[gnasher]

han, on 2012-March-21, 19:05, said:

I redid the simulation, now with 2000 hands. 3NT makes on 89% of the hands, 4S on 83% of the hands. So maybe it is obvious that we should bid 3NT.

What were the shape constraints for your 1NT opener?

I used to bid 3NT on most 4333 shapes, but I'm less inclined to now, because it's more common to open 1NT on a 5422, 4441 or even 5431 type. I'm more inclined to look for a heart fit than for a spade fit, because partner will open 1NT on 24(52) more often than on 42(52).

Quote

So to paraphrase, against 4S it doesn't matter too much what they lead, but against 3NT they'd better lead the right suit. Another argument not to investigate too much when we very likely heading towards 3NT.

To me that sounds like an argument against playing Stayman, not an argument against investigating on a 4333 shape. But I agree that within the parameters of the question it's relevant.

[han]

What were the shape constraints for your 1NT opener?


I don't know, I selected the box "balanced". I suspect it's 4333, 4432 or 5332, but the latter is not possible since I forced exactly 4 spades.

I have never opened 1NT with a 5431 shape, and certainly never with 4 spades. I wouldn't open 1NT with a 4441 shape and 4 spades either. So I certainly agree with you that this is an argument for investigating more often with hearts than with spades. (but to be honest, I'm almost always going to jump to 3NT with 3433 and 12 HCP.)


To me that sounds like an argument against playing Stayman, not an argument against investigating on a 4333 shape.


I don't follow your reasoning, how can the double dummy results obtained for this specific hand be an argument against playing stayman in general, rather an argumemt against investigating on this specific hand? It is true that one convention may do a better job in hiding information than another, but they all give information aboutdummy's hand or opener's hand, there is no way around it. Of course auctions in which we discover a 4-4 major suit fit but later make an informed decision to play 3NT anyway are absolutely worst in this respect, it's interesting that they are named as a plus by several people in this thread.

If you meant to write that there are conventions that investigate a 4-4 major fit without showing as much about opener's hand then I certainly agree. There are many options and for this purpose Stayman is probably the worst. I don't see why we should dump stayman though, it is a very useful convention for many other purposes. By which I again don't mean to say that it is better than other methods.

[gnasher]

han, on 2012-March-22, 15:59, said:

I don't know, I selected the box "balanced". I suspect it's 4333, 4432 or 5332, but the latter is not possible since I forced exactly 4 spades.

If so, I don't think your results tell us very much. I suspect that we'd want to play in 4♠ opposite any 5422 with four spades, and almost any 5332 with five spades.

[MrAce]

Han, it is a very interesting statistics that you brought up about the leads that defeats vs both 3NT and 4M. And i think thats the main advantage. But ty for the simulation, when you wrote it i started to think how accurate the result u posted with my experience, and i think it is very accurate, at least accurate according to my own experience.

[gnasher]

han, on 2012-March-21, 19:22, said:

What's also interesting is that against 3NT the lead matters more. A standard feature in dealmaster pro is that, only for the hands where the contract can be beaten, it keeps track of from how many suits the best lead is. These are the results:

Against 3NT:

1 suit: 115
2 suits: 23
3 suits: 57
4 suits: 28

Against 4S:

1 suit: 60
2 suits: 36
3 suits: 116
4 suits: 132

So to paraphrase, against 4S it doesn't matter too much what they lead, but against 3NT they'd better lead the right suit. Another argument not to investigate too much when we very likely heading towards 3NT.

I think I misunderstood this. If these figures are only for deals where we have a 4-4 spade fit, I agree that it doesn't tell us anything about the merits of Stayman as a convention. We'll just have to use our judgement about that instead.

[bluecalm]

Quote

To me that sounds like an argument against playing Stayman, not an argument against investigating on a 4333 shape. But I agree that within the parameters of the question it's relevant.

Sometimes 4M is much better than 3N facing 15-17 balanced with your 4M with 4-3-3-3 it never is.
I consider staymaning on such hands a mistake but somehow top Italians usually stayman and as I think they are by far the best bidders in the world who never open 1N with 5M-3-3-2 and very rarely with other off-shape hands some little doubts remain in my mind.

[barmar]

han, on 2012-March-21, 19:22, said:

What's also interesting is that against 3NT the lead matters more. A standard feature in dealmaster pro is that, only for the hands where the contract can be beaten, it keeps track of from how many suits the best lead is. These are the results:

Against 3NT:

1 suit: 115
2 suits: 23
3 suits: 57
4 suits: 28

I wonder how often the "right" lead is 4th from your longest and strongest, or a short suit from a very weak hand? If it's not that hard finding the best lead, then this advantage is diminished.

