Lead verus 3N

[hrothgar]

IMPS

Red v Red

 

Auction starts

 

P - (1N) - P - (3N)

All pass

 

You hold

 

♠ J9864

♥ JT2

♦ 3

♣ J842

 

Whats your lead?

Do you consider it close?

[Guest Jlall]

spade, not close at all.

[mike777]

spade, not close at all.

ditto.

[Cascade]

There is a huge premium on leading a major on this auction.

 

And a huge premium on leading a five-card suit.

 

I have a 5-card major do I really need to look elsewhere?

[Cascade]

If I did look elsewhere

 

♥ J10x is a fine lead.

[effervesce]

I like a heart lead. I don't have an entry so I want to try and hit partner's suit.

[jtfanclub]

I like a heart lead. I don't have an entry so I want to try and hit partner's suit.

Surely if you want to hit partner's suit a diamond stands out, no?

 

I have jacks under the strong hand. I have lots of hope that they will develop into tricks later. My spades aren't that bad. A spade for me.

[skjaeran]

A spade is pretty obvious at IMPs.

 

At MP, a ♥ lead might be more interesting. I'd still lead a spade though.

[TimG]

I think this would be an interesting hand for a double dummy lead analysis. I don't think Dealer (what I use for double dummy sims) does anything with leads. Please tell me if I am wrong about that. What do people use for their double dummy lead sims?

[jdonn]

I'm actually a bit torn between heart and spade. With Justin lead a spade for sure :)

[pclayton]

Funny, I'd lead a spade at MPs more than a heart. I don't think I want to try to swing on the OL here.

 

At IMPs its much tougher. With ♥JTx, pard is unlikely to make a double, even with something like ♥KQxxx. The problem with a spade lead is we need to catch pard with 4, or three really good ones like ♠KQx (with declarer Ax). On other layouts, the spades will cash later (unless declarer has 9 runners). I think a heart lead requires less. All we need is a little length in pard's hand.

 

I've talked myself into a heart. Maybe I'm not as afraid of Justin as Josh is :P

[hrothgar]

I think this would be an interesting hand for a double dummy lead analysis. I don't think Dealer (what I use for double dummy sims) does anything with leads. Please tell me if I am wrong about that. What do people use for their double dummy lead sims?

Hi Tim

 

I agree that this is the sort of problem where a double dummy approach might be valuable.

 

From my perspective there are two different ways in which one might conduct a double dummy type analysis:

 

Option 1: Run the double dummy solver 10,000 times or so. Count how often the solver choses leads X, Y, or Z.

 

Option 2: Run the solver 10,000 times with a constraint that forces lead "X". Run the solver 10,000 times with a constraint the forces lead "Y". Check and see whether the distribution describing the number of tricks is statistically significant.

 

I'm not sure which method is better.

 

(For what its worth, I chose a Heart lead at the table which happened to work this time)

[TimG]

Option 1: Run the double dummy solver 10,000 times or so. Count how often the solver choses leads X, Y, or Z.

 

Option 2: Run the solver 10,000 times with a constraint that forces lead "X". Run the solver 10,000 times with a constraint the forces lead "Y". Check and see whether the distribution describing the number of tricks is statistically significant.

I suspect you can get statistical significance with fewer than 10,000 deals. Then again, computer time is cheap.

 

The double dummy solvers that I have used (Double Dummy Solver, Deep Finesse and GIB on BBO) all highlight the cards that will produce the double dummy result, so it seems to me that a table should be able to be produced with Option 1 which would list all 13 possible leads and how often each produced the double dummy result, how many tricks on average the lead results in and how often the lead beats the contract.

 

With Option 2, you'd want to use the same random seed for each trial, which really makes things identical to Option 1. Doesn't it?

[Sambolino]

once i've read the article about similar topic. after computer simulation over large number of deals the conclusion was - if you are weaker than p, lead shorter major vs 3nt. i think it was written by luis argerich

 

i've experienced numerous times when playing with weak players that their leads vs NT keep coming into my singletons when i'm the one who holds more cards. i find that annoying, so i try to make my pd happy by leading from short suit

[helene_t]

Option 1: Run the double dummy solver 10,000 times or so. Count how often the solver choses leads X, Y, or Z.

