ATB

[beatrix45]

I happen to be interested in ordinary hands, because they crop up every day and how you handle them has a lot to do whether you win at money bridge or any other form of scoring.

The ones you seem to prefer, which crop up every other leap year, I do not care, not even for money bridge.

 

Rainer Herrmann

 

I completely agree with you that ordinary hands presenting ordinary bidding problems are by far the most important. The extensive discussion of this particular hand on this thread just seems odd to me. The north hand is a perfect 10-12 hcp Kamakazi one NT opener. If you are playing that system, then south can bid accordingly, ending in two hearts. Otherwise, north simply bid the hand like an untutored beginner, making every mistake in the book. She opened too light. She opened the wrong minor suit. She made a free bid of one NT in a competitive auction that by all bridge logic shows much better than minimum opening hand. The reasons for the poor outcome seem clear.

[rhm]

It would certainly surprise me.

 

There is some old DD data on RGB, from which you oculd conclude that 4333 is at least not worse than 4432 at notrump.

 

For example

 

http://groups.google.com/forum/#!msg/rec.games.bridge/pYo8MTID8Xw/cmazsS9Fg_4J

 

 

Rainer Herrmann

[PhilKing]

There is some old DD data on RGB, from which you oculd conclude that 4333 is at least not worse than 4432 at notrump.

 

For example

 

http://groups.google.com/forum/#!msg/rec.games.bridge/pYo8MTID8Xw/cmazsS9Fg_4J

 

 

Rainer Herrmann

 

For those willing to trawl through some of the data, we get a hint of just how irrelevant it is.

 

Look at the stats for "5322" (sic). Even these perform worse that 4333. Could it possibly be that DD simulations do not handle this sort of situation well? Or is it just that we have been over-valuing 5332 versus 4333?

 

We know from the stats available from the Richard Pavlicek site that in real play, declarer's single biggest advantage comes on the opening lead. After that, defenders give away fewer imps to par than declarer. If I lived in a world where the defence never erred on lead, I guess I have to prefer the 4333. And if I playing double dummy, I will have more two-way situations after the lead as well, whereas with the 5332 it's most likely just a race to set up the long suit.

 

It's true we were discussing 4432 versus 4333, but the principle is the same, if less extreme.

[rhm]

For those willing to trawl through some of the data, we get a hint of just how irrelevant it is.

 

Look at the stats for "5322" (sic). Even these perform worse that 4333. Could it possibly be that DD simulations do not handle this sort of situation well? Or is it just that we have been over-valuing 5332 versus 4333?

 

We know from the stats available from the Richard Pavlicek site that in real play, declarer's single biggest advantage comes on the opening lead. After that, defenders give away fewer imps to par than declarer. If I lived in a world where the defence never erred on lead, I guess I have to prefer the 4333. And if I playing double dummy, I will have more two-way situations after the lead as well, whereas with the 5332 it's most likely just a race to set up the long suit.

 

It's true we were discussing 4432 versus 4333, but the principle is the same, if less extreme.

What double dummy data suggests is, that distribution and long suits are not a deciding evaluation factor at notrump, that is has little impact on won tricks in general.

Of course it is easy to construct counter examples. If you have ♠A,♥A,♦JTxx,♣AKQJTxx you have 3NT in hand with 19 HCP.

Such hands exist, but are carefully constructed and not representative.

Double dummy results suggest adding one HCP for a fifth card in notrump is too much. People see the advantage of the fifth or sixth card, but not its disadvantage.

As far as I recollect a french study of actual table results came to the conclusion that the presence of a five card suit may affect actual table results by as much as a random ten on average.

The fact that you have a flat hand has almost no significant impact whether the other three hands round the table are flat.

So holding 4333 will almost have no impact on defenders race for 5 tricks except that if you are 4333, their lead can not be into doubleton, but could hit shortage in dummy or in RHO hand.

I also think the relative occurence for two way finesses is very little affected whether you have 5332 or 4333.

 

I agree Richard Pavlicek is right. But Richard's point is an observation, not a conclusion.

That is why I believe all this hogwash about two way finesse a double dummy declarer gets right is true but not relevant. It is just an observation, but the conclusion it skews results compared to single dummy results does not follow.

The important observation is again double dummy derives it tricks differently from single dummy. But this does not refute the observation that double dummy results may be a good proxy for single dummy results and that is what matters.

 

Rainer Herrmann

[johnu]

Except for the fact that KJx and even more KJ9x combos are greatly favoured by double dummy play

 

Having the jack in the same hand as a higher honor is makes them look good in simulations, single or double dummy. Opposite 3 small, you know the math. 25% - 2 tricks, 25% no tricks, 50% 1 trick assuming nothing unusual happens. Now look at Kxx. In simulations or real life, opposite Jxx, you need AQ doubleton onside to make 2 tricks without defensive help. Even opposite Qxx, you can only make 2 tricks without help if the ace is doubleton or singleton and you get it right.

 

Okay, double dummy you will never misguess a finesse or a suit distribution. But the defense will always lead the right suit, will always switch to the best suit in the middle of a hand, will never make a mistake which feels a lot more significant than knowing if a finesse is going to win.

