Another Tale of Woe

[rhm]

Rainer, it's still matchpoints, so I don't know why you're using the chance of defeating the contract as the measure of success. One of the biggest arguments against a spade lead is that it will often cost an overtrick.

I did not. Sorry you misinterpreted my results.

The first number shows, how often a card is the best lead in terms of tricks the defense gets.

However, I expected some would then ask, but I want to know how often the lead will actually defeat the contract.

 

So I decided to run the analysis twice.

What you can see is that the club lead is the best lead, working optimally on 731 of 1000 deals.

However, a low diamond defeats the contract 4 times more often. (not entirely clear whether this is significant over 1000 deals,but let's assume it is)

 

From that you can deduce assuming you do not object to double dummy analysis:

 

Matchpoint-wise a club is clearly best, IMP wise a low diamond has almost the same IMP expectation.

 

Rainer Herrmann

[Antrax]

The result seems a bit suspect. Is it possible leading a singleton seems good because DD analysis always has declarer picking up the club suit?

[gnasher]

Sorry I misunderstood. I'll try harder to understand next time.

From that you can deduce assuming you do not object to double dummy analysis:

 

Matchpoint-wise a club is clearly best, IMP wise a low diamond has almost the same IMP expectation.

I certainly do object to double-dummy analysis that doesn't take account of the entire auction. Partner had an opportunity to double 2♣ and didn't do so. Don't you think this is likely to affect the result of leading a club?

 

I also have doubts about using double-dummy analysis in a case like this, because it doesn't take account of the information given away by the different leads. In real-life bridge, leading ♣9 or ♠6 will often help declarer to pick up the suit, because it tells him so much about the location of the spot cards. A diamond lead has some information leakage too, in that it may help declarer to pick up a doubleton honour in partner's hand, but it seems to me that this will occur less often than the hands where a black-suit lead gives away useful information.

 

It would, of course, be nice to see some of the hands where each lead appears to gain or lose.

[rhm]

I certainly do object to double-dummy analysis that doesn't take account of the entire auction. Partner had an opportunity to double 2♣ and didn't do so. Don't you think this is likely to affect the result of leading a club?

I very much doubt it. Partner is weak. Doubling two-level bids is really dangerous, the weaker you are the more, unless you have a really long and strong suit.

I often redouble these bids when our side is strong, usually for good results and as far as I can tell from watching top-level play these disasters are not that uncommon there too.

There is an example in the last BW where Fred doubled 2♦ and payed up redoubled (-1160, one overtrick, if I remember correctly).

 

The club lead does not gain by establishing anything. The defense has no further communication in clubs and partner is too weak to get his suit established and run.

The club lead gains by giving nothing away in a scenario, where declarer is a favorite to make his contract anyway.

Remember, no specific lead beats him more than 18% of the time double-dummy.

 

I also have doubts about using double-dummy analysis in a case like this, because it doesn't take account of the information given away by the different leads. In real-life bridge, leading ♣9 or ♠6 will often help declarer to pick up the suit, because it tells him so much about the location of the spot cards. A diamond lead has some information leakage too, in that it may help declarer to pick up a doubleton honor in partner's hand, but it seems to me that this will occur less often than the hands where a black-suit lead gives away useful information.

Of course a double dummy solver does not require any information.

You can argue that a passive lead tells declarer that on a restricted choice basis opening leader is more likely to have a missing honor in any other suit declarer is interested in.

 

True, but whether this information is more damage than the information leakage coming from an aggressive lead is very debatable.

If your partner is not trained that you lead passive frequently (I personally was so far quite prejudiced against leading singletons in notrump contracts) and woodenly returns your suit when he comes in, then such leads will loose a lot of their efficiency.

Passive lead defenses may or may not be more error prone than aggressive hit and miss defenses.

 

It would, of course, be nice to see some of the hands where each lead appears to gain or lose.

I tried to save all deals in LIN and PBN formats but got an error. In Lin format it said it overflowed.

I have never done it before.

I can certainly rerun the simulation, but I will not go through 1000 deals.

If I find time I will check whether I can manage to save single deals of interest.

Nobody is bared from using my simple specifications and do his own simulations and check the results for himself.

 

Rainer Herrmann

[gnasher]

Nobody is bared from using my simple specifications and do his own simulations and check the results for himself.

And you're not barred from providing simulation results in a form which allows us to judge their value.

[Phil]

A club lead might be right. It is somewhat passive and gives away very little - at T1. However, there are so many combinations where leading the 9 allows declarer to pick up the suit for one more trick that would not normally happen, and a DD sim completely ignores this. In addition, there is a tertiary effect of a singleton lead v NT that allows a good declarer to start methodically picking apart our hand.

[ggwhiz]

And you're not barred from providing simulation results in a form which allows us to judge their value.

