Hand Evaluator(s) and System Design

[hrothgar]

The following posting might be a bit esoteric; none-the-less, it’s a topic that I am starting to find increasingly interesting: Is it practical to use multiple different hand metrics for hand evaluation within a single bidding system? Case in point: Many people claim that it is appropriate to use a “simple” 4-3-2-1 metric for balanced hands that fall into their NT ladder. These same players will use more complex metrics for unbalanced hands. (Examples of more complex metrics might include Kaplan-Rubens, Binky points, Zar points, yada-yada-yada…)

 

I’m not so much interested in questions surrounding the accuracy of any of the different metrics, rather, I am interesting in a more broad ranging question: Does the use of multiple metrics “stress” the bidding system? If so, how much? Obviously, the primary motivation behind creating this thread is a set of comments by Jimmy and Mike in “Bits and Pieces”. Roland posted the following hand

 

♠ 6

♥ K853

♦ AQ6

♣ AKT92

 

And asked about Dealers best rebid following the auction 1C – (1H) – 1S – (P)

 

Lets assume that the partnership agrees to use K+R to judge the strength of unbalanced hands and Thomas Andrews “Fifths” count for balanced hands. (Please note that these metrics are only being used as an example. I don’t want a fight about whether K+R is better or worse than Zar)

 

When Dealer initially picks up his cards, it seems perfectly reasonable to plan to treat the hand as unbalanced. Accordingly, we plug the hand into K+R and low and behold, the hand evaluates at 19.5 “K points”. Now, consider what happens after the 1H overcall. At this point in time, the most likely game is now 3NT Accordingly, we switch to our “balanced” hand evaluator. Andrew’s “fifths” evaluator says the hand is worth 15.8 “5 points”. The two evaluators have radically different judgements about the playing strength of the two hands. (Please note: I’m aware that one “K point” is not necessarily equal to one “5 point”. Even if we can’t necessarily normalize one scale into the other, the basic point still stands.

 

I argue that the need to shift between radically different hand valuation metrics places stress on the bidding system. How can you hope to have meaningful auction if valuation can be this plastic. Case in point: In Roland’s example hand, some players are arguing in favor of a 1NT rebid… Flexibility is all fine and dandy, however, it can be taken too far.

 

In case anyone believes that this example is overly contrived, I’d like to bring up a much more common case. You need to choose an opening bid holding

 

♠ KT3

♥ AT93

♦ 6

♣ AQT74

 

The “textbook” opening playing standard North American methods is 1C. However, consider what happens if the auction starts 1C – (P) – 1S. You have no effective rebid: Both 1NT and 2S substantially misrepresent hand strength… Many people (myself included) would argue in favor of a 1NT opening in order to sidestep this issue. Potentially, this entire class of debate is a special case of the skew between different metrics…

[EricK]

I have been thinking along the same lines. The type of hand that started me off is a balanced 5332 or 4432 at the top of one's NT range with a good 5 card suit or two good 4 card suits. If there is a fit, the hand can be very strong for a suit contract - stronger than partner will expect for a 1NT opening. But if you upgrade the hand (eg with 1x followed by 2NT) and there is no fit, partner will drive to game when it isn't there.

 

Eric

[mike777]

The following posting might be a bit esoteric; none-the-less, it’s a topic that I am starting to find increasingly interesting:  Is it practical to use multiple different hand metrics for hand evaluation within a single bidding system?  Case in point: Many people claim that it is appropriate to use a “simple” 4-3-2-1 metric for balanced hands that fall into their NT ladder.  These same players will use more complex metrics for unbalanced hands.  (Examples of more complex metrics might include Kaplan-Rubens, Binky points, Zar points, yada-yada-yada…)

 

I’m not so much interested in questions surrounding the accuracy of any of the different metrics, rather, I am interesting in a more broad ranging question:  Does the use of multiple metrics “stress” the bidding system?  If so, how much?  Obviously, the primary motivation behind creating this thread is a set of comments by Jimmy and Mike in “Bits and Pieces”.  Roland posted the following hand

