[mikeh]

Cascade, on Mar 20 2008, 05:35 PM, said:

Texas is clear 3NT maybe horrid opposite two small and a bad break but we might still make 4♠ with a trump loser.

Justin is exactly right (as he often is) - why do we want to play 3NT opposite two small?

I'll do a simulation too.

OK, I'm up to 160 hands now And I think I will stop.

So far, both contracts are strong favourites to make opposite a 2 card spade holding (no surprise there). When one is better than the other, it is so far 13 for 3N making and 10 for 4♠.

So the gap has narrowed, but 160 hands is starting to resemble a statistically significant number (well, my knowledge of stats is based on undergraduate courses more than 30 years ago, so I may be overstating matters here... dangerous with so many mathematicians posting )

I should add that the figures are much different at BAM or mps.

On some hands where spades are 3-2, a declarer in 3N should be ducking a spade because he only needs 5 tricks in the suit, and is not at risk of going down. On those hands, at imps, he will often be taking fewer tricks than the 4♠ declarer because of the safety play, even tho both make.

Of course, when spades are 4-1, 3N will now frequently fail since the BAM/mp declarer is not going to want to take the safety play.

And even when we have Jx of spades and 6 spade winners, there are lots of hands on which 10 tricks are easy in spades and not in notrump... one obvious example is when declarer's diamonds are Jxx or so, and they can cash 4 diamonds in 3N and only 3 in spades, etc.

There are other ramifications/combinations so don't take this as an exhaustive listing of possible scenarios.

But, on the whole, at mps or BAM, committing to spades appears to be strongly best, while at imps, there may be a very slight edge to 3N as the safer contract (altho my guess is that the margin so far revealed isn't meaningful for any purpose than saying that it is close). But, the fact that the 4♠ bidders will often gain an imp over 3N cannot be ignored.

So I think, on reflection, that using Texas is best. But not by much, and this may be the ideal hand type on which to swing if outgunned by your opps. Bid 3N... you may lose an imp fairly often but you have the chance to generate (either way) a game swing without being silly.

[TimG]

One thing that I have not seen factored in is second hand's pass. Some of the time that 3NT is going down it is because a suit is wide open. Sometimes when opener's LHO holds a running suit, they will be able to act over 1NT. When I started looking just at hands that made in 3NT or 4S and failed in the other, the first hand that I saw was one where 2nd hand was 0715 with AK of hearts and KJ of clubs. Surely that hand was going to act over 1NT!

So, I think second hand passing is a factor in favor of 3NT. Not only are the chances of there being a wide open suit reduced, but when 4th hand holds the long, strong suit, opening leader won't always find it.

[dburn]

TimG, on Mar 20 2008, 07:06 PM, said:

So, I think second hand passing is a factor in favor of 3NT.  Not only are the chances of there being a wide open suit reduced, but when 4th hand holds the long, strong suit, opening leader won't always find it.

His chances of doing so are greatly increased if you bid hearts. 4th hand may double 2♥; if he does not, 2nd hand will lead a non-heart if holding a weak balanced hand and trying to find partner's suit.

It would never occur to me to bid other than an immediate 3NT, at any form of scoring. Declarer is given a trick on the lead against 3NT far more often than against 4♠, and even if he is given a trick on the lead against 4♠, it may well be his ninth rather than his tenth.

[Cascade]

My double dummy simulation suggested that 4♠ was better.

The chance of 3NT failing was 27.6% and 4♠ failing was 17.8. I am surprised that this is so different than what Mike is finding by looking at hands. My experience of comparing double dummy and single dummy is that the results usually work out reasonably close from each method.

I also extended the study by considering when would you bid 3NT with these spades. Is there any number of extra values that make 3NT better than 4♠. Surprisingly to me the double dummy simulation suggested not. Here are the results:

 [space] [space] [space]Responder's HCP
 [space] [space] [space] 9 [space]10 [space]11 [space]12 [space]13
3NT 276 208 149 [space]93 [space]46
4S [space]178 120 [space]67 [space]35 [space]17

The numbers represent the number of times in 1000 that each contract failed.

