Simpleish question for simmers

[Jinksy]

With exactly 24 points between you, roughly evenly split, and no 4-4 major fit, do you have better expectation at IMPs from playing in 1N or 3?

 

If it depends on the vul, to what extent?

 

Thanks,

 

S

[aguahombre]

If I ever play in 1NT with 12-12, I will let you know. Our expectation based on style would be playing in 2NT if left alone, or 3m in competition.

[Endymion77]

Without any calculations to back it up, my gut feeling is that 3NT would gain quite a bit long term when vulnerable and would be close to a wash when not vulnerable but probably would still be slightly better than 1NT. Against less than perfect defense, and especially if one of the hands has a 5 card suit, just play 3NT.

[mycroft]

Years(*) of weak NT "received wisdom" is that 12 opposite 12 plays quite well in 3NT, especially after 1-3 and a blind lead. Invites are on 11s and 10s-with-extras.

 

No, that's not a sim :-)

 

(*) earliest entry I have for this is my original K/S book, so 50?

[jeffford76]

Double dummy results depend on exactly what constraints you use.

 

This is 1000 hands where north and south have exactly 12 balanced (5cd minor ok but not 5cd major), with no 8 card major fit. It also excludes any deals where the east hand has 9+ hcp and a 6+ card suit or 5+/5+ suits and any east hand with 11+ hcp and 5+/4+ under the theory most of those hands would come in. I didn't put any restrictions on west.

 

Tricks <7

59

Tricks =7

170

Tricks =8

393

Tricks =9

295

Tricks >9

83

 

Note that as mycroft said, this doesn't mean these numbers are right single dummy - I think 12 opposite 12 is harder to defend accurately than to play accurately.

[johnu]

I probably have as much faith in double dummy simulations as anybody, but with the caveat that the results represent perfect playing (and obviously, double dummy knowledge). 12 opposite 12 3NT contracts with 2 flat hands frequently require excellent planning, execution, and maybe less than optimal play by the defenders to make. IMHO, less than expert declarers would do better to stop in a part score if they can figure out that they only have 24 HCP.

[mr1303]

In England, those declarers play 1nt 2S as showing 11 points and 1nt 2nt as 12 points...

[spaderaise]

Double dummy results depend on exactly what constraints you use.

 

This is 1000 hands where north and south have exactly 12 balanced (5cd minor ok but not 5cd major), with no 8 card major fit. It also excludes any deals where the east hand has 9+ hcp and a 6+ card suit or 5+/5+ suits and any east hand with 11+ hcp and 5+/4+ under the theory most of those hands would come in. I didn't put any restrictions on west.

 

Tricks <7

59

Tricks =7

170

Tricks =8

393

Tricks =9

295

Tricks >9

83

 

Note that as mycroft said, this doesn't mean these numbers are right single dummy - I think 12 opposite 12 is harder to defend accurately than to play accurately.

 

If my calculations are right, then jefford's numbers correspond to (as a gain for the side bidding 1N):

 

non-vul: .059*3 + .17*5 + .393*5 - .378*6  = 0.724 IMPs / board
vul:     .059*5 + .17*7 + .393*6 - .378*10 = 0.063 IMPs / board

 

i.e. rather better in 1N than 3N when non-vul, and very close when vul.

[mikestar13]

Double dummy probably works to the defense's advantage in this case (totally non-revealing auction to game level). If this is true, 3NT on 12 opposite 12 should be a winner vulnerable and a wash non-vulnerable. Relative skill level of the partnerships (if known) is quite important to this decision at the table.

[Cascade]

Here is some data that I posted in rec.games.bridge in two messages in 2006. I hope that it is helpful.

 

Here are the numbers for 10000 hands for each table entry:

Percentage chance of making 9 tricks double dummy - declarer's points
shown on the side (rows) and total combined points shown on the top
(columns)

     23     24     25     26
0    2.14   6.07  13.99  30.56
1    3.73   8.83  19.33  36.8
2    5.39  13.73  27.48  48.7
3    7.62  17.47  34.32  55.82
4   10.47  21.68  39.62  59.86
5   12.63  25.52  44.53  65.7
6   14.51  28.4   47.57  67.49
7   15.56  31.02  50.89  70.95
8   16.83  32.8   51.76  70.84
9   18.66  33.94  53.91  72.12
10  18.83  36     54.96  73.58
11  19.71  36.23  56.26  74.36
12  19.87  36.9   56.04  74.53
13  19.6   36.85  56.26  74.6
14  18.7   35.55  56.42  74.63
15  18.04  36.3   56.37  74.17
16  17.17  34.52  55.2   74.5
17  15.65  32.14  55.41  74.21
18  13.69  31.21  53.54  73.59
19  11.97  28.15  51.7   73.48
20   8.5   24.36  47.61  70.75
21   6.33  20.11  43.58  68.37
22   4.34  16.61  38.41  65.28
23   2.73  10.95  32.63  60.8
24          6.67  23.09  53.98
25                17.46  42.19
26                       35.47

They are similar to the previous numbers.

