[pescetom]

[ ATTENTION: THIS POST WAS REVISED NOV 5th TO CORRECT A BUG, INITIALLY THINGS LOOKED UNDULY FAVORABLE TO 3NT ]

I cleaned up the script a bit and turned to my next point of interest, to see whether it was vantageous in terms of score for responder with a 5m332 to go through some kind Stayman (capable of locating a 3-5 major fit) or just bid in NT.
Initially I look at responder with enough for a thin game (10 HCP) or a safe game (11-12 HCP), as the script isn't ready for invitational yet.
The script assumes that opener is vulnerable, that we can find the 3-5 major fit if we wish and that the opponents keep quiet.

At first sight the results seem to say that in a thin game we are much better off in 4S than in 3NT, but this advantage disappears with safe game strength.

#
# NT vs Spades comparison
#
op_nt1517= hcp(north)>=15 and hcp(north)<=17
rs_weak= hcp(south)>=4 and hcp(south)<=7
rs_invite= hcp(south)>=8 and hcp(south)<=9
rs_thingame= hcp(south)>=10 and hcp(south)<=10
rs_safegame= hcp(south)>=11 and hcp(south)<=12
rs_5m3S32= shape(south, any 5332) and hearts(south)<=3 and spades(south)==3
rs_xfer= spades(south)>=5 or hearts(south)>=5 or clubs(south)>=6 or diamonds(south)>=5
op_s5 = spades(north)==5 and shape(north, any 5332)
produce 10
##### unquote ONE condition below at a time
condition op_nt1517 and op_s5 and rs_thingame and rs_5m3S32
#condition op_nt1517 and op_s5 and rs_safegame and rs_5m3S32
x3N_tricks= tricks(north,notrumps)
x4S_tricks= tricks(north,spades)
x3N_score= score(vul,x3N,x3N_tricks)
x4S_score= score(vul,x4S,x4S_tricks)
outcome = x3N_score == x4S_score ? 0 : (x3N_score > x4S_score ? 1 : -1)
##### unquote ONE action below at a time
action frequency(x3N_tricks,6,13)
#action frequency(x4S_tricks,6,13)
#action frequency(outcome,-1,1)

THIN GAME =======================================================

Frequency of tricks in 3NT

Frequency :
    6	       7
    7	      11
    8	      27
    9	      20
   10	      16
   11	      13
   12	       6
   13	       0
Generated 4986840 hands
Produced 100 hands
Initial random seed 1604580564
Time needed   60.142 sec

Frequency of tricks in 4S

Frequency :
    6	       0
    7	       0
    8	       6
    9	      21
   10	      41
   11	      24
   12	       8
   13	       0
Generated 4863518 hands
Produced 100 hands
Initial random seed 1604580675
Time needed   56.812 sec

Score is better in 4S / same in both / better in 3NT

Frequency :
   -1	      54
    0	      10
    1	      36
Generated 5100876 hands
Produced 100 hands
Initial random seed 1604580768
Time needed   77.179 sec

SAFE GAME =======================================================

Frequency of tricks in 3NT

Frequency :
Low	       1
    6	       1
    7	       3
    8	      10
    9	      20
   10	      27
   11	      26
   12	      11
   13	       1
Generated 4162578 hands
Produced 100 hands
Initial random seed 1604581057
Time needed   56.394 sec

Frequency of tricks in 4S

Frequency :
    6	       0
    7	       0
    8	       1
    9	      14
   10	      29
   11	      40
   12	      15
   13	       1
Generated 3433173 hands
Produced 100 hands
Initial random seed 1604581164
Time needed   42.272 sec

Score is better in 4S / same in both / better in 3NT

Frequency :
   -1	      47
    0	       4
    1	      49
Generated 3803028 hands
Produced 100 hands
Initial random seed 1604580938
Time needed   64.302 sec

[mycroft]

Interesting, but you have limited the hands in question to "5♠332 opposite 5m332". NT fails badly (when it fails) when the doubletons line up; 4♠ likely fails as well as there's no ruffs available, but not as badly. There's a maximum of one ruff for an extra trick, even if the doubleton's don't line up.

If South has two spades (33% of the hands), then you're going to end up in 3NT anyway. Yes, that's a losing case for checking, because you've got the defence off to a great start in counting the hand (not that that's relevant to double-dummy analysis, but in real life, anyway).

How do the numbers change if you remove the 5332 in the south, and just give it "hearts <= 3 and spades <= 3" ? What if, as in real life, you give South "hearts <= 4 and spades <= 4 and not (hearts==4 and spades == 4)" (those that play 1NT-3♣ gf Puppet will do it with GF 4=3, too, because they can find both fits)? Okay, (43)(60) probably has another sequence, and (43)(51) may, as well (as may (31)(45)), but (42)(43), or even (41)4=4, might go through puppet instead of just regular Stayman.

Again, this isn't going to be seen in double dummy analysis, but people are moving to 1NT-3♣; 3♦ being "no 5 card major (says nothing about 4-card)", which allows people to ask with 3=3 or 32 in the Majors and just bid 3NT after, without giving away too much. Not necessarily relevant to the analysis either, but a point to consider.