[rogerclee]

Cascade, on Mar 20 2008, 08:05 PM, said:

My double dummy simulation suggested that 4♠ was better.

The chance of 3NT failing was 27.6% and 4♠ failing was 17.8. I am surprised that this is so different than what Mike is finding by looking at hands. My experience of comparing double dummy and single dummy is that the results usually work out reasonably close from each method.

I also extended the study by considering when would you bid 3NT with these spades. Is there any number of extra values that make 3NT better than 4♠. Surprisingly to me the double dummy simulation suggested not. Here are the results:

 [space] [space] [space]Responder's HCP
 [space] [space] [space] 9 [space]10 [space]11 [space]12 [space]13
3NT 276 208 149 [space]93 [space]46
4S [space]178 120 [space]67 [space]35 [space]17

The numbers represent the number of times in 1000 that each contract failed.

I think it is not surprising that double dummy, 4♠ is better. One big factor of bidding 3NT is that we will get a bad lead very, very frequently. In fact, if I bid 3N, the one thing I do not really want led is spades.

[jdonn]

rogerclee, on Mar 21 2008, 12:33 AM, said:

Cascade, on Mar 20 2008, 08:05 PM, said:

My double dummy simulation suggested that 4♠ was better. 

The chance of 3NT failing was 27.6% and 4♠ failing was 17.8.  I am surprised that this is so different than what Mike is finding by looking at hands.  My experience of comparing double dummy and single dummy is that the results usually work out reasonably close from each method.

I also extended the study by considering when would you bid 3NT with these spades.  Is there any number of extra values that make 3NT better than 4♠.  Surprisingly to me the double dummy simulation suggested not.  Here are the results:

 [space] [space] [space]Responder's HCP
 [space] [space] [space] 9 [space]10 [space]11 [space]12 [space]13
3NT 276 208 149 [space]93 [space]46
4S [space]178 120 [space]67 [space]35 [space]17

The numbers represent the number of times in 1000 that each contract failed.

I think it is not surprising that double dummy, 4♠ is better. One big factor of bidding 3NT is that we will get a bad lead very, very frequently. In fact, if I bid 3N, the one thing I do not really want led is spades.

You are allowed to get the wrong lead against 4♠ too you know. I also think "very very frequently" is quite an overbid.

[Trinidad]

I share Mike's objections to double dummy analysis. Obviously, the best way to study this situation is to predeal a large set of hands and let some players play it out (without noting the fact that dummy always is the same ) But that would be hard to do in practice.

But isn't there a possibility to let a competent bridge playing program (e.g. GIB, Jack, Wbridge) play these hands single dummy? Obviously this creates a bias towards computer play rather than human, but this bias is a lot smaller than the bias caused by double dummy play.

Rik

[TimG]

rogerclee, on Mar 21 2008, 12:33 AM, said:

Cascade, on Mar 20 2008, 08:05 PM, said:

My double dummy simulation suggested that 4♠ was better. 

The chance of 3NT failing was 27.6% and 4♠ failing was 17.8.  I am surprised that this is so different than what Mike is finding by looking at hands.  My experience of comparing double dummy and single dummy is that the results usually work out reasonably close from each method.

I also extended the study by considering when would you bid 3NT with these spades.  Is there any number of extra values that make 3NT better than 4♠.  Surprisingly to me the double dummy simulation suggested not.  Here are the results:

 [space] [space] [space]Responder's HCP
 [space] [space] [space] 9 [space]10 [space]11 [space]12 [space]13
3NT 276 208 149 [space]93 [space]46
4S [space]178 120 [space]67 [space]35 [space]17

The numbers represent the number of times in 1000 that each contract failed.

I think it is not surprising that double dummy, 4♠ is better. One big factor of bidding 3NT is that we will get a bad lead very, very frequently. In fact, if I bid 3N, the one thing I do not really want led is spades.

