[Scarabin]

Vampyr, on 2014-March-04, 07:32, said:

Well, any line of play can be reduced to a percentage. So obviously by theory you mean something other than taking the highest percentage play or catering for the most layouts, or the most likely ones given the opponents' bidding.

So what do you mean exactly?

I will have to explain by examples. Anything else would be too obscure and boring.

I think the current theory of declarer card play may be regarded as a combination of analysis and knowledge, applied slightly differently to playing to the defenders opening lead(s) and to declarer's development of the hand. This approach seems to be totally historical and follows a well trodden path.

GIB introduced a different approach which might be characterised as search and evaluate or generate a sample and test double-dummy. GIB's approach however has little relevance to human players and could only be implemented with the help of a computer. It has achieved only limited success.

An approach which evaluates each card-state ( declarer & dummy's cards and any cards played & any inferences) as a numerical value and aims at monotonically increasing the numerical value of each successive card state, could perhaps be derived from computer analysis and provide an alternative theoretical approach. As far as I know no one has attempted this.

In a nutshell, that is what I mean (or at least my best attempt at an explanation).

[Scarabin]

Not entirely on topic but seeking information without opening a new topic: Some time ago Edgar Kaplan wrote a book on declarer play in which he repeatedly analysed a single hand which could be made by multiple strategies including, I think, both a squeeze and a throw-in.

Does anyone recall the book's title or perhaps the hand?

[nige1]

Scarabin, on 2014-March-17, 00:26, said:

Not entirely on topic but seeking information without opening a new topic: Some time ago Edgar Kaplan wrote a book on declarer play in which he repeatedly analysed a single hand which could be made by multiple strategies including, I think, both a squeeze and a throw-in.

Does anyone recall the book's title or perhaps the hand?

Perhaps Scarabin is referring to Edgar Kaplan's

Card Play at Contract Bridge (1964)

With the deal on the left
South is declarer in 3N on a ♠ lead.

---

[Scarabin]

nige1, on 2014-March-17, 19:30, said:

Perhaps Scarabin is referring to Edgar Kaplan's

Card Play at Contract Bridge (1964)

With the hand on the left
South is declarer in 3N on a ♠ lead.

---

Thanks nige1. I think it might be interesting to discover which strategy the robots would choose.

[Vampyr]

Scarabin, on 2014-March-04, 21:59, said:

An approach which evaluates each card-state ( declarer & dummy's cards and any cards played & any inferences) as a numerical value and aims at monotonically increasing the numerical value of each successive card state, could perhaps be derived from computer analysis and provide an alternative theoretical approach. As far as I know no one has attempted this.

This is untrue. People often recalculate the odds of a line based on the cards they have seen so far.

[Scarabin]

Vampyr, on 2014-March-18, 05:28, said:

This is untrue. People often recalculate the odds of a line based on the cards they have seen so far.

Not a lack of veracity, more that we are on different pages: you, I think are talking about how experts currently play the hand, I am dealing with theory as applied to how the great mass of players, the hoi polloi like me, are taught to play declarer's hand.

Let me first admit this is pure speculation. I am not actively researching play theory just reacting to one academic's comments on the present state of play theory.

It's not easy to formulate a feasible alternative to the current, traditional approach. If players were allowed to access computers (instead of note pads) then Ginsberg simulations might represent a viable alternative. This is not entirely fantasy I remember when it would have been unthinkable to allow calculators in examination rooms.

I also speculated that it might be feasible to program computers to generate random deals and investigate double-dummy which moves would increase the probability of making the contract. This would require enormous time so would have to be done in advance, and the results stored or if possible analysed into principles.

Take a simple example: when the opening card is led declarer may have a 50% chance of making the contract. After declarer holds-up this may increase to 100%. The player handles this by recognising the position, or by projecting the play (expert) or by rules (average player).

What I'm saying is that I do not think anyone has written a program to generate the best move for every possible/ probable situation.

I did warn you it would be boring!

