[cherdano]

hotShot, on May 17 2008, 06:45 AM, said:

cherdano, on May 17 2008, 09:15 AM, said:

Could someone please point me to the evidence that double dummy analysis is good enough?
And no, it is not sufficient to show me a study of 10,000 hands where on average, declarers took almost as many tricks as they would have double dummy.

Let me give an example. Han recently posted a double dummy lead analysis where leading from xx was exactly as good as leading from  Kxxx. My immediate objection was that the lead from xx assumes that partner will always find the right shift later, which is much harder when you led from a short suit than when you lead from one of your good suits. So maybe single dummy leading from xx is not such a great idea, after all.
I would be very surprised if the studies that Stephen and Richard cite  addressed my objection.

It's said that there are math gods, human number cruncher that can really calculate odds in bridge. Mere mortals like me use simulations to estimate the odds. Knowing that it is a very good estimate, is good enough for me.

Who says it is a very good estimate on a specific hand?

[helene_t]

Obviously, BridgeBrowser analysis has some flaws, simulation-based SD results some other flaws, DD analysis, Bermuda Bowl statistics and expert judgement yet other flaws. Ideally, all five methods should be applied simultaneously. When they all point in the same direction, we can probably draw a conclusion. When they don't, maybe it can be explained why a particular method gave deviating results.

FWIW, after having read "The wisdom of the crowds" I tend to have more faith in BridgeBrowser than in DD. Of course I may be wrong, but in any case which method is more adequate is likely to depend on the particular question being asked. I would be very surprised if the truth turned out to be anything resembling "BB is always right and DD is always wrong". (Or the converse).

[hotShot]

cherdano, on May 17 2008, 04:30 PM, said:

Who says it is a very good estimate on a specific hand?

Who's interested in the odds of a specific hand, when you don't know the specific hand?
I have to make my (bidding/playing) decisions based on the knowledge of a set of hand that the specific hand belongs to.
By a simulation I gain information about a subset of the set of hands I'm interested in. I hope that the decision that is right for most hands from my sample subset will be right for my real deal as well.

[cherdano]

hotShot, on May 17 2008, 09:02 AM, said:

cherdano, on May 17 2008, 04:30 PM, said:

Who says it is a very good estimate on a specific hand?

Who's interested in the odds of a specific hand, when you don't know the specific hand?
I have to make my (bidding/playing) decisions based on the knowledge of a set of hand that the specific hand belongs to.
By a simulation I gain information about a subset of the set of hands I'm interested in. I hope that the decision that is right for most hands from my sample subset will be right for my real deal as well.

Come on, it was obvious that I meant the results of a simulation knowing MY specific hand.

[sfbp]

cherdano, on May 17 2008, 07:21 AM, said:

hotShot, on May 17 2008, 09:02 AM, said:

Come on, it was obvious that I meant the results of a simulation knowing MY specific hand.

Ah - but when you try and apply ONE specific hand the simulator biases things by exactly how much you specify. Hard to explain this.. but the reality seems to be that BridgeBrowser allows you to see patterns in data that you didnt realise you asked for.

Whenever *you* overspecify a hand for a simulation, you have thrown out a whole bunch of hands that actually show the same underlying point that a more general approach would have done.

The advantage of looking at these in BRBR is that you get some feel for frequency. A treatment that only fits one hand in a million is great if you simulate it, but no use in practice - why? because you usually end up stealing some other (natural?) meaning more commonly assigned.

When someone (too lazy? to find it themselves) asks me a question, I generally end up answering some *other* question, heheh.

Sometimes the question is too general (eg something about the first 2 bids at the level of 1), sometimes too specific (all the people who overcalled 5C after an opening of 3H and a response of 4S). In the first case you can be sure that most actions are roughly neutral - interestingly any trends at all can be taken as quite significant since the sample is so huge; in the second case you might find one or two hands in the database.

Simulation doesn't give any weight to these numbers. BRBR does by showing the number of samples and the standard error of the mean of the matchpoint results.

