Good quantitative/statistical bridge books?
#1
Posted 2013-October-29, 09:58
Are there more books like these? I find myself more likely to believe suggestions based on statistics than I am to believe an author's gut feeling. Are there any other authors I should look for, or relevant academic papers that I can access without being at a university?
#2
Posted 2013-October-29, 11:07
wyman, on 2012-May-04, 09:48, said:
rbforster, on 2012-May-20, 21:04, said:
My YouTube Channel
#3
Posted 2013-October-29, 11:41
#4
Posted 2013-October-29, 17:52
rsm3000, on 2013-October-29, 09:58, said:
Are there more books like these? I find myself more likely to believe suggestions based on statistics than I am to believe an author's gut feeling. Are there any other authors I should look for, or relevant academic papers that I can access without being at a university?
I quit reading the second after counting up about 10 errors in the first 20 pages. The first book seemed pretty good. The work was so sloppy in the second that I wondered which data were good and which were wrong. Maybe some day I will have the time to finish the work I am doing and repeat the experiments.
Note: Not only is it not truly peer-reviewed, but apparently they can't even afford decent editors who can check the simple factual stuff. If they ever put up an errata page to go with the book, maybe I would return to it.
#5
Posted 2013-October-29, 20:02
GreenMan, on 2013-October-29, 11:41, said:
This is putting it rather mildly. When people begin to regularly match the performance of DF or whatever double-dummy solver shows up on their hand records, then I will believe that the matter is up for debate.
#6
Posted 2013-October-30, 04:10
rsm3000, on 2013-October-29, 09:58, said:
I doubt that you're going to find anything particularly useful.
If you approach bridge from an game theoretic or information theoretic perspective, you're quickly going to start veering into strong pass systems, relay methods and the like.
The first is pretty much banned where ever you go. The second is severely restricted in North America.
I did my first graduate degree in game theory.
The primary reason that I chose to work on poker rather than bridge was was the restrictiveness of the regulatory system.
#7
Posted 2013-October-30, 04:42
rsm3000, on 2013-October-29, 09:58, said:
Just like many studies in science have statistical errors, Better Balanced Bidding is based on flawed statistical evidence.
The examples from tournament play are chosen carefully and are biased.
For a short refutation, but based on sound statistical double dummy evidence, look at:
http://bridge.thomas...ions/cowan.html
Rainer Herrmann
#8
Posted 2013-October-30, 06:08
"The Mathematical Theory of Bridge", by Emile Borel and Andre Cheron.
This is from like 1940 or so and is basically a long statistical table with probabilities of breaks, shapes, suit combinations, some lines of play, etc. At the time a scholarly masterpiece (Borel was a mathematician of the top 10), it is now made more or less redundant due to computer simulations, which can crank out all those numbers in like 10 secs
It does have one extremely interesting chapter, which is about opening leads. Basically, it evaluated the odds of blowing a trick leading away from an honor in a "neutral" situation (see book for definition). It was based on those results that people came up with the "never lead away from a king" stuff, which was common lore for a long time. Then, as bridge systems evolved and more information was exchanged, people started to realize that king underleads were sometimes necessary (e.g. long suit in dummy).
It's interesting that the "winning xxx leads" series has results that are very, VERY much in common with Borel's book: ace/king/queen underleads are indeed statistically very dangerous, whereas jack or xxx underleads are realatively safe. So yes, Borel was proven right.
The book is hard to find in English. But if you can do with French, it's still printed regularly and you can order it from amazon.fr.
#9
Posted 2013-October-30, 09:18
In the realm of declarer play, there is such a thing as a "best" line of play, given what you know about the opponents, and there are many books that concentrate on helping declarer know the odds of certain things happening. The introductory books look at only one suit at a time. I know of only two that get serious about coming to grips with how two or more suits interact: Expert Bridge Simplifed by Jeff Rubens, and Bridge, Probability and Information by Ian MacKinnon. Some of the practice hands in the Rubens book are fiendishly hard. Some of the MacKinnon book goes far out of its way to take easy topics and smother them in a sea of words so they sound complicated. But they are what we have.