[Scarabin]

nige1, on 2014-March-17, 19:30, said:

Perhaps Scarabin is referring to Edgar Kaplan's

Card Play at Contract Bridge (1964)

With the deal on the left
South is declarer in 3N on a ♠ lead.

---

Interesting hand to analyse with Bridgify. I tried it on 4 robots - Jack, Gib, Wbridge5, and Sharkbridge. All tried to set up hearts. Wbridge played correctly: 4 & A on opening lead to give 2nd entry to dummy and made oontract. All the others played Q or J on opening lead. Gib persevered by trying to create an entry in diamonds - down 2. Jack & Shark seemed to try to recover with Jack discarding the SA on the third heart trick, and Shark overtaking the SQ with the SA on the opening trick. Both down 3.

Nobody tried holding-up (ducking the spade lead in both hands).

Thanks again.

Added later: On the inferior opening lead of a club all the robots successfully set up diamonds. Shark flirted with a heart finesse along the way. Perhaps combining chances?

This post has been edited by Scarabin: 2014-March-22, 05:54

[nige1]

Scarabin, on 2014-March-19, 01:37, said:

I also speculated that it might be feasible to program computers to generate random deals and investigate double-dummy which moves would increase the probability of making the contract. This would require enormous time so would have to be done in advance, and the results stored or if possible analysed into principles.

AFIK, at single-dummy, Gib and other robots adopt such a tactic. Hence, the longer the programmers give them to think, the better the bots play. I believe the programs also include principles and heuristics. I'm unsure whether the programs themselves learn new principles by trial and error and by battling amongst each other, thus improving their evaluation functions for use on future deals. Early adaptive AI programs. (e.g. Arthur Samuel's Checkers program 1956) used that practical approach.

[Scarabin]

nige1, on 2014-March-17, 19:30, said:

Perhaps Scarabin is referring to Edgar Kaplan's

Card Play at Contract Bridge (1964)

With the deal on the left
South is declarer in 3N on a ♠ lead.

---

With two fairly obvious ways of making the contract, one may query why we should look for any other methods. However I think Kaplan's premise is that the contract may be made in any of four ways provided we keep to a cohesive plan.

I can see that the contract may be made by throwing East in with the third round of hearts, after eliminating diamonds and clubs, and forcing him to lead a spade into dummy's ten-ace. This requires West's co-operation in not rising with the Ace of diamonds and leading a spade or heart. Perhaps this can plausibly be accomplished by South's leading the diamond queen as if to force an entry to dummy?

The only possible squeeze is a criss-cross squeeze against East in spades and hearts. This requires the defenders' co-operation in taking their diamond honours on the first two rounds, cashing the two long clubs and leading only clubs or diamonds.

In these two cases, the opening lead is a spade and dummy plays the queen or jack. Now criss-cross squeezes must be the easiest to break-up, and many of declarer and the defenders' plays are less than optimal.

Despite my best efforts to co-operate in defence, I cannot get any robot to adopt these latter plans.

[Lovera]

Hallo. My library counts 29 books ( with 3 books in foreign language -i'm italian - and my knowledge of bridge techinque is learned by books of R. Trezel ( Le chaiers du bridge) very clear but also B. Romanet (Le squeeze au bridge, Bridge gagnant , Bridge total on planning ). About Love's text "Bridge squeeze complete or the winning endplay strategy " this basilar book , i must say , is different from other books 'cause non use schemes (that is right thing to do). However if anyone needs scheme i suggest too "The squeeze at bridge" by Chien Hwa-Wang (on clash squeeze). I had to suggest also the italian book "Eliminazione e colpi nel bridge" by Adolfo Giannuzzi and "The complete Stayman systeme of contract bidding" by S. Stayman. (My nickname Lovera is a tribute to Clyde Elton Love, my real name is Carlo Ingravallo from Bari, Italia).