 

♠ 6

♥ K853

♦ AQ6

♣ AKT92

 

And asked about Dealers best rebid following the auction 1C – (1H) – 1S – (P)

 

Lets assume that the partnership agrees to use K+R to judge the strength of unbalanced hands and Thomas Andrews “Fifths” count for balanced hands.  (Please note that these metrics are only being used as an example.  I don’t want a fight about whether K+R is better or worse than Zar)

 

When Dealer initially picks up his cards, it seems perfectly reasonable to plan to treat the hand as unbalanced.  Accordingly, we plug the hand into K+R and low and behold, the hand evaluates at 19.5 “K points”.  Now, consider what happens after the 1H overcall. At this point in time, the most likely game is now 3NT  Accordingly, we switch to our “balanced” hand evaluator.  Andrew’s “fifths” evaluator says the hand is worth 15.8 “5 points”.  The two evaluators have radically different judgements about the playing strength of the two hands.  (Please note:  I’m aware that one “K point” is not necessarily equal to one “5 point”.  Even if we can’t necessarily normalize one scale into the other, the basic point still stands.

 

I argue that the need to shift between radically different hand valuation metrics places stress on the bidding system.  How can you hope to have meaningful auction if valuation can be this plastic.  Case in point:  In Roland’s example hand, some players are arguing in favor of a 1NT rebid…  Flexibility is all fine and dandy, however, it can be taken too far.

 

In case anyone believes that this example is overly contrived, I’d like to bring up a much more common case.  You need to choose an opening bid holding

 

♠ KT3

♥ AT93

♦  6

♣ AQT74

 

The “textbook” opening playing standard North American methods is 1C.  However, consider what happens if the auction starts 1C – (P) – 1S.  You have no effective rebid:  Both 1NT and 2S substantially misrepresent hand strength…  Many people (myself included) would argue in favor of a 1NT opening in order to sidestep this issue.  Potentially, this entire class of debate is a special case of the skew between different metrics…

Rebidding 2c regardless of 1h overcall or not seems fine on example hand one. Poor heart spots to reverse. Take away the K of hearts and I still rebid 2c. Playing this rebid with a range of K of h or not seems eminently playable.

Rebidding 2s on example hand 2 seems clear.

 

I see no reason for 2 different bidding metrics per hand.

 

Perhaps not opening hands without tricks or controls in first and second seat would solve many of these issues?

 

IMHO this seems to go back to the debate over constructive or destructive bidding in first or second seat openings. The great mass of bridge players having voted in favor of destructive/pressure opening bids.

[mikeh]

I find all of these discussions of marginal relevance to learning to play bridge. Zar points, Binky points, yada-yada points. Any arithmetical method is of maximum use in the initial evaluation prior to any bidding taking place. But any method that is capable of the degrees of precision and subtlety required to resolve late-stage decisions or competitive decisions will be unplayable for virtually all real-life bridge players.

 

In the interim, it seems to me, those who dream of finding the holy grail of such a method are missing out on exploring the beauty of the game by trying to formalize rules rather than improving judgment.

 

Do any of the world's top players use any such method? Not to my knowledge.

 

The top players use judgement: an appreciation of the comparative values of their holdings having in mind their partner's actions and those of their opponents, within the confines of an initial evaluation method that uses such indicators as high card points, controls, losing trick count and (for all I know) maybe Zar/Binky points as well. But no player, to my knoweldge, uses a single metric other than, perhaps, for notrump openings.

 

What is judgement? I cannot pretend to know how the world's top players think, beyond what can be gleaned from reading their descriptions of their thought processes, but some of the factors would include the afore-mentioned LTC, hcp, controls etc while other factors (some of which would be included in initial evaluation as well) would include in-out valuation (upgrade honours in long suits, downgrade those in short suits), positional values, depending upon opposition action, 'working cards' : honours in partner's suits; cover cards, internal texture, inferences from table action, inferences from opponents' style, an understanding of partner's style, anticipation of how the play might proceed in a possible final contract and so on. I wrote this list as a stream-of-consciousness exercise so it will not be exhaustive.

