[inquiry]

Misho and I were discussing classifying hands after the opponents open the bidding.

Of course there are balanced hands of various ranges, semi-balanced, one suiters weak, intermediate and strong, two suiters weak, intermediate and strong. And three suits same ranges.

So until someone expands on this start, lets call these,

bw, bi, bs (for balanced weak, intermediate, strong)
sw, wi, ss (for semibalanced weak, intermediate, strong)
1w, 2i, 1s (for one suited hand, weak, intermediate, strong)
2w, 2i, 2s (for two suited weak, intermediate, strong)
3w, 3i, 3s (for 3 suiter weak, intermediate strong).

Have we classified everything? Do we need very weak classification? Superstrong? Would a three suiter including their opened suit be considered a 2 suiter or a three suiter?

After classifying the hands, Misho and I want to discuss different bidding strategies with each type.

Ben

[pbleighton]

1) Do you classify 6332s as semibalanced or one-suited? Can they be either, depending on the distribution of honors?
2) I believe that you may be lumping together as two suiters 64 hands with a good 6 and 4 small with 54 hands where the 4 is stronger than the 5. If so, is this appropriate/useful? Generally, what is a "suit"? It seems to me that if it is not simply 4 cards, that the criteria would vary by strength ranges.
3) "Would a three suiter including their opened suit be considered a 2 suiter or a three suiter?" Good question - but applies equally well to two suiters. Also really applies to 4432 - much different I think if the 2 is in the opps suit than if a 4.

[inquiry]

Pbleighton,

Your questions get right to the heart of the discussion that Misho and I wanted to get started. First, to classify the hand types (I listed the obvious ones to start the thread). Now you have added three more...

s2s = semi two suited (a 6-4 or 7-4 perhaps)
o2s = opponent 2 suiter (one suit is theirs)
o3s = opponent 3 suiter (one suit is theirs)

As for how I would classify a 6-3-3-2 hand, I would call it a miss-deal. Now if it was 6-3-2-2, I would treat it as semibalanced or one-suited, just as you guessed. But maybe a one suited hand should always include a short suit (singleton or void).

Ben

[Cave_Draco]

If the hand is 2-6-3-2? One suited!

I would extend the distinction that I apply to 5-3-3-2s, if the 5 is a minor, it is balanced.

If the loser count is <6+ & the Milton Work count <15? at best semi-balanced; never explored the correlation but it seems a starting point?

[mishovnbg]

Thank u for opening this topic Ben. It isnt only theoretical discussion, target is to make SYSTEMof competitive bidding, rather then a heap of agreements. I know that it is very difficulte, but if we never try we never do it. As i read in one topic here author ( sorry cant remember name ) wrote that bridge system cant be make by method of mistakes-corrections because this way can lead to falshe optimum and i agree. Right way is "From top to bottom" and it lead to use objects of bridge bidding, its properties and methods. Some of systems of bidding without intervention was made same way. But i never meet complete SYSTEM of competitive bidding. But to make it, we must furst define and classify situations (objects of bidding) that appear during competitive bidding and its characteristics. I cant do it and hope that BBO forum members will help. I now will share few of situations i already identify, but i am not sure that i am right.

Classification of competitive situations ( naive try :)
1. First open: ( Ben - Inquiry correction )
1.1. No ( System )
1.2. We ( opp Intervention )
1.3. They ( our Intervention )
2. Vul: ( Ben - Inquiry correction )
2.1. They
2.2. No
2.3. Both
2.4. We
3. Level reached: ( Peter - pbleighton correction )
3.1. 1 level
3.2. 2 level
3.3. 3 level
3.4. 4+ level
4. Strength shown:
4.1. Score- (Pre-empt)
4.2. Score+ (1/1)
4.3. Game try
4.4. Game forcing (2/1)
4.5. Slam try (splinter)
Questions: Is strength (HCP) is most important? We must define other or subsituations for: distribution of strength between partners; statistical expectation for HCP; side which show strength; extent of limitation;...?
5. Distribution shown:
5.1. No major fit shown:
5.1.1. No majors shown
5.1.2. One major shown
5.1.2. Both majors shown
5.2. Fit shown:
5.2.1. One side shown major fit
5.2.1.1. One major shown
5.2.1.2. Two majors shown
5.2.2. Both sides shown major fit
Questions: How important are minors and must we include it in classification and where? Is length of majors lead to different situations or subsituations?
6. Relations between strength and distribution - offensive (off) / defensive criteria
6.1. No off hand shown
6.2. One side off hand shown
6.3. Both sides off hands shown
Questions: Best books i read about is Robson&Segal "Partnership bidding" and Lawrence "Evaluation of hands". I just feel a "black hole" here, even in noncompetitive systems and agreements?
7. Exceptions: ( Peter - pbleighton correction )
7.1. Preempts
7.2. Strong conventional opening
Questions: Exceptions only establish rules 8). Can u add more, but general, because if many it become a rule ;D