[inquiry]

Scarabin, on 2014-February-18, 20:06, said:

The only genuine textbooks I can remember are Love's "Bridge squeezes complete", Mollo & Gardiner's "Card Play Technique" and, perhaps Culbertson's Blue Book.

whereagles, on 2014-February-17, 15:32, said:

Bridge is obviously under-studied. The academic work on it is scarce. Here are some of the milestones I know of
1. "The Mathematical Theory of Bridge", by Emile Borel and Andre Cheron. A bridge statistics book from the 1940s. It's out of print in english, but you can order it in french.
2. Vernes article on The Law of Total Tricks. Foundation of many modern bidding ideas.
3. "I Fought the Law", by Mike Lawrence and Anders Wirgren. Criticism of law of total tricks abuses.
4. "Winning NT/suit contracts leads", by David Bird and Taf Anthias. Major modern work on opening leads. Validates some of the results of Borel & Cheron.
5. "Partnership Bidding at Bridge", by Andrew Robson and Oliver Segal. One of the very few systematic books on bidding. A masterpiece.
6. "Capitancy for Advancing Bridge Players", by Denis Klein. Leadership principles on bidding and defense debated.

There were a number of post talking about (good) textbooks for bridge. I iown and have to agree with the one listed above (OTHER than "the mathematical theory of bridge" which I do not own). You can add some others to this list, show below which I do own. The most obvious missing ones are the first four in the following list (it is unthinkable The Rodwell Files was left off any such list).

1. The Rodwell Files: Secrets of a Bridge Champion, by Eric Rodwell and Mark Horton -- both defense and offense in play. A huge oversight to leave this one out.
2. Expert Bridge Simplified: Arithmetic Shortcuts for Declarer, by Jeff Rubens, A textbook to cover statistical approach to the game.
3. Guide Dog, volume 1 and volume 2, by Marti
4. Master Play - by Terrence Reese
5. The Play of the Hand at Bridge, by Louis Watson (reprinted as Watson;s classice play of the hand at bridge).
6. Bridge Play from A to Z, by George Coffin
7. Many but certainly not all of the "MASTER BRIDGE SERIES" (some actually suck) which includes the wonderful "Adventures in Card Play" (Ottik and Kelsey), and especially also "Blocking and Unblocking and Safety plays at Bridge (Reese and Trezel)
8. Entry Management, by David Bird and Marc Smith,
9. Bridge Probabilities and Information, by Robert MacKinnon (another statistical book. I preferred Rubens book (#2), because of better Bridge issues (and hands), but MacKinnon's textbook is much more clinical and perhaps for a serious student might be a better "textbook".
10. Mike Lawrence's guide books (Guide to.. "overcalls", "passed hand bidding", "doubles", etc), plus his book on Falsecards


I will add an odd one to this list, that would be suitable for some people, "Bridge Squeezes Illustrated" by Fook Eng. It is a mathematical approach to studying bridge, with lemmas (ok, theorems might be the word we would use), postulates, and (gulp) equations. The approach try to explain mathematically how squeezes work, it is clearly not for everyone, but it has to fall into the "textbook" category. I happened to like it (just liked love's squeeze book better).

Of course any bidding system book or book dedicated to one convention or one type of play are also text-book like.

[benlessard]

Im sorry ive missed this thread but I want to say that bridge is seriously undeveloped compared to other classic games.

Books like "Adventures in Card Play" & "Diaboliquement votre" are for childrens compared to some Edo ages Shogi and Go problems.

In 1755 there is a book of 100 shogi problems called Mukou written by Kanju Itou. Each of these problems will take me at least a 30 hours to solve (moving pieces is allowed and ive got 2 boards) and Im a pretty good player for a non-japanese. His brother also made 100 problems that are as tough and maybe tougher Zukou. One of the tough modern shogi problems is going to be tougher for me to solve than all the hands in Adv iCP combined.



For go its about the same thing there is some game commentary and analysis that will take me many hours of work to grasp what is really happening. The go problems are not as sharp as shogi problems but the openings (joseki) are out of this world. Each years in sure there is 2 or 3 Rodwell files equivalent in japanese language. In go there is also the chinese and korean books. Probably less that 1% of the go books get translated into english and there is more great go books than there is great bridge books.

In chess the database and books and openings repertoires are at least 10 times more developped than in bridge. In chess there is many books that will take 100+ hours to read them.