[kenrexford]

Where's all the deal generators out there? It's about time for a scientific analysis we can all pounce on a ridicule.

[awm]

I randomly generated hands with 20-21 hcp, 2-3 cards in each major, 2-6 in each minor. This probably weights hands a bit on the "heavy" side since people upgrade some 19s (I have generated no 19s) and might upgrade some of the 21s I generate to open 2♣. Anyway, here are the hands:

1.
♠KQ
♥K5
♦AKT7
♣AQ843

2.
♠K97
♥A4
♦AKQ98
♣AJ9

3.
♠AJ6
♥K8
♦AK7
♣KQ843

4.
♠KT8
♥K84
♦AK7
♣AKJ3

5.
♠QT
♥AK6
♦AKT
♣KQT73

6.
♠A6
♥AT8
♦K86
♣AKQT4

7.
♠AKQ
♥KT6
♦AKQ
♣9542

8.
♠QT6
♥AK7
♦AKQ87
♣Q6

9.
♠A9
♥A95
♦AQ986
♣AKT

10.
♠A9
♥K54
♦AQ98
♣AKJ6

11.
♠AJ7
♥K9
♦AKQT8
♣AT5

12.
♠QJ7
♥A9
♦AK9
♣KQJ87

13.
♠KQJ
♥A84
♦AQ9
♣KQT6

14.
♠AKJ
♥A8
♦A976
♣KJT8

15.
♠K87
♥AT
♦K6
♣AKQJ52

16.
♠KJ
♥KT
♦AKQT9
♣A852

17.
♠KJ8
♥AT
♦AQ9
♣AQ643

18.
♠T6
♥A6
♦AKQ
♣AQJ975

19.
♠K7
♥A6
♦AQ7
♣AKT542

20.
♠AK8
♥A8
♦KQ86
♣AJ42

I will leave it to others to analyze these.

[Mbodell]

kenrexford, on Jun 15 2008, 06:13 PM, said:

Where's all the deal generators out there?  It's about time for a scientific analysis we can all pounce on a ridicule. 

Alright, ask and get an answer. Assumptions:

1. The x's in OP could include Ts. (I think this biases pro-NT as the T will carry more weight in NT).

2. The 2NT bidding is on the range HCP=20-21 semibalanced (no void, no stiff, no 6 card major, no 7 card minor) and has <4 cards in each major. So no upgrades, no stiffs, etc. which again will slightly bias pro-NT I think.

3. The analysis is DD. I think in this case with the strength so unbalanced and the opening lead being extra important in NT this is a possible bias anti-NT. Since I don't think declarer is getting to make lots of choices of which finesses to take when.

4. Nobody ever doubles. This is obviously a pro-NT bias.

Ok, so I dealt 1000 hands with us V and then, for fun, compared the scores when we were NV (on a different run of 1000 hands since I thought of this during the first run). Note when we don't make in NT we are not uncommonly down 2 or 3 tricks (which is why I wanted to try the NV numbers too, as if it was always = or -3 then it would score better NV than V since it is down nearly twice as often as it makes) and 3♦ nearly always makes. Basically if the scoring was MP or the vulnerability was NV then we should pass 3♦ for sure. But vulnerable at IMPs the results very slightly favor 3nt over my simulation.

We make 3♦ 91.6% of the time for an average score of 113.82
We make 3nt 36.7% of the time for an average score of 129.59
Diamonds is better than nt 58.9% while nt is better than diamonds 37.2% and 3.9% they are the same


We make 3♦ 91.9% of the time for an average score of 118.72
We make 3nt 38.7% of the time for an average score of 74.73
Diamonds is better than nt 56.8% while nt is better than diamonds 39.7% and 3.5% they are the same

I'll try and run a larger sample tomorrow while I'm at work as my program isn't optimized for speed (partially because I let the x's from the OP weak hand change rather than setting the hand as 1 fixed hand and partially because of the 2 DD solver runs per matching hand) and can only do <10 matching hands/minute.

[kenrexford]

Mbodell:

I would imagine that this sample has one other pro-NT bias, illustrated by this actual two-part problem.

When Opener has no four-card major, the chances of one five-card minor increase, or six-card for that matter. Because of the re-evaluation going on here, your numbers would need to include a few 19's and eliminate a few 21's.

Almost half of your sample 20 may be upgrades, some clearly.

[P_Marlowe]

#1 2NT, as far as I know, 2C is forcing to
game, and i am not going to force to
game with the given hand

#2 Stayman

With kind regards
Marlowe

[Little Kid]

I thought 2♣ is only forcing to game if the 2♣ bidder does not rebid 2NT? But I'm probably mistaken...