[johnu]

rhm it's fine to believe "I trust that SD is not far from DD," but we were just wondering whether this difference relatively bigger or smaller in case of 4432 or 4333. These subtle effects were what I was confused about. Please don't just quote me out of context and make it sound like I'm/we're confused about everything. Thanks!

 

To clarify: do you think that based on the numbers we were presented with Kxxx KJx KJx xxx is indeed an average 11 count for notrump purposes and people who disagree are only doing so because they are nurturing their prejudices?

 

With a good double dummy analyzer and simulator, I expect that the hands will be perfectly analyzed and tabulated. I haven't seen any hands with analysis mistakes, and I don't remember seeing anyone reporting mistakes.

 

How good are single dummy studies? If the actual hands are from expert play at world championships or late rounds of US national team trials or knockouts as examples, assuming you can get a large enough sample size for a given hand type to decrease the variance, how do the results apply if you are an beginner or intermediate player? Sample size can also be a major problem with single dummy studies from high level competitions. Looking at DMPro simulations, you can easily get 5-10% variations for smaller simulation runs. If the results are from everyday results from BBO or OKB, how do the results apply if you are an expert since most of those results are will probably be coming from beginner/intermediate players who make lots of mistakes. Sure, a few hands are pretty easy to play, but most require guessing or are at least somewhat complex. How do those beginner/intermediate results affect the overall results when both declarer and defenders are making all kinds of mistakes? I think they more or less balance out but I'm sure they favor one side more than another.

 

I've seen studies where DD and SD results are pretty close for specific examples. I've done DD simulations on generic hands, e.g. 15 points balanced opposite 9 balanced, 15 points balanced opposite 10 balanced, and 15 points opposite 11 balanced, and the results are in the right ballpark for my expected results. I don't have anything available to check single dummy results and I have no idea how my results would compare to the participants in a theoretical single dummy study. Thus, I'm satisfied using double dummy results until something better comes along, understanding that single dummy results may be several percentage points different, one way or another.

 

As to whether the given 11 point hand is average for a balanced hand, if simulation results show it is average for trick taking, it's probably average. Is it an ugly 11 point balanced hand? Absolutely, but to tell the truth, when you deal out a bunch of 11 point balanced hands, most of them look ugly if you know what I mean :P 11 points spread over a balanced hand just look a little lonely to me. That being said, the 2 KJ combinations give this some trick taking potential so I would have guessed it wasn't far off from average even before a simulation. The simulation confirms what my initial valuation expected.

[cherdano]

As far as I recollect a french study of actual table results came to the conclusion that the presence of a five card suit may affect actual table results by as much as a random ten on average.

 

Well now that would be interesting and possibly convincing data.

[Zelandakh]

Well now that would be interesting and possibly convincing data.

As I recall, the study concluded that a 5 card suit was worth a little under 0.5 hcp (close to 0.4 I think). I always felt this matched the evaluation that you upgrade a hand for opening 1NT with a 5 card suit and another plus factor. I have never seen it put in terms of being worth a ten before. Anyway, if we work on the basis of 3hcp per trick, that makes a 5 card suit work ~0.15 of a trick. I could easily believe that DD biases for 4333 hands might be greater than this.

[rhm]

As I recall, the study concluded that a 5 card suit was worth a little under 0.5 hcp (close to 0.4 I think). I always felt this matched the evaluation that you upgrade a hand for opening 1NT with a 5 card suit and another plus factor. I have never seen it put in terms of being worth a ten before. Anyway, if we work on the basis of 3hcp per trick, that makes a 5 card suit work ~0.15 of a trick. I could easily believe that DD biases for 4333 hands might be greater than this.

And you do not believe that a random ten tends be worth on average 0.15 tricks or 0.4 HCP in notrumps?

To me that looks about right.

 

Rainer Herrmann

[cherdano]

So to summarize:

• rhm is aware of a double-dummy study in which 5332 hands do worse in 3NT than 4333 hands.
  
• rhm is aware of a single-dummy study in which 5332 hands do better than more balanced hands in NT
  
• rhm makes many disparaging remarks about those who prefer to use hand evaluation based on experience over blindly trusting double dummy results.
  

No comment.

[rhm]

So to summarize:

• rhm is aware of a double-dummy study in which 5332 hands do worse in 3NT than 4333 hands.
  
• rhm is aware of a single-dummy study in which 5332 hands do better than more balanced hands in NT
  
• rhm makes many disparaging remarks about those who prefer to use hand evaluation based on experience over blindly trusting double dummy results.
  

No comment.

This may be a summary of what you have understood.

But unfortunately you have understood very little of what I said.

My personal opinion is more differentiated than you seem to be able to grasp.

 

Rainer Herrmann

[Zelandakh]

Rainer, how about taking each of cherdano's 3 points in turn and saying where you agree and disagree. You did say on the other thread that 4333 was a negative so I am assuming there is not as much difference here as there appears to be.

[rhm]

Rainer, how about taking each of cherdano's 3 points in turn and saying where you agree and disagree. You did say on the other thread that 4333 was a negative so I am assuming there is not as much difference here as there appears to be.

Sorry, I certainly said enough in this thread, to give anyone who has an open mind enough information what my position is.

I do not think that Cherdano "summary points" deserve any further reply from my side beyond my previous reply.

 

Rainer Herrmannu