[kenrexford]

I just deal out the cards until I have the hand I'm looking for.

God that takes forever.

[cherdano]

sfbp, on May 17 2008, 10:49 AM, said:

cherdano, on May 17 2008, 07:21 AM, said:

hotShot, on May 17 2008, 09:02 AM, said:

Come on, it was obvious that I meant the results of a simulation knowing MY specific hand.

Ah - but when you try and apply ONE specific hand the simulator biases things by exactly how much you specify.

????
I don't think I can say it more clearly but you and hotshot completely misunderstood my point.

Let me give you one example. A few days ago, I heard about a double dummy simulation for a competitive auction. The hand that involved one suit AKJxxx, and another of KJx, and very likely dummy would have very few entries. So the simulation told us this hand would take 12 tricks 124% of the time (numbers evidently made up). Double dummy. Well, with this particular hand I would expect the double dummy declarer to do much better than a single dummy declarer, since declarer might have to guess very well for which queen to use the entry. How much? I have no clue.

But since double dummy declarers on average take as many tricks as single-dummy declarers, or because of some other statistical study Richard and others know about, I don't have to worry about that??

[hotShot]

cherdano, on May 18 2008, 08:24 AM, said:

hotshot completely misunderstood my point.

Well at least I did not make my point clear enough.

I don't care about the actual deal, because I don't know all the cards.
But I can say that statistics over similar deals show that XXX is a good move for most of the similar deals. So if I don't know better, I will make the move hoping it works.

Maybe I will lose this actual board, but I know that most of the next times when a similar situation comes up, it will work in my favor.

[Trumpace]

cherdano, on May 18 2008, 01:24 AM, said:

Let me give you one example. A few days ago,  I heard about a double dummy simulation for a competitive auction. The hand that involved one suit AKJxxx, and another of KJx, and very likely dummy would have very few entries. So the simulation told us this hand would take 12 tricks 124% of the time (numbers evidently made up). Double dummy. Well, with this particular hand I would expect the double dummy declarer to do much better than a single dummy declarer, since declarer might have to guess very well for which queen to use the entry. How much? I have no clue.

But since double dummy declarers on average take as many tricks as single-dummy declarers, or because of some other statistical study Richard and others know about, I don't have to worry about that??

A simulation result does not equal proof. It is virtually impossible (IMO) that we can even prove that DD simulations give similar results as SD simulations in a great % of time. Best we can do is simulate, and there seems to be some empirical evidence which supports that statement, as hrothgar points out.

Simulations aren't the answer, they are a tool. It is upto you to give correct input, interpret and use the results of the simulation.

For instance, in the example you gave, what % of the hands will there be a guess for the Q (perhaps you can try simulating that)? Assuming the problem was vul small slam vs vul game in IMPS, even if that guessing % is say ~90% and the DD simulation says 100%, you should bid the slam. If you could tell that the guess % is more than, say 95% during the bidding itself, why do you even need a simulation?

If you can figure out something about the hand, which will affect DD vs SD simulation, incorporate it in your simulations. For instance, you could have done a simulation to find out the % of time you will need to guess the Q. This % combined with the DD result should determine your confidence in bidding the slam.

If you have absolutely no clue about a particular problem, then results of a correct simulation only increase you confidence in a certain solution to the problem. It still won't mean that it was the right solution, it would only mean that, it was the best solution you have based on the information at hand.

So if a DD simulation gives certain results, and there is empirical evidence that DD simulation gives approximately same results as SD, and you have absolutely no clue about the problem, then why won't you go with the results of the DD simulation? Giving just one sample situation (like the one you gave) does not contradict the DD = SD analysis which hrothgar was referring to.

What simulation to use is completely upto you. If you don't believe the DD vs SD simulation, why would you even try to do a DD simulation if you won't trust the results? For instance, Zia might possbily claim, that DD simulations are good for him, based on his table feel and all that... :-)

Again, simulations don't give proof. They are just a tool which provide information to help determine your confidence level in a certain hypothesis.