There is the anti-field question to ask as well; sure, it's a tossup whether 3NT or 4M is better; but at least now, a huge majority of the field are going to be in the 8-card major fit, so when 3NT wins, you get a near-top; when it loses (including when 3NT= and 4M=), you get the near-bottom. When you bid 3NT here, you are actually bidding 4NT; gaining only over actually bidding 4NT when 3NT= meets 4M-1 (or 3NT-1 meets 4M-2, or...). You'll be spending a lot of time looking at dummy and trying to work out the actual contract you're playing at trick 1. Note that I don't consider that a downside if there are corresponding benefits; I play a weak NT in the ACBL, so I'm used to it. But it will increase your volatility, and it will take brainpower away from the rest of your game. Of course, "following rules in the auction" like "not deciding whether to open 5M332 1M or 1NT based on..." or "not deciding whether to puppet look for 4M with 5m332 instead of just Hamman's Rule-ing it" saves some brainpower in exchange.

[pescetom]

Thanks for the thoughtful reply.

mycroft, on 2020-November-04, 10:26, said:

Interesting, but you have limited the hands in question to "5♠332 opposite 5m332". NT fails badly (when it fails) when the doubletons line up; 4♠ likely fails as well as there's no ruffs available, but not as badly. There's a maximum of one ruff for an extra trick, even if the doubleton's don't line up.

We currently do usually play in 3NT without looking for the 3-5 fit, but I am not certain it is the better line, hence this end of year system maintenance check. As you say later it is an anti-field line, especially over here, and my gut feeling was that playing 3-5 in NT may cost overall. As you say here, NT can fail badly. I recall a few times too often when I was struggling to bring home 3NT while others were in a comfortable and better scoring 4M. These numbers seem to say that playing in the major is safer with a thin game but not with a safe game.

Another possible factor here is the difference between double dummy and single dummy play (not just mine either). I suspect 3NT= double dummy might be harder to make than 4M= double dummy, being more likely to depend upon an improbable line of play. That would make playing in the major even more compelling than the numbers suggest. Would be interested to hear the thoughts of mikeh and other experts on this.

mycroft, on 2020-November-04, 10:26, said:

How do the numbers change if you remove the 5332 in the south, and just give it "hearts <= 3 and spades <= 3" ? What if, as in real life, you give South "hearts <= 4 and spades <= 4 and not (hearts==4 and spades == 4)" (those that play 1NT-3♣ gf Puppet will do it with GF 4=3, too, because they can find both fits)? Okay, (43)(60) probably has another sequence, and (43)(51) may, as well (as may (31)(45)), but (42)(43), or even (41)4=4, might go through puppet instead of just regular Stayman.

Again, this isn't going to be seen in double dummy analysis, but people are moving to 1NT-3♣; 3♦ being "no 5 card major (says nothing about 4-card)", which allows people to ask with 3=3 or 32 in the Majors and just bid 3NT after, without giving away too much. Not necessarily relevant to the analysis either, but a point to consider.

I am doing my best to avoid going into detail about Stayman/Transfer methods, because ours are unusual and I don't want to side track the discussion. Just please assume that Stayman could find any 3-5 or 4-4 major fit at invitational or higher level and has minimal leakage about opener's hand.

I don't see that it is relevant to examine distributions of responder with a 4 card major. If there is a fit we will find it and so (usually) will the rest of the field, so most will be in 4M. It's the special case of 3-5 fit that is interesting.

[mycroft]

If you do something else with 4=3 in the majors other than puppet to find the 5-3 or the 4-4 fit, then sure. But otherwise, ignoring those hands is skewing your results. How, I don't know.

[awm]

It seems like your statistics include hands where responder has 2♠-3♥-(35). No one is playing these in 4♠, and they will surely do significantly better in 3NT. So by including these you are making 4♠ look worse and 3NT look better. You want to distinguish between:

A. Play in the 5-3 spade fit when it exists and otherwise 3NT.
B. Play in 3NT always.

Not between:

C. Play in 4♠ always.
D. Play in 3NT always.

It's easy to believe D >> C but this does not imply anything about A vs. B.

[pescetom]

mycroft, on 2020-November-04, 11:56, said:

If you do something else with 4=3 in the majors other than puppet to find the 5-3 or the 4-4 fit, then sure. But otherwise, ignoring those hands is skewing your results. How, I don't know.

We put 4=3 in the majors through our Stayman, which is initially a true puppet. I don't see how it matters. If there is a 4-4 fit we will find it and usually play there, like most competitors, so it is not relevant to my results. My results are only about the specific situation of a 3-5 fit, where two different lines are possible.

[pescetom]

awm, on 2020-November-04, 12:18, said:

It seems like your statistics include hands where responder has 2♠-3♥-(35). No one is playing these in 4♠, and they will surely do significantly better in 3NT. So by including these you are making 4♠ look worse and 3NT look better. You want to distinguish between:

A. Play in the 5-3 spade fit when it exists and otherwise 3NT.
B. Play in 3NT always.

Not between:

C. Play in 4♠ always.
D. Play in 3NT always.

It's easy to believe D >> C but this does not imply anything about A vs. B.

I agree that I want to distinguish between A. and B. but not that I am failing to do so.
With 2♠-3♥ we will play in 3NT opposite 5 spades, not in 4♠.
We would play in 4♥ opposite 5 hearts.


OK, the penny dropped I see what you are getting at, thanks.
There was a bug in the script that made it consider 2♠-3♥ as a potential fit with 5 card spades.
[Now fixed in the OP, because otherwise we would have to start the thread all over again.]