[jdeegan]

karlson, on Mar 5 2010, 05:39 AM, said:

I think it's right, it's going to be very noisy with such a low probability of success. I ran 10000 again and I got 107 beatable, 10 must lead a spade, 69 anything but a spade. I can run a million overnight just for kicks.

I think the first 1000 was just an incredibly lucky set. I didn't change the code at all.

Obviously the probability of beating it in practice will be rather higher, since these hands include plenty of hands where they're just cold for grand and might try to bid it. Could try to do something a little more intelligent there, but I don't see why it would affect the spade/non-spade decision.

If someone wants, they can see the 107 hands from my last set where 6h was beatable and see if they think there's a significant double-dummy bias.

http://www2.decf.ber...jdeeg_prob.html. The winning leads are in the corner where the makeable contracts usually are.

♠ lead is a 10-15% play according to the simulations. At the table I think: "I know about half the HCP's - ♠AK, ♥ AKJ, ♦A and ♣ A or (maybe) K. About 20 HCP left and pard has 6- or about 1/3. So, Probability of ♠ Q about 1/3 minus a little. But, wait! Lottsa ♠Q hands where ♠ lead won't help (that's about as far as I got, and decided it was worth $20 US to get the damned problem off my mind for a while since I had long age solved the problem as a practical matter)."

So, if you had asked me, I would have thought 7 to 1 against a ♠ lead was a shade too high. Still, I maintain that it is a clear error much, much weaker than playing for the Q drop vs. finesse with eight trumps and no other info. There are some larger issues from this hand as well. #1 Don't underlead kings (esp. from length) vs. small slams w/o a plan, esp. vs. balanced hands and strength on your right. #2 Don't believe everything you read.

[jdonn]

Don't believe everything you read. Or see on a Mike Lawrence CD.

[jdeegan]

jdonn, on Mar 5 2010, 12:39 PM, said:

Don't believe everything you read. Or see on a Mike Lawrence CD. 

I am not saying I learned much from Mike's CD's - they are too elementary- but his lessons on opening leads really helped me. Lots of very fine players are not so good when it comes to opening leads.

Scoring: IMP

2♣-P-2♦-P
2NT-P-3♦-P
4♥-P-5♦-P
6♥-P-P-P

---

strikes me as a pretty good case in point. About half of the responders to this thread led a spade - which is a pathetic 7 to 1 underdog.

[jjbrr]

[OleBerg]

jdeegan, on Mar 6 2010, 05:57 AM, said:

...About half of the responders to this thread led a spade - which is a pathetic 7 to 1 underdog.     

We prefer the term "undercat".

[jdonn]

jdeegan, on Mar 5 2010, 10:57 PM, said:

strikes me as a pretty good case in point. About half of the responders to this thread led a spade - which is a pathetic 7 to 1 underdog.

And they are such accurate constraints too

[OleBerg]

Where this is headed:

[karlson]

You know, every once in a while someone comes along with a didactic and slightly condescending posting style, and tells us that we all suck at bridge. Now oftentimes the bridge arguments are laughably and provably wrong (think: foo), and it's not worth paying attention to. But sometimes they have a valid bridge point, and I don't see why the rhetoric should distract us.

I know many people like the spade lead, but mostly it's a matter of instinct, no one has a particularly convincing argument. Everyone is guessing, as they should be, since no one ever has enough datapoints from experience here. When the simulation(*) suggests so strongly that the spade lead is wrong, whatever caveats you apply to the DD analysis, I think it's worth rethinking. And I admit that on this hand I'm convinced. Neither declarer nor dummy is likely enough to have an outside source of tricks, and if they're marginal on high cards, one trick from the lead will often be enough.

(*)I've redone the sims with what I think are slightly more accurate constraints (gave 2n bidder 2245 hands, put an upper range on hcp, didn't let responder have 5-5 or 4-5 in the majors, tried to take away some hands that would texas. I didn't try to model what exactly the 5d bid should look like.) The spade still comes out at least a 3-1 to 4-1 underdog. You can look at hands here.

(**) ok, I noticed one bug, I only looked for leads that led to -1, sometimes there will be some leads for -2 that don't show up. If you think that this unfairly biases against a spade lead, I think you're crazy, but of course I'll redo it.

[rogerclee]

I called jdeegan an idiot, and I stand by that statement (this thread does little to counter that). However, I never said I thought his arguments for a spade lead were baseless. In fact in my first post I actually thought it was somewhat close (spade or club). It should be noted that among my friends, I am probably the least likely to lead aggressively (underlead a K in particular) against a small slam where I think my RHO is strong.

Anyway I have no problem with being told that a computer simulation probably says a club is right. It's certainly more convincing than anything else so far.

[gwnn]

I can't help but notice that most slams in karlson's simulation are terrible. Like responder cooperating with

xx
J9xxx
AQTx
xx

and then surprise, a spade lets it through. I am unhappy with these opponents and don't think most real life opps of reasonable quality would bid like this.