 

Anybody that puts this kind of effort into the discussion is good with me.

[gnasher]

Just to show that it's not really very hard to do a simulation whilst providing the exact criteria and the hands:

 

[Deleted, because I also managed to show that it's easy to mess up a simulation.]

[gnasher]

Sorry about the false start. Here's my simulation:

 

http://www.barbu.co.uk/tmp/BBO_forums_20120709_lead_vs_3nt.xls

 

There are two sheets. On the first sheet you'll find 1000 hands, with the number of tricks on each of four possible leads (sorry Phil, I left out ♠8), average number of tricks, and frequency of beating the contract. On the second sheet you'll find my code, which is for Thomas Andrew's Deal.

 

I note that my results are different from Rainer's. This may well result from the different critera I used. I make no claim that my simulation is better than Rainer's (well, not yet anyway). I do claim that the information I have provided allows an sceptical reader to evaluate it properly.

[gnasher]

I note that my results are different from Rainer's.

In fact, we were measuring two different things. I was measuring the average number of tricks after each lead, whereas Rainer's measure of success, as I understand it, was "How often is this the best, or equal best, lead?" I suspect that both of these are wrong: we should be measuring the matchpoints gained or lost against what we think the field will do.

[gnasher]

Sorry about the stream-of-consciousness style. I've added a few more figures to my spreadsheet:

                                           S     H     D     C

Average tricks                          9.69  9.79  9.66  9.60

Beats contract                           203   183   217   204

Is best lead                             604   553   652   694

Is worst lead                            626   682   572   469

Matchpoints if everyone else leads D     -51  -107     0    79

[daveharty]

The title seems to suggest that we are going to be subjected to more unpleasant discards on the 5 rounds of ♣s that will likely follow at the opening lead <_<...

The title--perhaps I should have called this hand "A Comedy of Errors"--will become clear when you look at North's actual hand:

 

[hv=pc=n&s=s7ha943dat98ck832&w=saq952h76dj73c764&n=sj43hqj52dqcaqjt5&e=skt86hkt8dk6542c9]399|300[/hv]

 

So, in order, North (who is not normally a joker in any sense of the word):

 

1. Mis-sorted his hand

2. Mis-counted his points

3. Mis-responded to Stayman

 

I thought it was close between a spade and a diamond, but eventually went with my five card suit, for -400. This wasn't a bad result, as there were enough people making +420 in 4H to protect us from Mr. and Mrs. Feeble who couldn't get to game; but if I had led a spade we would have had the only plus score in the EW column.

[Zelandakh]

In fact, we were measuring two different things.

Despite the differences, it seems that you have both come up with the same answer, that a club is the best MP lead assuming DD play thereafter. Perhaps GIB is onto something with these short suit passive leads...

 

Also interesting is that your numbers are generally supportive of results on other hands of this type, that the 5 card suit is the better DD lead for defeating the contract but that the 4 card suit has a higher number of tricks on average.

[gnasher]

Despite the differences, it seems that you have both come up with the same answer

This probably wasn't clear from my ramblings, but I haven't yet come up with any answer.

 

As I may have mentioned before, I believe that one should be sceptical about double-dummy results, and that before drawing any conclusion one should examine some of the hands to check them for reasonableness and for any bias caused by the use of double-dummy analysis. I'm still doing that.

[Zelandakh]

This probably wasn't clear from my ramblings, but I haven't yet come up with any answer.

Hence my caveat "assuming DD play thereafter". I find it very unlikely that a club would be the best lead here in real play.

[rhm]

Sorry about the stream-of-consciousness style. I've added a few more figures to my spreadsheet:

                                           S     H     D     C

Average tricks                          9.69  9.79  9.66  9.60

Beats contract                           203   183   217   204

Is best lead                             604   553   652   694

Is worst lead                            626   682   572   469

Matchpoints if everyone else leads D     -51  -107     0    79

 

I have not looked at your result in detail, but see a lot of broad agreements with mine

Overall the club lead looks best on your simulation too, which is the real surprise, at least for me.

I have long been skeptical about the beginner advice to lead from your longest and strongest, particular at matchpoints.

But I used to avoid singleton leads even when partner bid the suit.

However, I suspect single dummy you need to be an above average defender to reap the full reward of a club lead.

You relinquish control to declarer without fighting.

You will have to make some crucial discards and we all know how difficult this can be when no source of tricks have been established for your side.

 

Rainer Herrmann

[JLOGIC]

Club is always the winner in DD simulations on hands like this. No surprise, DD loves passive leads from no honors even in short suits since declarer will guess everything correctly anyways.

[lalldonn]

You don't even need to run a sim to know the short suit lead wins double dummy on hands like this. You just have to observe that it's the lead GIB almost always chooses.