[helene_t]

Little Kid, on Jun 16 2008, 02:05 PM, said:

I thought 2♣ is only forcing to game if the 2♣ bidder does not rebid 2NT? But I'm probably mistaken...

No you are probably right (unless playing reverse Benji or some other way of showing a hand slightly stronger than a 2NT opening).

[Little Kid]

I play 2/1, or at least I try to.

[Mbodell]

kenrexford, on Jun 16 2008, 03:35 AM, said:

Mbodell:

I would imagine that this sample has one other pro-NT bias, illustrated by this actual two-part problem.

When Opener has no four-card major, the chances of one five-card minor increase, or six-card for that matter. Because of the re-evaluation going on here, your numbers would need to include a few 19's and eliminate a few 21's.

Almost half of your sample 20 may be upgrades, some clearly.

I think that was my point 2. How would you define the 2NT then? 20-21 points (HCP + length points)?

The sample 20 presented weren't my hands.

[kenrexford]

Mbodell, on Jun 16 2008, 11:05 AM, said:

kenrexford, on Jun 16 2008, 03:35 AM, said:

Mbodell:

...Almost half of your sample 20 may be upgrades, some clearly.

I think that was my point 2. How would you define the 2NT then? 20-21 points (HCP + length points)?

The sample 20 presented weren't my hands.

Whoops! Yeah, I missed that sample ownership problem.

What are "length points?" Does that roughly mean 19-20 if a five-card suit exists? I suppose that would be a fair approximation. I'm not sure how defined you can get.

For me, I would expect a 2NT opening that is defined as "20-21" to be:

Normally 20-21.
19 allowed, 21 disallowed if a decent 5-card suit and 7+ controls (A=2; K=1).
19 allowed, 21 disallowed if 8+ controls.
19 allowed, 21 disallowed if 4432 with 7+ controls and at least one well-placed 10.
18 allowed, 20 discouraged (no 10's), 21 disallowed if a good six-card minor and 7+ working controls.

Something like that. I'm not sure if that can be programmed in. I would also imagine that the percentages likelihood of success of 3NT will tweak slightly one way or the other depending on individual definitions of what misleadingly appears to be a consistent "20-21."

This tweaking does seem relevant in this specific instance, however, as the preliminary question coupled with the secondary question seems to suggest an underlying problem to address. Was Opener too strong for the 2NT opening, or was Rersponder wrong in not bidding 3NT, or did both people bid wrong. I would imagine that Responder's position, if these two hands were opposite each other, is that he would have bid 3NT after 2♣...2NT but not after 2NT directly. Thus, to actually answer the assumed debate, one must analyze 2NT openings in the context of a majority position that the actual hand merits an upgrade to a 2♣...2NT sequence.

My criteria for upgrades (just now created; not sure what I actually do in practice except for a "feel") may be different than others' criteria.

[awm]

One comment about Mbodell's post:

The "x" cards were presented as "x"s. It seems reasonable to assume they are quite small. Some of us (including me) would in fact force to game if you add a couple of tens to the responder hand, but are passing 3♦ with the actual cards. In fact it is not hard to see that there are many hands where a ♦T provides substantial additional chances of setting up the diamonds, or where a ♠T provides a critical second stopper in spades.

[ArcLight]

To me x is less than a 9. In fact if a hand were AT987 I would list all the cards,

10s and 9s, especially in combination will be listed. x means crap, that will only win a trick in All through the Pack.

>Some of us (including me) would in fact force to game if you add a couple of tens to the responder hand

Not 10s or 9s, more like 6s and 7s

[kenrexford]

OK. "x" means the deuce, "xx" 32, "xxx" 432. Now what?

Actually, I wonder what the lowest critical pip would be.

On pure merits, assuming 4432 pattern in the suit, J-Q-K-A, 7-8-9-10, 4-5-6-pitch, 2-3.

So, I suppose Q853 is much better than Q852.

Not to mention the impact of intermediates on signals. Higher pips make high-low signals more costly.

Yep, deuce and up for "x" value.

[Mbodell]

Ok, my larger sample size with moving x's which can include T and 9 was:

=================================
After 30000 hands the V results are:
We make 3d 91.2933333333% of the time for an average score of 113.231333333
We make 3nt 36.9866666667% of the time for an average score of 129.702
Diamonds is better than nt 58.4866666667% while nt is better than diamonds 37.62
% and 3.8933333333% they are the same
After 30000 hands (0 attempted) the NV results are:
We make 3d 91.2933333333% of the time for an average score of 118.444666667
We make 3nt 36.9866666667% of the time for an average score of 66.767
Diamonds is better than nt 58.4866666667% while nt is better than diamonds 37.62
% and 3.8933333333% they are the same
=================================

which still favors the 3nt bid slightly. But if you eliminate T and 9 you definitely shouldn't bid 3nt!

