[TimG]

hanp, on Dec 20 2009, 03:09 PM, said:

I have often done double dummy simulations for opening lead problems. They tend to favor passive leads in a ridiculous way, I think they are completely useless.

When comparing double dummy results to real world results, I found that the opening lead was all of declarer's advantage and in fact that the defense performed better vs double-dummy after the opening lead was made.

I was looking at average tricks taken (in various NT contracts played at IMPs), rather than frequency of one lead being better than another. So, it's quite possible that the passive lead produces:

9, 9, 9, 9, 9, 9, 9, 9, 9, 9 = average of 9 tricks

while the more aggressive lead produces

8, 8, 8, 8, 9, 9, 9, 10, 11, 11 = average of 9.1 tricks

but that you would want to pick the aggressive lead because it wins with higher frequency (at MP) or wins more IMPs.

It's quite possible that the real worlds results suffer from people sometimes choosing the passive lead, I made no attempt to classify any leads as passive or aggressive, standard or non-standard, etc.

[Mbodell]

TimG, on Dec 20 2009, 01:35 PM, said:

When comparing double dummy results to real world results, I found that the opening lead was all of declarer's advantage and in fact that the defense performed better vs double-dummy after the opening lead was made.

That isn't that surprising when I think about it. In general, the declarer has more control and more decisions and hence more chances to vary from DD correct play. The defense have fewer opportunities (although they may be more important) after the opening lead which is their big advantage on most hands (if they get it right) and big disadvantage (if they get it wrong). Plus the opening leader is the only person who makes a play decision looking at only 13 cards instead of 27+ (hand+dummy+played cards).

[eyhung]

Jlall, on Dec 20 2009, 01:04 PM, said:

The fact that double dummy likes a diamond better than a club shows in my mind that a diamond is going to be much better lead (which also backs my initial judgment that a diamond is much better than a club).

Actually, the sim results say the opposite : a diamond (10.49) is slightly worse on average than a club (10.45) DD, but the margin of error is significant enough that it's possible a diamond is better. (Higher avg # of tricks is bad for the defense, not good.) I took the position, unsupported by the statistics, that a diamond is better than a club, single-dummy.

If you want to give me constraints where LHO holds 4 spades or 5332, I'd be happy to add that to the simulation. But I believe that most people will investigate for their 4-4 or 5-3 major suit fit unless they know the combined count is at least 28 HCP -- and some always investigate. Even in the Blue Ribbons, I was cussed out by an opponent sotto voce when I raised 2NT to 3NT with a 4-card major (and a combined count of 28) and she led my major, giving us an extra trick over the field in 4M.

I can also try to rerun the sim using gnasher and TimG's statistic -- frequency of win is important and can be lost in the averaging.

[Jlall]

eyhung, on Dec 20 2009, 05:40 PM, said:

Jlall, on Dec 20 2009, 01:04 PM, said:

The fact that double dummy likes a diamond better than a club shows in my mind that a diamond is going to be much better lead (which also backs my initial judgment that a diamond is much better than a club).

Actually, the sim results say the opposite : a diamond (10.49) is slightly worse on average than a club (10.45) DD, but the margin of error is significant enough that it's possible a diamond is better. (Higher avg # of tricks is bad for the defense, not good.)

Misread this, thanks, I was actually quite surprised that a diamond lead came up better double dummy, so this makes sense

[Phil]

I had two similar opening lead problems over the last week, except the suit was Axxx instead of AJxx.

I'd be curious to know how much the presence of the J makes (my guess is a LOT).

[eyhung]

I looked at some of the simulated hands and realized that there were some hands where East (partner of opening leader) would overcall 1S, improving the results of the spade lead. After adding a primitive pass_over_1H filter for East (so no hands with 8 HCP and 5 spades, few hands with 6 spades, no takeout doubles of 1H, no 2C overcalls, etc.), and allowing North (dummy) to have 4 spades or 5 hearts in a balanced hand when the combined count is known to be 28 HCP, I get the following results for 1000 hands:


Spade lead = 10.461
SD = 1.430568252662393
Diamond lead = 10.663
SD = 1.2381292194128852
Club lead = 10.632
SD = 1.2783623157032638

Spade better than Club = 228
Club better than Spade = 270

Spade better than Diamond = 234
Diamond better than Spade = 252

Diamond better than Club = 36
Club better than Diamond = 61


However, if we go back to the simple dummy constraints, where dummy always tries to investigate for a 8-card major-suit fit if one is possible (1000 hands):


Spade lead = 10.215
SD = 1.4376515022222305
Diamond lead = 10.488
SD = 1.2959689562391281
Club lead = 10.459
SD = 1.3296943585226983

Spade better than Club = 262
Club better than Spade = 238

Spade better than Diamond = 265
Diamond better than Spade = 219

Diamond better than Club = 44
Club better than Diamond = 69


So it looks like a spade is actually the worst at BAM dbl dummy vs. Justin-type players, despite having the best trick average (you are more likely to gain multiple tricks on a spade lead, but the lead loses the board more frequently). Vs. typical Norths who always look for an 8-card major-suit fit, the spade is best.