There are also many other books that, while not quite so rigorous, lay out the evidence why they recommend a certain method -- say it solves 2 or 3 common problems in exchange for causing 1 new problem -- and let you decide whether you believe them enough to try the new method. Journalist Leads by Rosler and Rubens is getting hard to find, but if you can find it, it does a better job of enumerating the (single-dummy) ups and downs of the different methods considered than any other leads book I can think of.
Finally there are experiments you can do yourself: let's consider how you accept two-under transfers like 1NT-2S showing clubs. Does "3C denies good clubs" or "3C promises good clubs" wrongside more contracts? Most people's intuition runs along the lines of "we're more likely to stop in 3C when opener denies a good fit, so we better make the in-between bid the super-accept and the transfer completion the signoff." In fact, you are going to stop in 3C when responder is weak. When responder is strong you're either headed for 3NT, or will play in 5C/6C with both hands strong and not be too worried about wrongsiding. And opener is much more likely to have a big club when responder has Jxxxxx than when responder has HHxxxx, just because there are more big clubs left for opener to have in the former case. In fact, "super-accept shows the strong club holding" wrongsides many fewer hands that stop in 3C, when it matters. You can work that out with a pencil and paper, or the aid of a simple sim to get approximate frequencies, or even by dealing out several dozen hands with a real deck of cards.
#10
Posted 2013-November-01, 16:37
Siegmund, on 2013-October-30, 09:18, said:
Finally there are experiments you can do yourself: let's consider how you accept two-under transfers like 1NT-2S showing clubs. Does "3C denies good clubs" or "3C promises good clubs" wrongside more contracts? Most people's intuition runs along the lines of "we're more likely to stop in 3C when opener denies a good fit, so we better make the in-between bid the super-accept and the transfer completion the signoff." In fact, you are going to stop in 3C when responder is weak. When responder is strong you're either headed for 3NT, or will play in 5C/6C with both hands strong and not be too worried about wrongsiding. And opener is much more likely to have a big club when responder has Jxxxxx than when responder has HHxxxx, just because there are more big clubs left for opener to have in the former case. In fact, "super-accept shows the strong club holding" wrongsides many fewer hands that stop in 3C, when it matters. You can work that out with a pencil and paper, or the aid of a simple sim to get approximate frequencies, or even by dealing out several dozen hands with a real deck of cards.
Great observation and I never understood why bidding the in between suit was taught as standard. Also bidding 2NT certainly doesn't prevent right siding when responder still wants to go beyond 3♣.
#11
Posted 2013-November-01, 17:24
Siegmund, on 2013-October-30, 09:18, said:
There is a far better reason to play that one bids 3♣ with good hands and 2N with bad (apologies in advance for the thread-jack, which I hope is only a minor diversion).
In standard bidding, with 4 suit transfers, 1N 2N is artificial, and thus with an invite in notrump, and no major, responder bids 2♣ then 2N over all responses by opener.
This means that the opps gain useful distributional information from opener, which information was of NO use to responder at all.
The way around this is to use 2♠ as a range ask/club transfer.
Opener bids 2N with all hands that would reject an invite to 3N, and responder passes or bids on (3♣ would be to play, and other bids confirm clubs and are gf).
Opener bids 3♣ with all hands that would accept the invite. Responder can pass, bid 3N, or bid something else...all other calls confirm clubs.
#12
Posted 2013-November-01, 17:37
rsm3000, on 2013-October-29, 09:58, said:
Are there more books like these? I find myself more likely to believe suggestions based on statistics than I am to believe an author's gut feeling. Are there any other authors I should look for, or relevant academic papers that I can access without being at a university?
As far as I know, the more common lead agreements aren't merely based on the gut feeling of any one or indeed any small group of players.
Leads have evolved over centuries....the concept of the opening lead was a relevant at Whist as it is at bridge, and such rules as the Rule of Eleven pre-dates our game by at least 40 years.
Do you know the principle of the simulation? Did the program assume single dummy play by declarer or double-dummy?
IOW, did the computer have only its hand, dummy and the cards played by the defenders at any point or did it 'know' all the hands?
If the latter, and I have always, tho perhaps mistakenly, assumed that is how most computer simulations analyze hands, then the results may be a curiousity but they bear so little resemblance to the game of bridge as to be essentially worthless.