Maybe it's just because I'm biased for underleading honours or against condescending posts.

[gnasher]

gwnn, on Mar 6 2010, 06:32 PM, said:

I can't help but notice that most slams in karlson's simulation are terrible. Like responder cooperating with

xx
J9xxx
AQTx
xx

and then surprise, a spade lets it through. I am unhappy with these opponents and don't think most real life opps of reasonable quality would bid like this.

If they reach a slam which requires either a king onside or an underlead of the relevant king, and you'll nearly always lead from a king if you have one, doesn't that make it quite a good slam?

[gwnn]

good point. I will need to look at the hands again.

[karlson]

gwnn, on Mar 6 2010, 09:32 AM, said:

I can't help but notice that most slams in karlson's simulation are terrible. Like responder cooperating with

xx
J9xxx
AQTx
xx

and then surprise, a spade lets it through. I am unhappy with these opponents and don't think most real life opps of reasonable quality would bid like this.

Maybe it's just because I'm biased for underleading honours or against condescending posts.

Remember also that you're only looking at the ones that are beatable at all.

[jdeegan]

rogerclee, on Mar 6 2010, 11:45 AM, said:

I called jdeegan an idiot, and I stand by that statement (this thread does little to counter that).

At last something we can all agree on. You only agree with my wife and most of my bridge partners.
Two or three years ago this thread was a treasure for serious players, but imho it has been going down hill lately. Back then, whenever I had a bridge hand that baffled me (and they came up all the time), I would submit it and learn something useful. Nowdays, the post-mortem discourse is sometimes little better than one finds at the average suburban duplicate club. I just wanted to stir things up a little.

[jdonn]

karlson, on Mar 6 2010, 02:15 PM, said:

gwnn, on Mar 6 2010, 09:32 AM, said:

I can't help but notice that most slams in karlson's simulation are terrible. Like responder cooperating with

xx
J9xxx
AQTx
xx

and then surprise, a spade lets it through. I am unhappy with these opponents and don't think most real life opps of reasonable quality would bid like this.

Maybe it's just because I'm biased for underleading honours or against condescending posts.

Remember also that you're only looking at the ones that are beatable at all.

Something is still obviously wrong with the constraints. You have some, for example 2 and 8, where south is super-accepting on average (at best) 22 counts that clearly aren't worth it. You have others, for example 4 and 9, where south would rebid 2♠ with his 5-4 in the majors (at least most would). And you have quite a lot where I just don't believe responder would try for slam at all (1, 5, 11, 13, 24, etc.) And I believe all of those flaws are generally bad for a spade lead. Obviously if south has spades that is bad, and if NS are stretching a lot then rather than a need to take advantage of whatever strength we have, we more often have a need to not give anything up.

What were your constraints exactly? They certainly don't match the ones given in this challenge.

[effervesce]

Out of interest, after the transfer, superaccept, what did their 5♦ bid mean (apart from cuebid)?

Possible options:
1) Looking for a specific control (perhaps denying a club control)
2) Shortage
3) A sidesuit

I'd rule out option 1, since opener superaccepted: on a 2♣ opener the superaccept direct to 4♥ should show a control in each suit, as otherwise the opener could have cuebid something else as a superaccept.

Two is possible, and possibly the best method.

For option 3, where it shows a sidesuit, I think a spade lead would be much better than what the previous simulations have suggested as a passive lead can lead to spade losers being ditched on dummy's diamonds.

[karlson]

First I gave south 22-24 balanced (4333,4432,5332) with 4+ hearts, a spade control, and a club control, responder 5+ hearts, and NS 31+ hcp, which was basically the original constraint except that I didn't make the superaccept very restrictive.

Then I restricted responder's hands a bit. Gave him no 4-5 or 5-5 majors or 7 hearts, and gave him 9+ hcp or 8 or fewer losers. Allowed opener to have some 4522 shape also (hence those 4-5 majors hands).

Now I've tried a couple other things. I've beefed up opener, as you suggested (and the op said too). Gave him 23-24 with at least 8 controls, and at least 6 hcp in hearts. Took away the 4-5 and 5-4 majors hands.

I've tried giving responder real diamonds (4+). I've also tried beefing up responder, giving him only 7-loser or better (or 9+hcp) hands, with real diamonds. hands.

The spade never comes close to being a favorite. We can pick more at criteria, but I don't believe that slight changes are going to change it from a huge underdog to a favorite. Moreover, as far as I can see, most of these hands are pretty easy to play, so I don't think there's a huge DD bias, but of course one can always argue with that.