Below is a table extracted from the above numbers that illustrates the
declarer's advantage in terms of percentage chance of making 9 tricks:

Marginal percentage increase of making 9 tricks with the stronger hand
as declarer.
Declarer's points (rows) vs Total points (columns)

    23    24    25    26
26                    4.91
25              3.47  5.39
24        0.6   3.76  5.28
23  0.59  2.12  5.15  4.98
22  0.61  2.88  4.09  5.42
21  0.94  2.64  3.96  2.67
20  0.88  2.68  3.08  3.26
19  1.5   2.63  4.13  2.53
18  1.06  2.81  2.65  2.75
17  1.14  1.12  3.65  2.09
16  1.61  1.72  1.29  0.92
15  1.21  2.36  1.41 -0.19
14  0.04 -0.45  0.16  0.1
13  0.77  0.62  0.22
12  0.16

 

The simulation data included the exact number of tricks taken for each
hand.  From this I could do an analysis of IMPs won in 3NT vs 1NT etc.
This sort of analysis overcomes the faulty assumption in the standard
analysis of the percentage needed for vulnerable or not vulnerable
games that the contract will never be 2 or more down.  I have included
tables for 3NT v 2NT and 3NT v 1NT both vulnerable and not vulnerable
(four tables).  3NT v 2NT means if the game is not bid then we will
play 2NT.  On the other hand if you can stop in 1NT then look at the
3NT v 1NT table - I understand in the bidding the decision is not that
clear cut.
         Vulnerable 3NT v 2NT
      23        24        25       26
26                                 0.0181
25                       -2.3235   1.0723
24             -3.3281   -1.5187   2.8819
23   -3.3006   -2.8257   -0.1822   3.9257
22   -3.1771   -2.1841    0.5968   4.6083
21   -3.1116   -1.8038    1.3859   5.0883
20   -3.0194   -1.2252    2.0004   5.479
19   -2.5713   -0.7265    2.623    5.896
18   -2.4116   -0.2849    2.8799   5.9115
17   -2.2063   -0.1583    3.1722   6.0101
16   -1.9658    0.1871    3.1635   6.0643
15   -1.8758    0.4674    3.315    6.0154
14   -1.7861    0.3564    3.3524   6.0872
13   -1.6628    0.5593    3.3208   6.0788
12   -1.6346    0.5556    3.3024   6.0703
11   -1.6425    0.4691    3.3289   6.0443
10   -1.747     0.4305    3.1002   5.9192
9    -1.8081    0.1537    2.9592   5.703
8    -2.0232   -0.0125    2.6293   5.5087
7    -2.185    -0.2538    2.5282   5.5131
6    -2.2774   -0.6547    2.0096   4.9796
5    -2.4744   -1.0417    1.5634   4.6833
4    -2.7435   -1.5544    0.8186   3.7951
3    -3.0093   -2.0936    0.0744   3.1625
2    -3.1939   -2.4775   -0.864    2.1163
1    -3.1688   -2.9765   -1.9547   0.3236
0    -3.272    -3.2555   -2.669   -0.6442

         Vulnerable 3NT v 1NT
      23        24        25       26
26                                -0.4301
25                       -3.4671   0.6517
24             -5.3021   -2.5239   2.5687
23   -5.858    -4.5715   -0.9378   3.6735
22   -5.5973   -3.5873    0.0304   4.4029
21   -5.2758   -3.0016    0.8649   4.9017
20   -4.8932   -2.3046    1.5228   5.2796
19   -4.3385   -1.6627    2.1886   5.7238
18   -4.032    -1.1587    2.4887   5.7395
17   -3.7097   -1.0037    2.7948   5.8417
16   -3.4506   -0.6165    2.7593   5.8889
15   -3.3192   -0.3236    2.9542   5.8362
14   -3.1975   -0.4308    2.9516   5.9084
13   -3.0508   -0.2377    2.9322   5.9078
12   -2.9954   -0.2158    2.8948   5.8911
11   -3.0173   -0.3365    2.9291   5.8653
10   -3.1694   -0.3727    2.7304   5.7426
9    -3.1915   -0.7099    2.5512   5.5032
8    -3.5028   -0.8949    2.2005   5.2947
7    -3.7052   -1.1958    2.05     5.3141
6    -3.8826   -1.6195    1.5198   4.7548
5    -4.203    -2.0957    1.0244   4.4685
4    -4.5607   -2.7408    0.2204   3.5209
3    -5.0289   -3.4502   -0.6558   2.8687
2    -5.3709   -4.0667   -1.7806   1.7019
1    -5.6222   -4.8969   -3.1487  -0.2346
0    -5.8664   -5.3707   -4.056   -1.252