 

Option 2: Run the solver 10,000 times with a constraint that forces lead "X". Run the solver 10,000 times with a constraint the forces lead "Y". Check and see whether the distribution describing the number of tricks is statistically significant.

I suspect method 2 will be more efficient since it will not waste time on other leads than the ones we are considering. OTOH if one stands out by a mile it would be able to prune the other leads away quickly so I'm sure it's possible to construct a deal on which method 1 will be more efficient.

 

Another issue is that if the DD solver only makes one choice (and not highlights equivalent options) it may have a bias when the two are equivalent.

[han]

For what it is worth, I ran a double dummy analysis with the following specifications:

 

partner: at most 11 HCP.

 

dealer: 15-17 balanced.

 

dummy: 9-14 HCP, no shortness, no shortness, only a 4-card major if 4333.

 

Out of the 1000 hands that I dealt, 3NT could be beaten 116 times.

 

For each time that the contract go down the program gives a point to each lead that would give the defense the largest number of tricks. The different leads received the following number of points:

 

spade jack: 51

other spade: 60

heart honor: 62

low heart: 45

diamond: 31

club jack: 26

club 8: 36

low club: 37

 

I don't claim that these specifications are best or that a double dummy search is useful at all for lead problems.

[TimG]

For what it is worth, I ran a double dummy analysis...

 

For each time that the contract go down the program gives a point to each lead that would give the defense the largest number of tricks. for lead problems.

What program are you using?

[rogerclee]

For what it is worth, I ran a double dummy analysis...

 

For each time that the contract go down the program gives a point to each lead that would give the defense the largest number of tricks. for lead problems.

What program are you using?

Han just thinks very, very quickly.

[han]

dealmaster pro.

[hrothgar]

For what it is worth, I ran a double dummy analysis with the following specifications:

 

partner: at most 11 HCP.

 

dealer: 15-17 balanced.

 

dummy: 9-14 HCP, no shortness, no shortness, only a 4-card major if 4333.

 

Out of the 1000 hands that I dealt, 3NT could be beaten 116 times.

 

For each time that the contract go down the program gives a point to each lead that would give the defense the largest number of tricks. The different leads received the following number of points:

 

spade jack: 51

other spade: 60

heart honor: 62

low heart: 45

diamond: 31

club jack: 26

club 8: 36

low club: 37

 

I don't claim that these specifications are best or that a double dummy search is useful at all for lead problems.

Silly question:

 

I would expect the sum of the number of points to be less than or equal 116 (One point each time there was a "best" lead with some sort of tie breaking algorithm)

 

The total number of points actually sums to 348 (or precisely 3 x 116)

[Finch]

I understood it that if all leads concede 13 tricks (say) then all leads get one point i.e. no tie-breaking algorithm.

 

Of course the biggest problem with the analysis is that 'most number of tricks' is a long long way from 'best lead'. Can you not differentiate between 'beats the contract' and 'doesn't beat the contract' ?

 

(I've carefully missed out my rant on the benefits of DD vs single D analysis for once)

[TimG]

I would expect the sum of the number of points to be less than or equal 116 (One point each time there was a "best" lead with some sort of tie breaking algorithm)

If a heart and a spade both result in 5 tricks for the defense (and diamonds and clubs produce fewer tricks for the defense) then both the heart and the spade get a point.

 

I think it is just a coincidence that the sum is 3 x 116. The table could read:

 

spade jack: 51

spade nine: 60

spade eight: 60

spade six: 60

spade four: 60

 

etc.

 

and have the same meaning, if I understand correctly.

[han]

Silly question:

 

I would expect the sum of the number of points to be less than or equal 116 (One point each time there was a "best" lead with some sort of tie breaking algorithm)

This is not a question, this is a statement.

 

The total number of points actually sums to 348 (or precisely 3 x 116)

 

OK, not very interesting to my mind but I trust it does.

[han]

Yes what Tim said is correct.

[Fluffy]

So all in all it is close.

 

I would lead a spade and don't think any more, I hate looking silly when partner has 4+ spades.