[han]

gnasher, on 2012-March-22, 16:20, said:

If so, I don't think your results tell us very much. I suspect that we'd want to play in 4♠ opposite any 5422 with four spades, and almost any 5332 with five spades.

Why are you so dismissive of these results? Frances wrote that with this hand we would obviously search for a 4-4 spade fit. I think that these results show very clearly that this is far from obvious. Suppose that I had just written that from my experience I don't consider it obvious at all, do you think anybody would have paid attention to my claim? I doubt it, even I would not!

You have mentioned a lot of factors that should influence our decision, I agree with all of them. How often we open 1NT with a 5-card major, how often we open 1NT with a singleton or 4-2-5-2 shape, what our methods are, and whether we have hearts or spades, these are all good points.

That doesn't mean that these simulation results are worthless, we can take them for exactly what they are. That if partner is a disciplined (for lack of better word: old-fashioned) 1NT opener, it would not be clear that we'd want to play 4S even if we knew that partner had 4 spades. If anything, the results suggest that we should still opt to play 3NT, preferably after a blind auction.

How about this, tonight (US time) I will do another search where I give responder a 3-4-3-3 9-count (with 4 hearts) and manually specify that opener is "balanced" with 4 or 5 hearts, but is allowed to be 5332 or 2-4-(52) or 1-4-4-4 with a stiff queen, king or ace. This is more or less my style, and I think that this is not so far from your style? I don't know what the results will be, but I suspect that 4H will do significantly better than 3NT. Maybe not, it would be interesting to see the results either way. They would still be double dummy results, but we comparing the results with the earlier results could be interesting.

[han]

barmar, on 2012-March-23, 09:03, said:

I wonder how often the "right" lead is 4th from your longest and strongest, or a short suit from a very weak hand? If it's not that hard finding the best lead, then this advantage is diminished.

I can redo the simulation and check manually how often this is the case (although you haven't specified what a very weak hand is or which short suit they should lead. a major? spades?). DMP highlights all the best leads so this is easy. I won't do it for 2000 hands though!

[aguahombre]

han, on 2012-March-23, 09:28, said:

I won't do it for 2000 hands though!

Aw, come on. We can wait a few more minutes for your job to be done right

[bluecalm]

Quote

I can redo the simulation and check manually how often this is the case (although you haven't specified what a very weak hand is or which short suit they should lead. a major? spades?). DMP highlights all the best leads so this is easy. I won't do it for 2000 hands though!

Yeah this digging through hands as I call it is quite time-consuming. I did a lot of it in the past. That's why I have so strong opinion about puppet/transfers to 3N and quite strong opinion about this one too (undermined only slightly by Versace constantly staymaning on 4-3-3-3's).

[mgoetze]

barmar, on 2012-March-23, 09:03, said:

I wonder how often the "right" lead is 4th from your longest and strongest, or a short suit from a very weak hand? If it's not that hard finding the best lead, then this advantage is diminished.

GIB seems to simulate that the best lead is from a short honorless suit rather often.

[barmar]

mgoetze, on 2012-March-23, 10:21, said:

GIB seems to simulate that the best lead is from a short honorless suit rather often.

I was thinking about that when I wrote my post.

I haven't done any kind of analysis to determine how often GIB's leads are actually best. They do seem to work out reasonably well, but mainly because its partner also knows better than to return its suit -- it often leads back its own worthless suit, leading through declarer to find partner's strength. So is this a good strategy?

It frustrates the human players no end -- we get frequently complaints "Why doesn't GIB return my suit"?

[gnasher]

han, on 2012-March-23, 09:24, said:

Why are you so dismissive of these results?

Because, as I understand it, you excluded a category of hand that nearly all of us would open 1NT, where 4♠ is very likely to be better than 3NT.

If you were testing a new vaccine, you wouldn't start by excluding all the people who are most likely to catch the disease, would you? That's the equivalent of what you've done by excluding 5332 shapes.

Quote

Suppose that I had just written that from my experience I don't consider it obvious at all, do you think anybody would have paid attention to my claim? I doubt it, even I would not!

Why wouldn't I pay attention to that? You're a strong, experienced player, and apparently objective (except perhaps when it comes to double-dummy simulations). I'm more inclined to believe your experience of single-dummy results than to accept the results of a double-dummy simulation that we already know is flawed.