Often 3NT fails because there is a wide open suit. I'd be happy if they always led a spade and never managed to run their suit before I got a chance to take my tricks. I admit I will not always have nine runners.

[mikeh]

TimG, on Mar 21 2008, 07:49 AM, said:

Often 3NT fails because there is a wide open suit.  I'd be happy if they always led a spade and never managed to run their suit before I got a chance to take my tricks.  I admit I will not always have nine runners.

You must open some weird hands if, when opener has Jx or worse in spades, you often hold a wide-open suit on the side

Of course, if the idea is blasting 3N, that will happen more often, because you will miss some 9 and 10 (and even 11) card fits... but if you transfer and bid 3N, you should almost never be off an entire suit... and that was true in my simulations.. I think there was one hand on which 3N failed because the opps ran 5 tricks in one suit on opening lead: declarer had Qxx opposite xx.

[pclayton]

3N. We can never get pard to choose correctly because of the 2-3 card support paradox.

Even if we are wide open in a side suit, they might be 4-4.

I have an interesting problem if LHO makes a spec double on 5 good hearts with our without an entry. Might we expect a spade lead?

[TimG]

mikeh, on Mar 21 2008, 11:30 AM, said:

TimG, on Mar 21 2008, 07:49 AM, said:

Often 3NT fails because there is a wide open suit.  I'd be happy if they always led a spade and never managed to run their suit before I got a chance to take my tricks.  I admit I will not always have nine runners.

You must open some weird hands if, when opener has Jx or worse in spades, you often hold a wide-open suit on the side

Of course, if the idea is blasting 3N, that will happen more often, because you will miss some 9 and 10 (and even 11) card fits... but if you transfer and bid 3N, you should almost never be off an entire suit... and that was true in my simulations.. I think there was one hand on which 3N failed because the opps ran 5 tricks in one suit on opening lead: declarer had Qxx opposite xx.

I was talking about after blasting to 3NT, otherwise you lose some of the advantage, defenders will lead better after 1N-2H-2S-3N (or 1N-4H-4S) than after 1N-3N -- and meant to be referring to times that 3NT failed while 4S made (double dummy). Yes, many of these hands would choose 4S if given a choice.

Then again, there is a hand like J6 AKJ3 AJ43 JT9 that would likely choose 3N, but was down on a normal club lead while making on any other lead, including a spade. (Yes, I realize this is a very small sample size of one hand.)

But anyway, in browsing through hands it seems to me that a spade lead against 3N helps more often than the "normal" lead.

[joshs]

I wonder how well xfer then 3D will work, assuming that if you bid 4S next over a non-3N bid that partner doesn't take it as a slam try. I am just not tricky enough to bid a 2 card suit....


Personally, I think you should at least have a diamond honor in addition to perpetrate crimes like bidding a 3 card suit here, but I am endlessly curious.

Anyway, unless I have machinery to identify a 4333 hand with all fast or all slow tricks opposite, I would probably just blast 4S....

[pclayton]

joshs, on Mar 21 2008, 10:05 AM, said:

I wonder how well xfer then 3D will work, assuming that if you bid 4S next over a non-3N bid that partner doesn't take it as a slam try. I am just not tricky enough to bid a 2 card suit....


Personally, I think you should at least have a diamond honor in addition to perpetrate crimes like bidding a 3 card suit here, but I am endlessly curious.

Anyway, unless I have machinery to identify a 4333 hand with all fast or all slow tricks opposite, I would probably just blast 4S....

I held ♠Jx ♥Axxx ♦QJ9xx ♣10x yesterday.

1♣ - 1♦ - 3♣. I really thought about 3♠...

[TimG]

dburn, on Mar 20 2008, 07:40 PM, said:

It would never occur to me to bid other than an immediate 3NT, at any form of scoring.

And, you're probably playing those weak NTs.

[jdeegan]

4♠ unless I needed a board, because if 3NT does go down when partner has a weak side suit, it will demoralize him/her.