[effervesce]

Ok kalson,
I manually analysed the first 50 hands of your link (from the post above).
Given the constraints that opener would cuebid if missing a 1st/2nd round control in a suit (ie opener cuebids in a suit if missing AK in another suit), and my analysis of hands that would cuebid 5D as responder (if responder only bids 5D as showing a suit), I get the following results

1 - good, spade only lead
2 - good, spade only lead
3 - opener does not superaccept 4H - cuebids 3S
4 - opener cuebids 3S
5 - opener cuebids 4C
6 - opener cuebids 4C
7 - opener cuebids 4C
8 - opener cuebids 4C
9 - opener cuebids 4C
10 - good, non-spade
11 - cuebids 4C
12 - cuebids 4C
13 - cuebids 3S
14 - cuebids 4C
15 - good, spade only lead
16 - cuebids 3S
17 - good, spade only lead
18 - cuebids 4C
19 - good, non-spade
20 - dont think responder would cuebid 5D over 4H
21 - good, non-spade
22 - good, spade only lead
23 - good, spade only
24 - opener cuebids
25 - good, any lead
26 - opener cuebids
27 - opener cuebids
28 - good, spade only lead
29 - good, non-spade
30 - opener cuebids
31 - opener cuebids
32 - good, non-spade
33 - opener cuebids
34 - opener cuebids
35 - opener cuebids
36 - opener cuebids
37 - opener cuebids
38 - dont think responder would cuebid 5D
39 - opener cuebids
40 - good, spade only lead
41 - partner will lightner double!
42 - good, any lead
43 - opener cuebids
44 - good, non-spade
45 - opener cuebids
46 - dont think responder would cuebid 5D
47 - opener cuebids
48 - opener cuebids
49 - good, club only lead
50 - opener cuebids

That is, only 17 hands of the 50 qualify. Of the 17, 8 only a spade will beat it, 6 any non-spade, one only a club lead, 2 any lead.

Now, this is even with significant bias in the hands you gave - all of these are settable double dummy. When single dummy, ie when declarer won't know the diamond layout, I'd bet there are many more hands where a spade lead is best compared to other leads where a spade lead is necessary to set up a trick before spades are pitched on diamonds. With responder having hearts+diamonds I'll have to disagree with you and say that there would be a HUGE double dummy bias.

[effervesce]

Ok I've analysed the hands 51-100 and I get

51 - opener cuebids
52 - good, any non-spade
53 - good, any lead
54 - good, non-spade
55 - lightner double
56 - responder does not cuebid 5D
57 - good, spade only
58 - good, non-spade
59 - good, diamond only
60 - good, spade only
61 - good, non-spade
62 - responder does not cuebid 5D
63 - lightner X for dummy's sidesuit?
64 - good, non-spade
65 - good, any
66 - responder does not cuebid 5D
67 - opener cuebids
68 - good, club only
69 - opener cuebids
70 - opener cuebids
71 - good, non-spade lead
72 - opener cuebids
73 - opener cuebids
75 - lightner X?
76 - good, non-spade lead
77 - opener cuebids
78 - good, club only lead
79 - opener cuebids
80 - good, diamond only lead
81 - good, spade only lead
82 - good, spade only lead
83 - dont think responder bids 5D
84 - opener cuebids
85 - good, club only lead
86 - dont think responder cuebids 5D
87 - dont think responder cuebids 5D
88 - good, spade only lead
89 - good, spade only lead
90 - opener cuebids
91 - good, spade only lead
92 - opener cuebids
93 - good, non-spade only
94 - good, spade only
95 - opener cuebids
96 - lightner X!
97 - opener cuebids
98 - good, non-spade lead only
99 - lightner X?
100 - lightner X!

24 hands qualify, of which 7 only a spade works, 2 any lead, 8 non-spade, 2 diamond only, 3 club only

Summary: out of 41 qualifying hands from the first 100 hands from Karlson's link , 15 only a spade works, 14 only a non-spade works, 2 diamond only, 4 club only, 4 any lead.
And that's without taking into account any double-dummy bias; the hands only includes hands that are double-dummy down; other hands that may be single-dummy down since declarer may not pick the diamond suit (made more likely since partner's possible diamond honor(s) are over dummy's) and we may require a spade lead to set up a quick trick that is needed to be cashed when partner comes in with a diamond (or any other) trick before disappearing on dummy's diamonds (or elsewhere).

I don't suppose you could post up some hands where responder has a singleton diamond to analyse? I'd imagine it'd make a spade lead significantly worse than when dummy has a diamond suit.

Another interesting re-analysis is for opener having 3-4♥ for the superaccept - I'd imagine many players/expert partnerships would allow superaccepts to have only 3 card support over 2NT then responder transferring over 2NT. If the 2♣ opener is allowed to have only 3 hearts, then the likelihood that a spade lead risks a spade trick evaporating into thin air decreases (eg less chance of declarer having doubleton spade AQ for example and increasing chance of AQJ where a spade lead won't blow a trick anyway).

[OleBerg]

effervesce, on Mar 7 2010, 02:03 PM, said:

Ok I've analysed the hands 51-100 and I get

51 - opener cuebids...


... a diamond suit.

Nice posts.

They might even help jdeegan to "to gauge the quality of commentators."