[benlessard]

---

http://www.bridgebas...-maybe-obvious/
In this post it was suggested that leading low might be better i was highly suspicious of this and decided to do my own sim with dealmaster pro.

For 2000 hands

East is 15-17 bal (may have 5M but not all of them)
West is 9-14 without 4M (i didnt want to include the 4333/3433 for this sim)

double dummy a high H win 458 while a low H win 417.

----------------------low H win & high H lose when
xx---A9xx = 17 times
A6---987x = 2 times (playing the A at trick 1 block the suit.)
k7x---A8xx = 2 times declarer can establish a 3rd H trick

------------------------High H win & low lose when
Hx(x)---H9x = 9
Hxx--H9 (H9x--H8, H98--Ax ) = 7
Hx--A9xx = 4
xx(x)--AK9(x) = 10
98--AKx = 1
9x--AKx(x) = 2 (playing the 9 cost nothing here)
Kxx--xx = 8



case i didnt count

AKx---98x where declarer has 50% of ducking
763---AK8 leading H declarer duck and impossible switch
H9x--Hx(x) or 9xx AKx = 11
Kx--xxx = 6
case where opener has a 1S opening.

Even if i allow 4333/3444 for responder i think its clear that a high H is better.

Also if the opps are playing puppet stayman than many hands where responder has xx in H and 3S are to be removed wich mean that low H make no sense anymore.

[bluecalm]

Quote

West is 9-14 without 4M (i didnt want to include the 4333/3433 for this sim)

I doubt it matters but why not ?

Quote

----------------------low H win & high H lose when
xx---A9xx = 17 times
A6---987x = 2 times (playing the A at trick 1 block the suit.)
k7x---A8xx = 2 times declarer can establish a 3rd H trick

Did you count it manually going through the hands (as I am always doing) or is there any way to extract hands where it matters and look only at them ?

[bluecalm]

Overall results of your simulation are very similar to mine. I think letting it run on matchpoints (amount of trick taken) is better but then again, there will be more counting of cases

[benlessard]

There is a yellow screen in dealmaster pro wich scroll all the hands (review or preview screen i think) in the lower left corner there is a winning lead diagram i just click fast at all the hands and stop cases where QJT or 54 (but not both) is lighted. It goes pretty fast what takes times is eliminating hands that are to "much double dummy" and sorting them. Dealmaster is not obvious to use at first (poor design IMO) but its really quick and has tons of worthy options.

As for the 4333/3444
In Montreal most of top opponents open 1Nt with 5M on a regular base, many of them have puppet stayman, ive wanted to compensated some of the hands where responder with 3S and xx in heart would make a puppet stayman. But since the vast majority of the low winning case are xx---Axxx (with or without 98 support) its clear that if they play puppet forget about leading low.

For doing the sim and counting the cases it took me around 90 min for 2000 hands and around 50 worthy cases. (but i kind of like to take my time for these kind of things)

[Siegmund]

Double-dummy sims against 1NT-3NT say to lead the queen even from QJ9xx and QJxxx.

Deciding how to compare double-dummy against single-dummy results is just about impossible without looking at the cases where single-dummy play is different than double-dummy play (for QJxxx, the big one is that there are a bunch of holdings along the lines of KTx on the table, or Txx on the table K8x in hand, where declarer has a guess single-dummy that he never gets wrong double-dummy. If you find a good method for evaluating the size of this effect, by all means let us know. I've never taken the time to work through enough sub-cases to come up with good proof either way.

[FrancesHinden]

Double-dummy simulations also ignore the rest of the hand. Take this example:

Axx
Kxx
QJx
Qxxx


K9x
QJx
Kxx
AKxx

If the opening leader has, say,
QJ10xx
Axx
xxx
xx

then double dummy it makes on either lead but single dummy it will go off half the time against good defence, and slightly less than that on soft defence.

[han]

Good point, bad example.

[JLOGIC]

han, on 2011-June-17, 05:10, said:

Good point, bad example.

Haha, we will have to teach Frances about the hold up play

[bluecalm]

Quote

Double-dummy sims against 1NT-3NT say to lead the queen even from QJ9xx and QJxxx

Not really. Results for 97 QJ543 A65 765
Winning lead:

7 ♠ - 105
Q ♥ - 153
4 ♥ - 169
A ♦ - 74
6 ♦ - 79
6 ♣ - 93

But: best lead:
Q♥ 668
4♥ 600

On the same sample which suggests that Q♥ is better at matchpoints which may or may not be true but for sure it's reasonable hypothesis (btw my theorem that one should lead the same at MP's and IMP's is taking some heavy blows in simuls I recently made).