<That was a long post... sorry about that>

[jtfanclub]

Jlall, on May 16 2008, 11:16 PM, said:

And sorry, but by weak players I mean 99.9 % of online bridge.

Wow.

In your average Nationals or large Regional, what % of players would you consider "weak"?

[matmat]

jtfanclub, on May 18 2008, 09:09 AM, said:

Jlall, on May 16 2008, 11:16 PM, said:

And sorry, but by weak players I mean 99.9 % of online bridge.

Wow.

In your average Nationals or large Regional, what % of players would you consider "weak"?

is this a surprise to you?

have you not played online?

[matmat]

cherdano, on May 17 2008, 02:15 AM, said:

Let me give an example. Han recently posted a double dummy lead analysis where leading from xx was exactly as good as leading from  Kxxx. My immediate objection was that the lead from xx assumes that partner will always find the right shift later, which is much harder when you led from a short suit than when you lead from one of your good suits. So maybe single dummy leading from xx is not such a great idea, after all.
I would be very surprised if the studies that Stephen and Richard cite  addressed my objection.

this post, i believe?

ding

vs. 1nt-3nt:
Qxxx Kxxx Axx xx

I think a similar thing applies here that arend just pointed out.
The DD simulations show that it is better to lead from the Q than from the K. I am wondering if this margin of difference can't be overcome by noting that when you lead from the K, the declarer will find your Q anyway and will be able to pick it up more often than picking up your K (left to their own devices).

Another way of saying this is that if there is a two way finesse available against the queen, the declarer will always find it double dummy, whereas (to my knowledge) there is no two way finesse against a king in a no-trump contract. (hope this makes sense...)

[Cascade]

matmat, on May 19 2008, 06:47 AM, said:

Another way of saying this is that if there is a two way finesse available against the queen, the declarer will always find it double dummy, whereas (to my knowledge) there is no two way finesse against a king in a no-trump contract. (hope this makes sense...)

I recently had a two-way finesse against a king.

AJxx

Qxx

First I successfully finessed west for the king by playing low to the jack.

Then after some thought I decided to finesse east for the king by playing low to my queen. This also won when east popped king.

[FrancesHinden]

inquiry, on May 17 2008, 05:17 AM, said:

I will address the issue of the program Mikeh asked about, and Frances finds useless.

I said I found it useless in the context of this thread i.e. when used as a 'simulation' tool to look at questions such as "should we open 1NT with a 5-card major" or the example you gave "I also use the DB to answer some odd questions from time to time. Like after 3 passes, should you strickly adhere to Pearson Points in 4th seat"

You then explain other reasons you think the DB is a good idea. I haven't given my opinion of them (and I'm not going to in this thread).

[1eyedjack]

Cascade, on May 18 2008, 10:23 PM, said:

matmat, on May 19 2008, 06:47 AM, said:

Another way of saying this is that if there is a two way finesse available against the queen, the declarer will always find it double dummy, whereas (to my knowledge) there is no two way finesse against a king in a no-trump contract.  (hope this makes sense...)

I recently had a two-way finesse against a king.

AJxx

Qxx

First I successfully finessed west for the king by playing low to the jack.

Then after some thought I decided to finesse east for the king by playing low to my queen. This also won when east popped king.

There was a similar fine example in a book "The Best of Bridge" by Eric Jannersten containing a number of hands from what is known as the "Wohlin collection". Have lost my copy of the book which I lent to someone, but I vaguely recall one hand where you hold AJxx opposite KTxx of trumps and the only way to make is to finesse on the first round one way, which defender must duck holding Qxxx even though the finesse fails, and then declarer must finesse back the other way, in both cases before cashing a top honour. Would love to see that hand again if anyone can find and post it.

[brianshark]

Cascade, on May 18 2008, 09:23 PM, said:

I recently had a two-way finesse against a king.

AJxx

Qxx

First I successfully finessed west for the king by playing low to the jack. 

Then after some thought I decided to finesse east for the king by playing low to my queen.  This also won when east  popped king.

I could have sworn I read that in a book somewhere.