they do. If WesleyC is right, we'd expect to see the results from real play looking closer than the double-dummy comparison (not clear that the "worse" leads would turn out better, since occasionally the worse leads might be simply bad, but we'd expect the results to be closer).

[Cthulhu D]

One might be able to look at hands where two tables had the same auction but different opening leads, and compare the results from the leads which are "better double dummy given the auction" versus the leads which are "worse double dummy given the auction" and see how they do. If WesleyC is right, we'd expect to see the results from real play looking closer than the double-dummy comparison (not clear that the "worse" leads would turn out better, since occasionally the worse leads might be simply bad, but we'd expect the results to be closer).

 

It is relatively easily solvable - pick the single dummy bridge engine of your choice, and run the same test(s). If I could figure out how to do it I would :D (If anyone can tell me how to use a free single dummy solver to play the cards from a PBN file with a specified opening lead, I will do it)

 

Edit: It looks like it is only possible with GiB