Another sim request

[Cyberyeti]

I picked up this awe inspiring collection:

 

[hv=pc=n&s=sa73h98652d93ct62]133|100[/hv]

 

Partner opened a 20-21 2N.

 

2 questions:

 

1: assuming partner will open 2N on any 21-22 5332/4432/4333, what are the frequencies for the different numbers of hearts in his hand.

 

2: I have the following alternatives:

 

Pass

Transfer to 3♥ intending to play there, playing in 4 if partner has 4+♥ or ♥Hxx/any suit HHxxx H=AKQ

Transfer to 3♥ and bid 3N, playing in 4♥ if partner has 3 or more

 

Which is best ?

[nigel_k]

Is it 20-21 or 21-22?

[BillPatch]

1. Assuming 0-36 hcp any 5332, 4432, or 4333 rounded to nearest .o1% 5♥ 3.47%, 4♥ 20.73%, 3♥ 47.37%, 2♥ 28.42% total 99.99%

Calculated to exact number of hands.

 

2. Thomas Andrews has simulated the outcomes of the auction 2NT all pass and 2nt pass 3nt for various combinations of honors.

double dummy using his program Deal. For xxxxx Axx Txx xx for a sample size over 600 he computes 3nt makes 35.6+-1.4% and 2NT fails 21.8+-1.2%

Therefore it was worth a raise to 3NT at imps only when VUL

I only ran a sample of 20 reps 20-21 hcp the given NT dist for N and the given hand for S

For the auction 2nt all pass 9+ tricks made 50% exactly 8 45% and one down 5%

The good spot cards on my simulation provide a non-random bias from the Andrews result, but my low sample size provides a fairly large random error

Only reaching game on the superaccepts reached 7 4♥ making games, losing to the direct 3NT raise 10.5 to 9.5 at MP.

The clear winner was the jacoby 3♦ followed by 3nt, making 13 games, going down 7 times, only once for -500. at MP beat 2nt 65-35, the direct 3NT raise 60-40, and only reaching game on superaccepts 57.5-42.5%.

[Cyberyeti]

Is it 20-21 or 21-22?

We actually play "good 19" to 21 but you can't sim that which is why I said 20-21.

 

Scoring was matchpoints.

[Siegmund]

This is an unusually challenging sim to automate because of the interesting superaccept criteria.

 

Note that cyberyeti specified returning to hearts with 4333 after 2NT-3D-3H, and superaccepting with 3433 after 2NT-3D, neither of which is everyone's cup of tea, but the sim is as requested. (And the sim was run before yeti's post below with further details.)

 

I can mostly confirm BillPatch's findings with a larger sample size. Given three strategies

1. Pass 2NT

2. Transfer then pass (ends in 3H or 4H)

3. Transfer then rebid 3NT (ends in 3NT or 4H),

 

and a sample of 1,000 hands,

Strategy 1 scored 32.9±3% of the matchpoints vs. strategy 2;

Strategy 1 scored 45.3±3% of the matchpoints vs. strategy 3;

Strategy 2 scored 51.7±3% of the matchpoints vs. strategy 3.

 

The difference between 2 and 3 was too small to be shown conclusively with only 1,000 hands.

 

Edited to add: with 10,000 hands, Strategy 2 is the winner, 51.9±1% against strategy 3.

 

I would not be surprised to see the winner between 2 and 3 depend on how opener chooses to bid his 4333s.

 

***

 

For those of you curious about the Andrews Deal 3.1 program, here is the script I used to do the sim:

 

source lib/score2.tcl

source format/none

 

north is {A73 98652 93 T62}

 

shapecond balanced5S {$s==5&&($h*$d*$c==18)}

shapecond balanced5H {$h==5&&($s*$d*$c==18)}

shapecond balanced5D {$d==5&&($s*$h*$c==18)}

shapecond balanced5C {$c==5&&($s*$h*$d==18)}

shapecond ntopener { [balanced eval $s $h $d $c] || [balanced5S eval $s $h $d $c] || [balanced5H eval $s $h $d $c] }

 

holdingProc hons {A K Q} {return $A + $K + $Q}

 

sdev 1v2

sdev 1v3

sdev 2v3

 

main {

set h [hcp south]

reject if {$h<20} {$h>21}

reject unless [ntopener south]

# final contracts: 2NT=1, 3H=2, 3NT=3, 4H=4

# South always continues after xfer-then-3NT with 3 hearts

set hs [hearts south]

if {$hs>=3} {set final3 4} {set final3 3}

# South superaccepts with 4 hearts, or Hxx hearts + HHxxx another

if {$hs>=4} {set final2 4} {

if {$hs==3 && [hons south hearts]>=1 && (

[balanced5S south] && [hons south spades]>=2 ||

[balanced5D south] && [hons south diamonds]>=2 ||

[balanced5C south] && [hons south clubs]>=2 )

} {set final2 4} {set final2 2} }

 

set th [deal::tricks south hearts]

set tn [deal::tricks south notrump]

 

set sc1 [score {2 notrump} nonvul $tn]

if {$final2 == 2} {set sc2 [score {3 hearts} nonvul $th]} {set sc2 [score {4 hearts} nonvul $th]}

if {$final3 == 3} {set sc3 [score {3 notrump} nonvul $tn]} {set sc3 [score {4 hearts} nonvul $th]}

 

if {$sc1 < $sc2} {1v2 add 0} {if {$sc1 == $sc2} {1v2 add 50} {1v2 add 100} }

if {$sc1 < $sc3} {1v3 add 0} {if {$sc1 == $sc3} {1v3 add 50} {1v3 add 100} }

if {$sc2 < $sc3} {2v3 add 0} {if {$sc2 == $sc3} {2v3 add 50} {2v3 add 100} }

{accept}

}

 

deal_finished {

set a1v2 [1v2 average]

set a1v3 [1v3 average]

set a2v3 [2v3 average]

puts "$a1v2; $a1v3; $a2v3"

}

Edited June 3, 2013 by Siegmund

[Cyberyeti]

Post removed because of a stupid bug in a sim. Repost coming soon.

I should have specified that we would break the transfer on a 3433 21 but not 20.

 

We would return to hearts on a 4333 with 3 hearts because of the danger of being unable to enjoy the heart suit in NT, we'd actually do it on a case by case basis at the table, but that's not useful for a sim.

[rogerclee]

I am curious why you feel the need to have such robotic rules regarding your superaccept style, I thought they were just your sim conditions but it sounds like this is how you actually bid. I much prefer my set of rules, which is:

 

1) Bid 4m if you think you have a very good hand with a trick source

2) Bid 4H with a very good hand without a clear trick source

3) Bid 3H otherwise

[Cyberyeti]

I am curious why you feel the need to have such robotic rules regarding your superaccept style, I thought they were just your sim conditions but it sounds like this is how you actually bid. I much prefer my set of rules, which is:

 

1) Bid 4m if you think you have a very good hand with a trick source

2) Bid 4H with a very good hand without a clear trick source

3) Bid 3H otherwise

 

We are not quite that robotic, and plenty of upgrading/downgrading goes on.

 

The rules on the 5/3 superaccepts are to avoid you playing in 5♥ with Kxxxx opposite Jxx and everything else, and also so that the weak hand knows Qxx in partner's suit is now running.

 

Do you never have 5♠/3♥ and bid 3♠ ?

 

We bid 3N with any minimum with 4 card support or maximum 3433, 4♥ with a maximum non 3433 with 4 card support and don't break with a minimum 3433.