If I instead made the x's fixed as small as possible I get:

=================================
After 1000 hands the V results are:
We make 3d 85.8% of the time for an average score of 97.24
We make 3nt 25.1% of the time for an average score of 32.73
Diamonds is better than nt 68.5% while nt is better than diamonds 26.1% and 5.4%
they are the same
After 1000 hands the NV results are:
We make 3d 85.8% of the time for an average score of 105.94
We make 3nt 25.1% of the time for an average score of 17.43
Diamonds is better than nt 68.5% while nt is better than diamonds 26.1% and 5.4%
they are the same
=================================

If I instead make the x's as big as possible but no T or 9, so counting down from 8 I get:

=================================
After 1000 hands the V results are:
We make 3d 90.3% of the time for an average score of 109.93
We make 3nt 27.2% of the time for an average score of 53.21
Diamonds is better than nt 68.0% while nt is better than diamonds 27.8% and 4.2%
they are the same
After 1000 hands the NV results are:
We make 3d 90.3% of the time for an average score of 115.48
We make 3nt 27.2% of the time for an average score of 28.21
Diamonds is better than nt 68.0% while nt is better than diamonds 27.8% and 4.2%
they are the same
=================================

The above all assume basic 2nt is 20-21 hcp semibalanced no upgrads or downgrades. If I do as large as possible x's that have no T and 9 but use upgrades/downgrades that are sort of like a more conservative version of ken's suggestions (zar points between 36 and 39 inclusive with hcp between 18 and 22 inclusive) I get:

=================================
After 2000 hands the V results are:
We make 3d 79.15% of the time for an average score of 76.695
We make 3nt 16.65% of the time for an average score of -61.1
Diamonds is better than nt 76.5% while nt is better than diamonds 18.1% and 5.4%
they are the same
After 2000 hands the NV results are:
We make 3d 79.15% of the time for an average score of 90.42
We make 3nt 16.65% of the time for an average score of -29.725
Diamonds is better than nt 76.5% while nt is better than diamonds 18.1% and 5.4%
they are the same
=================================

People can chime in on upgrades, as mine was slightly more conservative then Ken's to keep it down to 1 zar number rather than all the different cases he outlined. From Ken's post my reading is he'd suggest that the following are all minimum 2NT openers (varies from 35-38 zar inclusive - I'd use 3/4 of these instead of 100%):
K87 A5 AK8 KQ642
KQ6 A8 AQ6 AT872
AJ6 K65 AK4 AT52
AJ7 AT8 AQ6 A962
62 A54 AQT2 AKQ9
K2 KT2 AKT4 AQ53
AK3 A6 K4 KJT754
KT2 A4 KT AKJ542

while the following were all too good to open 2NT (varies from 37-40 zar inclusive - I'd use 3/4 of these instead of 0%):
AK7 K4 AKQ65 Q92
AQ K75 AK9 KQ862
AQ6 AK7 A87 KJ98
AQ6 AK5 A87 AJ87
KQ4 K4 AKT3 AQ32
AKJ K6 KQ42 AJT5
K6 43 AK4 AKQJT2
QJ AT5 AQ AK9842

[kenrexford]

I would agree 100% with the upgrade 2NT openings you suggested, Mbodell. One was borderline (the KJ10xxx suit), but I would have no problem with 2NT. I might want the 9 in there also.

I'm almost 100% on the second group. I would definitely not upgrade the hand with the QJ tight on the side, though.

The sampel seems to suggest very strongly a pass, then. Unless there are 10's and 9's unmentioned. Even then, one sampel not run is the 10's and 9's plus upgrades, but I suspect that this favors passing 3♦ strongly as well.

Ultimately, though, the data seems to confirm the suspected problem. If these hands are a pair, Responder is accurate if he claims that Opener should have upgraded to a 2♣ opening and that this is why he was conservative in passing 3♦.

[kenrexford]

jdonn, on Jun 15 2008, 02:33 PM, said:

kenrexford, on Jun 15 2008, 02:17 PM, said:

So what if 3NT is game and 3♦ is not.  You have almost no chance in Hell oif making 3NT, but you have a lot going for 3♦.

Gove partner some ridiculously tremendous hand like Axx-AKx-AQxx-Axx, and he'd open 2♣, and that still takes a diamond hook.  Move the diamond Queen to clubs and you have a better shot, but Geez!

It seems to me that over 10 hands, 3NT with the positives and negatives summed will be less than 3♦ with the positives and negatives summed.

I mean, do you convert all 3M bids to 3NT just because 3NT is game?

You SEVERELY underestimate the chances of 3NT making. There isn't much else to say about it.

Mbodell had a few words. LOL