Misho

[pbleighton]

1) "...use objects of bridge bidding, its properties and methods" - do I smell a fellow computer programmer here?

2) This will get very hairy (and may turn out to be impossibly complex), but hairy can be fun.

3) Here is my naive attempt to put some hierarchical structure into the discussion:

Are you initially restricting your analysis to auctions where the bidding started with a non-preemptive bid, and the second side to bid entered with a double or simple overcall? If not, the most important classification may be auction type:
a) Neither side preeempted.
First to bid was a preempt.
c) Second to bid was a preempt.

Preempts suck up so much bidding space that they must be treated differently. A preempt by the second side to speak is different than the first, since the first (usually stronger) side was able to at least give some information. It is therefore somewhere in between the other two types.

I believe that his should be the "parent" object, with at least three "child" objects beneath it.

Then game invitation, game force, slam try, etc. are objects under the above objects - you are going to need hierarchical relationships, or your individual situations will multiply beyond control. This will be essentially strength - more specifically the balance of strength between the two sides.

Distribution/fit are objects at the same level as strength, both having parent/child relationships to the top level object auction type.

The third "child" object is level of bidding, or number of bids, or some measure of where in the auction's progress you are. Theoretically this isn't necessary, but in practice it will be, or else again the number of situations will quickly multiply beyond control.

Yours in bewilderment at complexity,

Peter

[pbleighton]

It strikes me that I indulged in tech-speak too much in my previous post. Two clarifications:

1) I am not implying, nor do I think Misho was, that we should try to write the specs for a computer program. However, system design has many similarities, whether you are designing computer programs, bridge bidding systems, or cars.

2) To clarify "objects, methods, and properties". I will use a car example. A "car" object could be considered the parent object (comparable to the "auction type" object above, with engine systems, transmission systems, and wheel/axle sytems as child objects (comparable to the strength, distribution, and auction progress objects above).

Misho mentioned methods and properties. Neither he nor I addressed them directly, but in case the discussion moves in this direction, here are car analogies.

Methods are actions which the objects (parent or child)can perform. The engine object has a start method. The wheel object has a turn method. The car object has a start method, which has a relationship to the start method on the engine object.

Properties are data about an object. The engine object has a horsepower property.

[mishovnbg]

I like "car" example, i also give it to my students, when i teaching them to work with computers ;D.
Yes, i am programmer
Yes, properties of bridge objects like: hcp, distribution, OD criteria, level... can help partners to idenify situation SAME WAY and have knoledge what METHOD they use in such SITUATION.
Yes, METHODS of bidding are agreements we have how to bid in different SITUATION (different opp dids like MULTI for example).
Thank u Ben for try of classification of type of distribution - we need it when we will build our METHODS against different SITUATIONS.
I will correct my classification of situations and will add with name of friend that help.
Ben we can start to write also methods for different situations. Different players use different methods ofcourse, but knoledge about their methods can only help.
Misho

[mishovnbg]

To classifying the hands we need some general terms and then reduce to absolute distributions and hcp described by Ben and Peter ( i will shortly call it BPC). I will try to write it.

1. Destructive: dont have hcp for own contract
2. Constructive: have hcp for own contract
3. Defensive: prefer opp to play contract
4. Offensive: prefer to play own contract

With this classification we have 4 general type of hands: DD(1+3), DO(1+4), CD(2+3), CO(2+4).