Quote

That doesn't mean that these simulation results are worthless, we can take them for exactly what they are. That if partner is a disciplined (for lack of better word: old-fashioned) 1NT opener, it would not be clear that we'd want to play 4S even if we knew that partner had 4 spades. If anything, the results suggest that we should still opt to play 3NT, preferably after a blind auction.

OK, I should have been more specific: they don't tell us much if we're trying to decide what to do with KJ10x Axx xxx Axx opposite a modern 1NT opener. I agree that they would be of value if I played some other style, but I don't.

Quote

How about this, tonight (US time) I will do another search where I give responder a 3-4-3-3 9-count (with 4 hearts) and manually specify that opener is "balanced" with 4 or 5 hearts, but is allowed to be 5332 or 2-4-(52) or 1-4-4-4 with a stiff queen, king or ace. This is more or less my style, and I think that this is not so far from your style? I don't know what the results will be, but I suspect that 4H will do significantly better than 3NT. Maybe not, it would be interesting to see the results either way. They would still be double dummy results, but we comparing the results with the earlier results could be interesting.

Yes, that would be interesting, but I note that you've changed two variables: you've changed opener's hand-types, but you're also moving from a specific responding hand to any 3433 9-count. When we compare these results with your earlier results, we may be left uncertain as to what caused any change.

[gnasher]

Regarding leads, one of the problems with methods like Stayman in Doubt, where you find out that you're both 4333 and then play in 3NT, is that you tell the opponents that it's a good idea to lead from a doubleton.

[phil_20686]

SO a few comments:

(1) It has always been my understanding that you are meant to bid 3N with 4333 when weaker, and look for a 4-4 major fit when in the 28-30 combined HCP range. The reasons for this seem obvious: when you have lots of points, 3N can go off when you are missing a suit, or most of a suit, and one other trick, whereas 4M maintains control. When you have fewer points you may just have 9 tricks, or you may go off against a poor trump break, when playing in 4M isnt giving you any extra tricks.

(2) The worst hands to look for 4M are those with slow honours in ever suit: xxxx KJx QJx QJx should absolutely never look for a spade fit. The best hands for 4M are those with strong trumps and obvious weak spots: AKxx Axx xxx xxx for example.

(3) The lead issue doesnt surprise me at all. 3N often goes off when the opponents can establish a suit. In 4M the cause is often is too many trump losers. Trump losers don't go away, and give you some control to find your other winners.

[manudude03]

Out of curiosity, suppose you decide to agree respond 2♠ with 4-4 in the majors, would anyone bid stayman and then 3NT over 2S even with a fit and 4333 shape? That way if partner pulls to 4H, you can bid 4S to play knowing partner has a ruffing value.

[FrancesHinden]

Han's simulation results are quite interesting, not least because my opening 1NT style is actually close to the one simulated than the less shape-constrained one gnasher prefers. It is a good point that you are better looking for a heart fit with 3433 than a spade fit with 4333 because (i) partner is more likely to open 1NT with 2452/2425 than with 4225/4252 (IMO) and (ii) partner is more likely to be 4333 when he responds 2S to Stayman, than he is to be 3433 when he responds 2H.

Maybe I should retrench slightly and say that I still think it's worth looking for a fit if you can offer partner choice of games and both of you can use your judgement. What's more you can't disprove it (ha!) because if we have perfect judgement we'll always play in the right contract. All we lose is what we reveal about our hands, and you aren't going to prove what that costs either (ha! squared) because your DD analysis won't help you there.

[han]

Opener: 15-17 HCP, 4-5 hearts, either balanced or 1-4-4-4 with a singleton spade top-honor or 2-4-(52).

Responder: 3-4-3-3 with 9 HCP.

1000 double dummy hands. Results:

3NT makes 444 times.

4H makes 375 times.

The leads numbers:

3NT: 154-58-111-233
4H: 68-64-134-359

[han]

gnasher, on 2012-March-23, 11:28, said:

I'm more inclined to believe your experience of single-dummy results than to accept the results of a double-dummy simulation that we already know is flawed.

I'm flattered, but how often do you think I've encountered a GF 4333 responding hand with a 4-card major, compared the results between jumping to game or bidding stayman, and actually remembered the results well enough to make an informed opinion now? In fact I only remember one hand: I opened 1NT with 5 hearts, my partner jumped to 3NT holding a 3-4-3-3 shape and I had 9 tricks after the lead, while 4H was completely hopeless. You can imagine that I remember this experience very well, it won our team IMPs in the league and we had a good laugh about it. This single hand has probably influenced my opinion on the subject a lot, so I don't think that my personal experience is that reliable at all.