 

I very much doubt that any method of metrics could be both actually usable and able to incorporate all of these factors.

 

It may be argued that it should be possible to develop a method that would permit lesser players to gain some of the advantages of a top player's judgement through the use of metrics, and maybe it is. But the energy spent pursuing that goal would probably generate greater returns if used developing judgement, and judgement will have a bigger upside.

 

I realize that I may come across like those who objected to the use of hcp: back in the 1940's and early 1950's many experts objected to the use of the Goren-style point count for he same basic reason as I raise here: they were proven wrong and maybe I will be as well ;)

 

In the meantime, I understand (I think) the attraction that the concept has for the mathematically inclined and I wish you well.

[Winstonm]

Not being nearly as bright as the author, I can't add anything to this discussion but wanted to point out that this is what makes bridge such a wondrous game - there is some facet for virtually everyone's interests. The challenge of the game IMO is to try to get two people to think as one.

 

Winston

[david_c]

You need to choose an opening bid holding

 

♠ KT3

♥ AT93

♦  6

♣ AQT74

 

The “textbook” opening playing standard North American methods is 1C.  However, consider what happens if the auction starts 1C – (P) – 1S.  You have no effective rebid:  Both 1NT and 2S substantially misrepresent hand strength…  Many people (myself included) would argue in favor of a 1NT opening in order to sidestep this issue.

Maybe this is a problem hand (though I think it's a clear 2♠ rebid), but I don't agree on what the cause of the problem is. Suppose you change the suits round a bit:

 

♠ KT3

♥ 6

♦ AQT74

♣ AT93

 

Now there's no problem: if you consider the hand too strong for 1♦:1♠,2♠ then you can rebid 2♣ instead. So with the original hand the cause of the problem is a hole in the system: 3154 hands are biddable but 3415 hands aren't. It's not so much a problem of having different methods of hand evaluation, because unless you're a big fan of off-shape 1NT opening bids you're going to be opening 1♣ on any hand of this shape regardless of strength.

 

Anyway, if this sort of thing bothers you, then it seems to me that frequent off-shape 1NT openings actually make the problem worse. That is, you'll be giving partner very accurate information about "NT strength" on hands where it is fairly likely that some other metric would be more appropriate.

[Robert]

Hi everyone

 

I think that HCP values 'according to any measure' are simply a matter of style.

 

I tend to 'know' what to bid in any system based on 'looking' at a hand.

 

For some close cases, I use A=4.3, K=3.1, Q=1.7 and J=.9 as my tie breaker.

 

Most hands you do not have to add any fractions up. Aces and Queen balance

 

each other and Kings and Jacks also balance each other out.

 

If you have a lot of Aces and Kings, the hand is 'upgraded.' With a lot of Queens

 

and Jacks, the hand is downgraded.

 

The auction tends to give even greater meaning to how the hands 'fit' If partner

 

shows 5-5 majors and you hold QJxx and QJxx in the minors, hit the brakes.

 

The K10x A109x x AQ10xx hand I would raise to 2Ss with almost any partner.

 

My partner 'knows' that I raise on suitable three trump hands.

 

I am somewhat old fashioned, in that my NT openings promise at least two cards

 

in each and every suit. I do open NT with 5332 majors and 6322 minors 'if' the

 

hands seems to fit that 1NT description.

 

Bidding is often best an estimate of how many tricks you 'think' declarer will take.

 

One partner of mine often had a shocked look on his face as he 'thanked' me for

 

dummy. He was the best card player that I have ever played with, so my rule with

 

him was to accept any invitation that he made. His card play more than made up

 

for any lack of values in dummy.

 

I often double the declarer rather than the contract when playing bridge. A

 

doubtful double of a bad card player will tend to reward you. Doubling a good

 

card player carries a built in warning. The warning 'double' may allow them to

 

make an unmakeable contract.

 

Happy Holidays everyone.

