GF or limit raise?

[Cascade]

continuing the stupid question route --

 

I don't remember whether GIB takes the auction into account. Obviously this is irrelevant for the double dummy analysis, but does it matter for SD?

The auction is certainly part of the input process. For these simulations I put in the auction:

 

1♠ 2NT (consistent with GF)

4♠

 

So a standard Jacoby 2NT auction. As far as I know GIB understands Jacoby.

 

And as far as I know the program considers the auction when making its lead and plays.

 

I hope it does as one of the simulations that I plan to do sometime in the future is 1NT 3NT and 1NT 2♣ ... 3NT where no major has been found and see if I can measure the cost of having the more informative auction.

[han]

I find these results very interesting Wayne, the search that you propose would also be very interesting.

[lexlogan]

Very interesting. My initial reaction to the OP was that this was a nothing special limit raise. Cascade's simulations make a convincing case for game-forcing. Now the question is: is that primarily because of the known 9 card fit, or the 5 controls, or both?

 

In Goren's methods, a minimum opener with a five-card major would be a 5332 12 count. He'd add one distribution point "for the doubleton", really a short-cut method of counitng length. After a raise, opener adds one for the fifth trump. So a 5332 12 count revalues to 14 points, the 11 hcp plus doubleton raise counts 12, and you'd reach the magic 26.

 

Cascade, could you simulate (double-dummy) 1000 5332 12 hcp hands opposite a more average 11 hcp dummy? Say, Axxx KQxx xx Qxx ? Or randomly generated 4432 11 counts with four-card support? I'm curious whether ol' Charlie had this one nailed or it's just the extra controls that make this a game force.

[Cascade]

Very interesting. My initial reaction to the OP was that this was a nothing special limit raise. Cascade's simulations make a convincing case for game-forcing. Now the question is: is that primarily because of the known 9 card fit, or the 5 controls, or both?

 

In Goren's methods, a minimum opener with a five-card major would be a 5332 12 count. He'd add one distribution point "for the doubleton", really a short-cut method of counitng length. After a raise, opener adds one for the fifth trump. So a 5332 12 count revalues to 14 points, the 11 hcp plus doubleton raise counts 12, and you'd reach the magic 26.

 

Cascade, could you simulate (double-dummy) 1000 5332 12 hcp hands opposite a more average 11 hcp dummy? Say, Axxx KQxx xx Qxx ? Or randomly generated 4432 11 counts with four-card support? I'm curious whether ol' Charlie had this one nailed or it's just the extra controls that make this a game force.

A random 5-3-3-2 12 count opposite a random 4-4-3-2 11 count gave these numbers double dummy:

 

Tricks Frequency
0        0
1        0
2        0
3        0
4        0
5        0
6        1
7       21
8      163
9      434
10     335
11      44
12       2
13       0

 

381/1000 produced game. By itself this is close to being a reasonable game on average vulnerable at IMPs. The 185 times we failed by 2 tricks or more may make this slightly worse than the odds we need.

 

I am just running a second simulation where I am correlating the double dummy tricks with responder's controls.

[Cascade]

Tricks versus Controls:             CONTROLS  (Responder)      
                      0       1       2       3       4       5     Sum
              0       0       0       0       0       0       0       0
              1       0       0       0       0       0       0       0
              2       0       0       0       0       0       0       0
              3       0       0       0       0       0       0       0
              4       0       0       0       0       0       0       0
              5       0       0       0       0       0       0       0
              6       0       0       1       0       1       0       2
  TRICKS      7       0       1       6       9       3       1      20
              8       0      10      31      77      47      12     177
              9       0      13      57     171     148      31     420
             10       0       3      28     118     116      65     330
             11       0       0       5      13      22       9      49
             12       0       0       0       0       1       1       2
             13       0       0       0       0       0       0       0
Sum                   0      27     128     388     338     119    1000
Prob 10+                  0.111   0.258   0.338   0.411   0.630   0.381

[Cascade]

Similar data but with a three-card limit raise. Otherwise everything is the same.

 

Tricks versus controls:                CONTROLS (Responder)     
                      0       1       2       3       4       5Sum
              0       0       0       0       0       0       0       0
              1       0       0       0       0       0       0       0
              2       0       0       0       0       0       0       0
              3       0       0       0       0       0       0       0
              4       0       0       0       0       0       0       0
              5       0       0       1       0       0       0       1
              6       0       0       0       1       1       0       2
   TRICKS     7       0       2      11      17       8       0      38
              8       0       9      54      93      63      25     244
              9       0      11      66     159     167      52     455
             10       0       1      31      91      86      27     236
             11       0       0       1       5      18       0      24
             12       0       0       0       0       0       0       0
             13       0       0       0       0       0       0       0
       Sum            0      23     164     366     343     104    1000
       Prob 10+           0.043   0.195   0.262   0.303    0.26    0.26

[lexlogan]

Thanks! Looks like you need both the 9 card fit and excellent controls for a 23 hcp game with two balanced hands (one doubleton each.)

[Cascade]

Here are the final simulation results for 1000 deals played single dummy.

 

It appears that GIB generates hands that it thinks are consistent with the bidding and uses Monte Carlo simulations to find the most successful plays based on the results of those simulations.

 

Tricks  DD      SD
      0       0       0
      1       0       0
      2       0       0
      3       0       0
      4       0       0
      5       0       0
      6       0       1
      7       1       4
      8      55      85
      9     346     294
     10     459     450
     11     132     150
     12       7      16
     13       0       0

 

Trick Diff DD-SD
     -2      20
     -1     195
      0     587
      1     177
      2      20
      3       1
average   -0.015
std dev     0.735742

 

 Double Dummy (rows) versus Single Dummy (columns)
              6       7       8       9      10      11      12
      7       1       0       0       0       0       0       0       1
      8       0       1      32      18       4       0       0      55
      9       0       2      44     190      99      11       0     346
     10       0       1       9      78     295      71       5     459
     11       0       0       0       8      51      66       7     132
     12       0       0       0       0       1       2       4       7
              1       4      85     294     450     150      16    1000

 

There are no glaringly obvious new conclusions from the bigger data set.

 

Judging by how small the 'advantage' has been in this sample for Single Dummy over Double Dummy play - 15 tricks in 1000 deals - we would need a much bigger sample before we could make any meaningful conclusion about whether Single Dummy or Double Dummy produces more tricks for hands of the type contained in this simulation which were 5-3-3-2 12 counts opposite the hand in the opening post.

 

Single Dummy with GIB as declarer the simulation does suggest that the control rich 4-4-3-2 11 count in the opening post should be treated as a game force.