         Not Vulnerable 3NT v 2NT
      23        24        25       26
26                                -0.7554
25                       -2.1965  -0.0372
24             -2.6616   -1.6732   1.1829
23   -2.4371   -2.3732   -0.8137   1.8857
22   -2.3941   -2.0146   -0.3237   2.3443
21   -2.4281   -1.8093    0.2069   2.6698
20   -2.4444   -1.4432    0.6199   2.9415
19   -2.1698   -1.134     1.038    3.222
18   -2.0961   -0.8454    1.2029   3.232
17   -1.9888   -0.7653    1.4017   3.2996
16   -1.8243   -0.5389    1.4035   3.3393
15   -1.7778   -0.3476    1.4965   3.3069
14   -1.7211   -0.4211    1.5314   3.3557
13   -1.6428   -0.2832    1.5078   3.3488
12   -1.6281   -0.2894    1.5004   3.3438
11   -1.628    -0.3424    1.5159   3.3263
10   -1.6885   -0.3695    1.3522   3.2402
9    -1.7411   -0.5433    1.2637   3.097
8    -1.8647   -0.6525    1.0413   2.9667
7    -1.963    -0.8048    0.9837   2.9656
6    -2.0029   -1.0747    0.6311   2.6051
5    -2.1059   -1.3177    0.3369   2.3983
4    -2.267    -1.6384   -0.1624   1.8021
3    -2.3903   -1.9671   -0.6416   1.3715
2    -2.4634   -2.164    -1.238    0.6813
1    -2.3553   -2.418    -1.9212  -0.5164
0    -2.379    -2.559    -2.3685  -1.1722

         Not Vulnerable 3NT v 1NT
      23        24        25       26
26                                -1.0893
25                       -3.0458  -0.3505
24             -4.1007   -2.4161   0.9514
23   -4.2403   -3.6449   -1.3703   1.6994
22   -4.1048   -3.0317   -0.7397   2.1934
21   -3.9671   -2.6809   -0.1764   2.5329
20   -3.778    -2.2302    0.2699   2.7951
19   -3.4325   -1.8137    0.7206   3.0962
18   -3.2505   -1.4787    0.9162   3.1057
17   -3.0623   -1.3776    1.1265   3.1767
16   -2.8801   -1.1212    1.1098   3.212
15   -2.8086   -0.9188    1.2343   3.1771
14   -2.7252   -0.9855    1.239    3.2257
13   -2.6304   -0.8591    1.2256   3.2254
12   -2.5935   -0.8436    1.2048   3.2131
11   -2.6015   -0.9239    1.2252   3.1962
10   -2.6975   -0.9492    1.0846   3.1124
9    -2.7224   -1.1658    0.9669   2.9518
8    -2.9163   -1.2884    0.7284   2.8108
7    -3.0414   -1.4858    0.6341   2.8203
6    -3.1411   -1.7724    0.2741   2.4407
5    -3.3329   -2.08     -0.0559   2.2406
4    -3.5571   -2.4992   -0.6008   1.6004
3    -3.8194   -2.9533   -1.1774   1.1548
2    -3.9925   -3.3113   -1.9082   0.375
1    -4.0639   -3.8015   -2.7977  -0.9294
0    -4.1782   -4.0831   -3.3913  -1.6246


Some comments:

Notation (x,y) will mean x points with declarer and y points with
dummy.

1. With 26 points it is almost always worth playing 3NT.

Vulnerable 3NT v 2NT only with 0 (declarer) opposite 26 (dummy) were
IMPs lost.

Vulnerable 3NT v 1NT - (26,0), (1,25) and (0,26) were negative.

Not Vulnerable any vulnerability (25,1) is added to the negative
results.