 

Regards,

Robert

[luke warm]

i don't see any reason for richard to *not* continue his thoughts, hopefully even coming up with something that the avg player can find useful... as mikeh said, the goren point count replaced "judgment" for many people, but the net result was an improvement in bidding (and play) for almost all who adopted it

 

as far as the hand richard posted, i would prefer to use canape openings on all such 'problem' hands... if i can't, then i'm forced to raise to 2♠.. there's nothing inherently wrong with 2♠ there, i just don't like it ;)

[bestguru]

excuse my drunkeness, but I don't think anyone has replied to the question he was asking.

 

I think the point is

 

1C – (1H) – 1S – (P)

 

now suppose opener makes a strong bid. 1) Was partner strong enough in bidding method A to make the strong bid initially? 2) Did the 1H bid encourage partner to change his bidding method to B which made his hand stronger than it would have been in A? 3) Does you response depend in any way on which of quesition one or two is true? 4) Is openers interpretation of your response influenced by any doubts about question 3?

 

I don't think I can answer any of these questions, as I play more pickup than repeat partnerships. If I had to take a guess, I would say that over time with your partner you develop an idea of what partner will typically do and this will become an unwritten part of your system. That is, the system is more than just the conventions played; it is a balancing act between the unique styles and personalities in the partnership. To add to that, this balance may be quite dynamic.

 

I think Mikeh's point is the most relevant. documenting every minute detail of a partnership and then trying to learn and use all of the minutae would be nearly impossible. Beyond the sheer scope of the rules to learn you also would have the conflict between your personality and the personalities involved in the original partnership. I am sure we have all been in situation where the correct system bid just doesn't feel right.

 

I had more to say, but Ii'm being fussed at, and I need a beer, and I think I'm rambling anyways.

[mikestar]

I argue that the need to shift between radically different hand valuation metrics places stress on the bidding system.  How can you hope to have meaningful auction if valuation can be this plastic.…  Flexibility is all fine and dandy, however, it can be taken too far.

A very correct observation -- you are touching on a fundamental aspect bidding theory. All hands fluctuate and sometimes fluctuate radically in value according to the probable contract. There are multiple metrics intrinsic to the situation, even for those of us (not me! :) ) whose sole metric is "expert judgement". The example hand is about two tricks better played in a 4-4 heart fit than in NT, even with good spade stoppers, and this remains true regardless of what methods one uses to crunch the numbers--if your evaluators say otherwise, they are inaccurate.

 

The 3.7 point = about an Ace disrepancy between Fifths and KR on this hand seems fairly accurate: a one trick advantage to playing in an eight card fit is implicit in evaluators that require the same count for both 3NT and 4M. So adding in 3 points for the implict trick, this becomes 6.7 = a bit more than 2 tricks.

 

One thing I liked playing Precision with fairly conservative responses (we passed most non-fitting 7 counts) was that we had some finely tuned sequences that could send messages like "invitational in NT, GF if you fit my major" and "invitational if you fit my major, no game in NT." Rather more useful than "invitational" of "GF."

[helene_t]

We probably all learned to start counting HCPs and then, as soon as a fit was found, either add points for shortness, prime cards etc, or switch to something else (LTC), which amounts to about the same.

 

This does not stress the system much sinse it's easier to upgrade than to downgrade (as one of the old Blue Teams people said, Belladona I think).

 

But when it comes to slam investigation, I think it does stress the system: early in the auction, hands are evaluated in terms of game potential. This means that as soon as we discover that the target may be slam, some bidding space has been wasted on the wrong information exchange (for example, after a 1♦-2♦ inverted raise, it is unclear whether we're looking for additional strength to reach 3NT or 6♦).

 

Some authors take this into account by making a formal distinction between slamish and non-slamish GF hands. E.g., in Lawrence's system

1M-1NT

2m-3NT

could be a hand that was upgraded after hearing the 2m rebid, but could also be a non-slamish GF response. The precision auction

1♣-1♦

1NT-3NT

is similar.