[eyhung]

Phil, on Dec 20 2009, 04:25 PM, said:

I had two similar opening lead problems over the last week, except the suit was Axxx instead of AJxx.

I'd be curious to know how much the presence of the J makes (my guess is a LOT).

Same hand with the S6 instead of the SJ and non-Justin dummy :

Spade lead = 10.434
SD = 1.251883066109296
Diamond lead = 10.727
SD = 1.0842724042631777
Club lead = 10.652
SD = 1.1993057718256541
Spade better than Club = 232
Club better than Spade = 150
Spade better than Diamond = 242
Diamond better than Spade = 116
Diamond better than Club = 49
Club better than Diamond = 104

Looks like a spade is a huge winner now at BAM -- which confirms Evan's findings that AJxx is a really poor suit to lead from a trick-blowing perspective.

[TimG]

eyhung, on Dec 20 2009, 06:25 PM, said:

However, if we go back to the simple dummy constraints, where dummy always tries to investigate for a 8-card major-suit fit if one is possible (1000 hands):

If responder investigates with 3♣, that gives partner a chance to make a lead directing double some of the time that a club lead is best.

[sathyab]

eyhung, on Dec 20 2009, 06:25 PM, said:

I looked at some of the simulated hands and realized that there were some hands where East (partner of opening leader) would overcall 1S, improving the results of the spade lead.  After adding a primitive pass_over_1H filter for East (so no hands with 8 HCP and 5 spades, few hands with 6 spades, no takeout doubles of 1H, no 2C overcalls, etc.), and allowing North (dummy) to have 4 spades or 5 hearts in a balanced hand when the combined count is known to be 28 HCP, I get the following results for 1000 hands:


Spade lead = 10.461
SD = 1.430568252662393
Diamond lead = 10.663
SD = 1.2381292194128852
Club lead = 10.632
SD = 1.2783623157032638

Spade better than Club = 228
Club better than Spade = 270

Spade better than Diamond = 234
Diamond better than Spade = 252

Diamond better than Club = 36
Club better than Diamond = 61


However, if we go back to the simple dummy constraints, where dummy always tries to investigate for a 8-card major-suit fit if one is possible (1000 hands):


Spade lead = 10.215
SD = 1.4376515022222305
Diamond lead = 10.488
SD = 1.2959689562391281
Club lead = 10.459
SD = 1.3296943585226983

Spade better than Club = 262
Club better than Spade = 238

Spade better than Diamond = 265
Diamond better than Spade = 219

Diamond better than Club = 44
Club better than Diamond = 69


So it looks like a spade is actually the worst at BAM dbl dummy vs. Justin-type players, despite having the best trick average (you are more likely to gain multiple tricks on a spade lead, but the lead loses the board more frequently).  Vs. typical Norths who always look for an 8-card major-suit fit, the spade is best.

If you have a combined HCP in the range of 28-30 with two balanced/semi-balanced hands I think a 5-3 major fit produces almost as many or more tricks or at NT as it does in a major suit. There were several examples from LM pairs and Blue Ribbon pairs. But it's probably rarer for a 4-4 major fit not to outperform 3nt with normal breaks. So my feeling is that LHO is a lot more likely to have five hearts and still bid 3nt rather than 4-4 in majors.

[sathyab]

Thanks for all the input and the Sim results. Here's the full hand:

Scoring: BAM

1d-1h
2n-3n

---

♠ lead works a trick better than either ♣ or ♦. After a ♠ lead, double dummy declarer can still finesse the ♥J, then ♦s and make 10 tricks. But in practice he's likley to make only 9 and that's what happened at teammates' table.

Tired of leading fourth best from balanced hands and always wondering at dinner break why I did, I decided to go passive and lead a ♣. Partner took the King and saw no reason to attack ♠s. Had he had the ♠8, this would have cost us big. In any case he continued ♣s, taken in dummy, ♦ finesse. Soon declarer discovered that I had 3-2 in the minors and therefore not more than 4 ♠s and finessed ♥s confidently after partner followed to rounds of the suit, making 4

[MarkDean]

I guess this might be a really bad lead from nobody else even mentioning it, but I would lead a heart. I really hate leading from any four card holding which includes the ace, particularly on this auction, and I think there is a decent chance I will cut down on declarer's communication. Obviously, this may cost a trick in hearts, and yes it is the only suit where we know partner has fewer than 4 cards, but for some reason it appeals to me.