For example, many, many times the effectiveness of the underlead of an honour will depend on the guess made by declarer at trick one. When the simulation ALWAYS has declarer guessing correctly, then the result is unrealistic and misleading. Such simulations will reward passivity on opening lead, whereas in the real world, many situations call for some degree of aggression.
#13
Posted 2013-November-01, 17:44
Bridge Odds for Practical Players (Master Bridge) by Hugh Walter Kelsey and Michael Glauert (Aug 2001)
Bridge, Probability & Information by Robert F. MacKinnon (Author)
Take All Your Chances by Eddie Kantar
Meta analysis:
Richard Pavlicek Odds & Theory
Richard Pavlicek Major Events
...get back to us if you find something!
Trust demands integrity, balance and collaboration.
District 11
Unit 124
Steve Moese
#14
Posted 2013-November-01, 22:48
mikeh, on 2013-November-01, 17:37, said:
Quite. Sometimes when the set of hands is particularly interesting, you can spend more time after the session working out how Deep Finesse got its results than you did playing the hands.
#15
Posted 2013-November-03, 06:47
I bought the leads book - it's not often I get mentioned in a book - but was disappointed somewhat in the methodology. Been meaning to write up a longer analysis, but I haven't gotten around to it.
As one of the first people to do serious double-dummy analysis, i'm always careful to avoid making blanket statements about expert practices. Sometimes, I think expert practice is right, but not necessarily for the reasons that the experts think they are right.
For example, there is apparently no different in play strength of the two hands ♠T8642 ♥AK4 ♦A83 ♣95 and ♠AK642 ♥AT8 ♦843 ♣95. This is counter to the standard expert view. But the first hand has a drawback. When you open 1♠ with this hand, partner will generally over-value the wrong hands and undervalue the right hands. For example, he will over-value ♠Q53 and undervalue a stiff spade. This makes the first hand a bad opening hand because, although it has the same playing strength, partner's re-evaluation of his hand's worth is going to be wrong. So expert practice to view the second hand as a "better hand" for opening can be correct, even if the expert belief that honors in long suits are stronger is (possibly) wrong.
I've always taken the view that if my research contradicts expert practice, I need to look deeper, not jump to any conclusions.
But then, I'm a mathematician more than a researcher, and I see my research more as puzzle.
In the opening lead research, the big problem is that, if you choose the double dummy best lead, partner can't take any inferences from your lead. For example, if you find a perfect lead of the jack from AJ32, the only way for partner to cooperate is for partner to have a double-dummy analysis, too, and imagine a bunch of hands where you'd make this lead, and then pick the defensive line which covers the most possibilities. That's a tricky proposition.
This is why every other book about leads emphasizes the *information* in the lead for partner. The idea of information is exactly what you lose when you use double dummy analysis, and bridge is primarily about information.
#16
Posted 2013-November-03, 13:36
If your goal is to satisfy your curiosity about certain aspects of bridge, then the books you listed are interesting, though their methodologies are a bit flawed. My feeling is that if double dummy analysis shows that something is overwhelmingly correct then I will just believe it, but if it suggests that something is quite a bit better then I will think about it and usually come to believe it is inconclusive or wrong.
However if your goal is to improve your skill at bridge, you want to work on the fundamentals. Stuff like this is interesting, but you should come back to it later.
I only bring this up because many youngish, enthusiastic players, especially the mathematically inclined, are eager to prove something about bridge (see hrothgar above). It is easy to get lost in this pursuit, for the worse.
#17
Posted 2013-November-03, 17:45
thandrews, on 2013-November-03, 06:47, said:
Does your research really say this? I am surprised and very interested to know more -- can you give more detail please?
#18
Posted 2013-November-04, 09:36
Vampyr, on 2013-November-03, 17:45, said:
I should have been more specific - there is no apparent double dummy difference between these two hands in terms of expected tricks.
I can't explain why this is true, except from data. It's possible that the double dummy trick values don't represent single dummy expectations. My explanation - that experts are right about what to do with this hand, but wrong about why - is really merely a hypothesis, not a proven fact.
#19
Posted 2013-November-04, 10:13
http://www.bridgebas...s-investigated/
But I realize this is a somewhat different question that the one we are discussing here. Obviously we don't know whether spades will be trumps if we pick the hand in question.