[rogerclee]

jdeegan, on Mar 22 2008, 02:27 AM, said:

4♠ unless I needed a board, because if 3NT does go down when partner has a weak side suit, it will demoralize him/her.

Yes, the issue is, how likely is it that partner has a weak side suit when we already know he has a weak spade suit? Is it more likely that:

1) We have 9 tricks in either contract all day?
2) A bad lead against 3N hands the contract to us?

[Cascade]

rogerclee, on Mar 23 2008, 01:24 AM, said:

jdeegan, on Mar 22 2008, 02:27 AM, said:

4♠ unless I needed a board, because if 3NT does go down when partner has a weak side suit, it will demoralize him/her.

Yes, the issue is, how likely is it that partner has a weak side suit when we already know he has a weak spade suit?

I got these numbers:

♥ xxxx 0.2274%
♥ Jxx or worse 2.8580%
♥ Qx or worse 3.3414%

♦ xxxx 0.1075%
♦ Jxx or worse 2.2504%
♦ Qx or worse 3.4468

♣ xxxx 0.2442%
♣ Jxx or worse 2.8605%
♣ Qx or worse 3.3308%

Altogether this is 18.667%

In addition exactly

♥ Qxx 2.8767%
♦ Qxx 2.5785%
♣ Qxx 2.8907%

Another 8.3459%

There are other combinations that are potential problems e.g Jxxx or Kx when the opponents can lead to an ace and switch etc.

There are also likely to be more problems when we consider the modern style is to open 1NT on a wide range of 6=3=2=2 hands and slightly weaker total high-card values. Personally I would be happy to open 1NT with a couple of Qx suits and a solid minor and I might do it with weaker doubletons (I certainly know many would).

I think we are getting close to or maybe even higher than 1/3 hands where we have a potential problem. Of course this problem could be solved by a favourable lead or a 4-4 break or blockage. Against that there might be other problems on the hand.

In my partnership we have a rule of thumb that we at least investigate a major suit contract via Stayman or a transfer whenever we have a weak doubleton. Here the hand in question has two such holdings.

[rogerclee]

Cascade, on Mar 22 2008, 01:06 PM, said:

♥ xxxx 0.2274%
♥ Jxx or worse 2.8580%
♥ Qx or worse 3.3414%

♦ xxxx 0.1075%
♦ Jxx or worse 2.2504%
♦ Qx or worse 3.4468

♣ xxxx 0.2442%
♣ Jxx or worse 2.8605%
♣ Qx or worse 3.3308%

Altogether this is 18.667%

In addition exactly

♥ Qxx 2.8767%
♦ Qxx 2.5785%
♣ Qxx 2.8907%

Another 8.3459%

This is much higher than I expected and enough evidence to convince me that 4♠ is likely better. Thanks.

[cherdano]

mikeh, on Mar 20 2008, 05:27 PM, said:

Cascade, on Mar 20 2008, 05:35 PM, said:

Texas is clear 3NT maybe horrid opposite two small and a bad break but we might still make 4♠ with a trump loser.

Justin is exactly right (as he often is) - why do we want to play 3NT opposite two small?

I'll do a simulation too.

OK, I'm up to 160 hands now And I think I will stop.

So far, both contracts are strong favourites to make opposite a 2 card spade holding (no surprise there). When one is better than the other, it is so far 13 for 3N making and 10 for 4♠.

So the gap has narrowed, but 160 hands is starting to resemble a statistically significant number (well, my knowledge of stats is based on undergraduate courses more than 30 years ago, so I may be overstating matters here... dangerous with so many mathematicians posting )

Sorry to take you up on the invite
When the result is so close (13 to 10), then 160 is unfortunately not statistically significant. Maybe I should really buy GIB so that I can try single-dummy simulations for questions like this. This wouldn't be perfect but avoid the obvious problems of double dummy analysis and still easily give statistically significant numbers. (Easy for me, not for my Laptop's CPU )