Now BPC looks: ( all of them can be modified with "o" - opp suit)

1. DD: bw, sw
2. DO: sw, 1w, 2w, 3w, s2sw
3. CD: bi, bs, wi, ss, 1i, 1s, 3i, 3s
4. CO: wi, ss, 1i, 1s, 2i, 2s, 3i, 3s, s2si, s2ss
Notes: same type in different general type mean that type can elavuate to other by vul, position... not clear yet ofcourse >

Absolute type of hands Ben-Peter classification:
bw, bi, bs (for balanced weak, intermediate, strong)
sw, wi, ss (for semibalanced weak, intermediate, strong)
1w, 1i, 1s (for one suited hand, weak, intermediate, strong)
2w, 2i, 2s (for two suited weak, intermediate, strong)
3w, 3i, 3s (for 3 suiter weak, intermediate strong).
s2sw, s2si, s2ss = semi two suited (a 6-4 or 7-4 perhaps - weak, intermediate, strong)
o2s = opponent 2 suiter (one suit is theirs)
o3s = opponent 3 suiter (one suit is theirs


In different SITUATIONS same hand can change it type - evaluation of hands( for example after preempts u jumps are not preempt ). We must first pick up a SITUATION.

I will describe my METHOD of bidding in SITUATION:
1 level of bidding, Score-(weak opening like Boian precision), 1 major shown - not yet fited, 1 offensive hand shown, no exception.

1. CD hand i will bid DBL or trap pass
2. DD hand i will bid Pass or bluff
3. DO i will bid overcall !!! or 1NT !!!:
4. CO i will bid jump overcall, 2NT, cue bid, jump cue bid, 3NT.
Note: I didnt describe exactly types because is only example. Else i must write much more:)
Misho

[hrothgar]

I'm not sure whether the approach that you are suggesting is feasible.

I don't think that hands can be nicely lumped into catagories like "Defensive" or "Offensive". This is a complex multi-dimensional construct which you are attempting to simplify into a binary expression.

Furthermore, I disagree with your definition of "Destructive" and "Constructive". Destructive is a very loaded term, which is typically applied by individuals seeking to achieve political ends. More significantly, the bridge scoring table often rewards me even if I achieve a "negative" score. 5HX- 2 might be an excellent result, if it does score up as -300.

At the very least, your analysis needs will need to be many times more complex to allow for the existence of mixed strategy equilibriums.

[mishovnbg]

Thank u for join to our discussion Richard.
I agree that any classification is very difficulte.
I know that my terms is not good, can u help me with better terms, please. I chose it, because most of players use terms like "offensive" and "defensive" how i wrote. I prefer "Active" and "Passive" for distribution and "offensive" and "defensive" for hcp.
I agree that great game Bridge :-* is "a complex multi-dimensional construct" like life. But ulike life, Bridge have less usefull "space" for analysis and may be possible to classify. All best ideas in human history was "simple", like wheel or E=mc² :.
Behind the word, when you use it as term, you use definitions of meaning, else u must invent new words :-
In terms i defined bid "penalty double of opponents contract" is Constructive-Defensive.
We here discuss exactly how to reach better partnership undersanding and have same nice results, like 5HEX-2 sacrifice.
I hope you help us to make this or another classification. I know you have great experience with realy systems and your friendly analysis is welcome. Thanks again.
Misho

[the hog]

I think some classification is useful eg general principles to be used over a multi, or perhaps over a 2 suited opening where 1 suit is specified and the other isn't. Similarly over a strong C - Precision, Moscito, and then over nebulous C such as Polish C.

If you have general principles and defenses available the partnership should be able to apply them in comparable situations.

For example -
Over a weak 2 including a multi 2D, we play that 2NT shows 15+ - 18
Pass followed by a x is takeout of that suit

Over a 2 suited 2 opening we play that 2NT shows -17 to 19
Why the range difference? When one player has shown 10 cards in 2 suits, our fits will not break, therefore we need a bit more high card strength to make the same bid. We apply this to any 2 suited opening regardless of its nature.

I think this is what you are talking about Misho, isn't it?