2. With 25 points bid game when declarer has:

Vul 3NT v 2NT - 3 hcp up to 22 hcp
Vul 3NT v 1NT - 4 hcp up to  22 hcp
NV 3NT v 2NT - 5 hcp up to 21 hcp
NV 3NT v 1NT - 6 hcp up to 20 hcp

3. With 24 points almost never bid game.  The only times where 3NT
gained IMPs versus 2NT or 1NT was when vulnerable and the decision was
between 2NT and 3NT.  Then it was only a winning proposition when
declarer had 9 up to 16 hcp.

4. With 23 hcp it was never a winning action on average.  They are
included because I had already done the analysis and because in
practice since we use wider range NT bids getting to 3NT with 24 or 25
points will mean that sometimes we get there with 23 or 26 etc.

The following assumptions and parameters were used in this simulation
and analysis:

a. The contract was never doubled

b. Declarer and dummy both had 4-3-3-3 or 4-4-3-2 distribution with no
major suit fit

c. The tricks taken were double dummy analysis calculated by Matthew
Ginsberg's GIB double dummy engine

d. For each hcp combination e.g. (15,10), 10000 hands were analyzed.
The hands were generated using the dealer program

e. Probably some other things that I have not thought of

[Siegmund]

I did a similar analysis in the winter of 2009-10, and reached very similar conclusions to Cascade's: http://taigabridge.net/articles/dd/hcpsplit.htm

 

I did allow a somewhat wider range of distributions (no 8-card major fit, but I allowed any balanced or semibalanced shape), and found a small profit in bidding 3NT vs 1NT with 24 points vulnerable at IMPs.

 

But it IS very close to break-even, sensitive to exactly what hand patterns you have, and, in the real world to how good of a declarer you are.

[Finch]

The problem with simulations is that they don't allow you to use any judgement at all. In the real world you can avoid playing in 3NT with KJx QJx Jxxx Axx opposite Qxx Axx Kxx Kxxx but bid it with Axx Kxx Jxx Axxx opposite xxx AQx AQ109x xx

[rhm]

...

But it IS very close to break-even, sensitive to exactly what hand patterns you have, and, in the real world to how good of a declarer you are.

Declarer has an advantage single dummy mostly because the defense does not always find the best opening lead.

I am not known to be very sceptical about double dummy simulations, but you have to interpret the results properly if you want to apply them to the table.

 

For some single dummy studies look at

 

http://www.rpbridge.net/9x54.htm

 

http://www.rpbridge.net/9x58.htm

 

Rainer Herrmann

[mike777]

The problem with simulations is that they don't allow you to use any judgement at all. In the real world you can avoid playing in 3NT with KJx QJx Jxxx Axx opposite Qxx Axx Kxx Kxxx but bid it with Axx Kxx Jxx Axxx opposite xxx AQx AQ109x xx

 

 

 

Thank you Frances. It may help to explain why example 2 bid 3nt ...example 1 don't.

 

I think we know the answer is example two is worth more than example one but why?

[rhm]

The problem with simulations is that they don't allow you to use any judgement at all. In the real world you can avoid playing in 3NT with KJx QJx Jxxx Axx opposite Qxx Axx Kxx Kxxx but bid it with Axx Kxx Jxx Axxx opposite xxx AQx AQ109x xx

Thank you Frances. It may help to explain why example 2 bid 3nt ...example 1 don't.

 

I think we know the answer is example two is worth more than example one but why?

I think the problem is not simulation but our point count system. It is very simple, which has merit for beginners, but it is not very precise.

If the first pair of hands is twelve opposite twelve, the second is thirteen opposite thirteen.

Simulations using point count are of course oblivious to the above, but at least you know that the effect between good and bad twelve counts will cancel each other out over larger samples.

I am always surprised how many otherwise good and experienced tournament players will rarely deviate from point count even on the most obvious examples.

I would not call this judgment, I call it stubbornness.

 

Rainer Herrmann

[NickRW]

Thank you Frances. It may help to explain why example 2 bid 3nt ...example 1 don't.

 

I think we know the answer is example two is worth more than example one but why?

 

2nd example has useful looking intermediates, the first does not.

 

2nd example has a 5 card suit in one hand, whereas the first has no shape at all.

 

2nd example has 9 controls, the first has 7 (which can be overrated at low level NT contracts - but is not be sneezed at - and does make a difference in suit contracts and high level NT contracts).

 

Various attempts have been made to quantify these things - but very very few people seem interested in counting quarters or fifths of a point. Bottom line is 2 decent 12 counts will make reasonable play for game, but you're pushing your luck if they're flawed.