Ron

[mishovnbg]

Thank you The_Hog for joining to our discussion , Ron. I read about you. (Thanks to Victor Mollo for...) ;D
Yes you wrote what i mean. I noticed that i play similar in many positions. Next step is to classify it. Classification must be not too complex to be usefull. So i hope BBO members to help me :
Misho

[inquiry]

This seems a very complicated project. Let me make a few comments to see if I am on the right track. First off, for the development of these concepts, we have to consider, in effect, did we open, did they open, and what type of bids have been show to date. I think you also have to consider what Robson/Segal refer to as "potentially competitive auctions" and rather or not your side has a known fit yet, and yes, rather or not they have a fit. So let’s hand fits (ours and theirs first). How does this change the bidding. For instance, assume the bidding has gone… (1H)-1S-(2H) to you. Would your choices, lets just assume two possible bids 2NT and DBL, be different than if the bidding had gone 1S-(2H) to you (they have no known fit. There is also the positional question. In the first auction, the "REAL" heart suit is behind you, in the second, you hare sitting behind the heart suit.

Clearly in the first example (hearts opened to your left) you are much LESS likely to want to defend 2H. First, for positional reasons. Second, they have a known fit. So playing your CD, DD, DO, and CO rules of engagement (see Misho reply 9 ), it is unlikely that with an defensive hand (DD or CD). Therefore double here should be an offensive type hand. 2NT is a possibility as natural or a limit Spade fit or better with defensive values. On the otherhand. If your partner opens 1S showing very limited values (9-15), and RHO overcalls 2H. Now, sitting behind RHO, you may very well want to let your partner know about the DEFENSIVE nature of your hand. Here a double as showing cards and at least reasonable expectations that a pass would be ok if fine. The obvious implication is that 2NT can no longer be natural, and shouldn't show spade fit and defense as well either. So here, 2N would be a takeout kind of thing.

I assume this is the type of classification that you are interested in… where position, specific bids, and maybe vulnerability all comes into play. This will be quite an undertaking, but I believe for the most part some general rules could be constructed. If one is willing to give up on negative doubles and find some other bid as a general takeout (of course, missing out when partner of the negative doubler is stacked in their suit).

Ben

[mishovnbg]

There are a number of different approaches to "system design". [At the moment, I am talking about designing any type of complex system, rather than designing a bidding system]. Two of the most important schools of thought are so-called evolutionary design methods and concious design methods.


Concious design is a top down development approach that attempts to use formal logic and modeling techniques to create an optimal solution to the problem being solved.

As the name suggests, evolutionary design is based on trial and error.

Random permutations are made to a complex system. A feedback mechanism rewards "good" changes while punishing bad changes. Over time, the system converges on a locally optimal solution.

Evolutionary design methods are capable of creating remarkably sophisticated systems. The application of evolutionary design is not limited to the biological arena. Evolutionary methods are currently being applied to designing computer programs, integrated cicuits, and even phyical constructs. There are numerous well documented examples in which evolutionary design methods have produced superior solutions than expert humans.

There are three main limitations to evolutionary design methods.

The first is time. Applying these methods requires extremely large numbers of generations.

The second limitation is the requirement to develop an appropriate feedback mechanism.

The final problem is local optimums. In many cases, a system will converge on a solution that is a local optimum, but globally sup-optimal. If mutuations are not large enough to perturb the local optimum, the system will never converge on the global. [There are some really ugly relationships between the "size" of a mutation and the amount of time required to converge on an equilibrium.] In a worst case scenario the search landscape is designed such that a series of local optimums block all potential paths from the starting point to the global optimum.

Now, back to bridge.

You are (essentially) suggesting a evolutionary approach to the design of bidding systems. I contend that this is impractical without significant computer assistance. There are a wide number of problems with the approach that you are suggesting. The most significant is that people aren't able to extract enough information about the specific elements of a bidding system to appropriately condition the design of the bidding system. [People are myopic and they don't play enough hands] This is not to say that many people don't design bidding systems in just the way that you suggest. However, they rarely arrive at anything thats particularly good.

I think that you will have far better luck designing a bidding system using a concious design approach starting from first principles.

You first goal really needs to be establishing a set of basic design goals for the system that you are playing. Decide what you want from a bidding system before worrying how you're going to get tere.