[hrothgar]

What is judgement? I cannot pretend to know how the world's top players think, beyond what can be gleaned from reading their descriptions of their thought processes, but some of the factors would include the afore-mentioned LTC, hcp, controls etc while other factors (some of which would be included in initial evaluation as well) would include in-out valuation (upgrade honours in long suits, downgrade those in short suits), positional values, depending upon opposition action, 'working cards' : honours in partner's suits; cover cards, internal texture, inferences from table action, inferences from opponents' style, an understanding of partner's style, anticipation of how the play might proceed in a possible final contract and so on. I wrote this list as a stream-of-consciousness exercise so it will not be exhaustive.

Here's the crux of the matter: Judgement is a wonderful thing. An expert's intuitive/subjective judge regarding the strength of different hands can be a thing of beauty.

 

However, the ethereal is pretty damn useless as a means of communication and communication is central to the game of bridge.

 

* Disclosure regimes require that players are able to accurately describe their methods.

 

* As computers become more and more central to the game, players require methods to formalize judgement.

[goodwintr]

The point about encoding judgment is valid, certainly, when it comes to computer bidding. In the real world, though, it is important to realize that an evaluation system is just shorthand, and an approximation. I used to wonder why good players would start talking about a hand by saying, "I had 12 points," when everyone realizes that "12 points" leaves a lot unsaid about a hand. The answer, I think, is that "I had 12 points" is just shorthand, and is understood to be that, for a more complicated expression that cannot be reduced easily to numbers -- simply a basis for conversation, and not a statement about the actual value of a hand. I also used to sort of sneer at "4-3-2-1" evaluation, until I came to realize that as a rough approximation, for purposes of easy communication, it isn't bad at all.

[hotShot]

Ever tiny bit of information, will change the hand evaluation. It does not matter who the source is.

 

Synergie effects are:

fit

shortness opposite length without wasted values

controls

values in opps suit placed behind the bidder

 

Downgrades are needed for:

misfits

values in opps suit placed in front of the bidder

shortness opposite length with values (possible double evaluation)

shortness opposite shortness

missing controls

 

I bet there is more and each rule will have exceptions.

[awm]

I agree that re-evaluation is important. This is one of the biggest reasons I've moved away from "immediate game force" approaches to bidding. While styles like two-over-one or precision tend to make auctions "easier" in some ways, there are pretty big prices to be paid. If you frequently open hands which are distributionally strong but lack high card values, you'll find yourself sometimes in really bad game contracts when partner looks at his 13-count and commits to game at first bid, only to find you with a misfitting ten count. On the other hand, if you pass with these distributional hands you can sometimes find yourself missing really good games when you have a fit and the auction is preempted quite high by your next turn. Similar problems come up when partner opens and you decide to make a direct invite, or bid a forcing notrump, only to discover that you have a nice fit in a side suit (radically upgrading your hand) and have no more forcing bids available.

 

My current approach is to delay the decision whether to game force as long as possible, to allow exploration for fit and re-evaluation of hands before determining whether we're looking for game, partscore, or slam.

 

The hand given as an example is a problem because of the point ranges inherent in standard bidding. Even if opponents pass throughout you have a tough rebid after partner's 1♠ call. I generally reverse with these hands (I reverse lighter than a lot of people); obviously there are downsides to that but at least it gets the shape resolved. It wouldn't occur to me to rebid 1NT with a decent 16-count; my evaluation is that this hand is worth a little more than 16 points towards 3nt with no suit fit, maybe a little less than 16 points after the 1♥ call, but still a lot more than the maximum of 14 for 1NT rebid.

[mike777]

et there is more and each rule will have exceptions.

Sorry I see this type of thing all the time and it really bugs me :).

 

A Rule should not have any exceptions, it should just be complete.

 

The good "judgement" players tend to have complete rules, the "poorer" judgement players tend to have incomplete rules or to put it more bluntly, nonsense rules.

 

"I don't know much about history, and I wouldn't give a nickel for all the history in the world. History is more or less bunk."

 

Henry Ford

[hotShot]

Well i don't know if one can give complete rules for evaluation, but i do know that those rules won't be simple.

Given the fact that you have to deal with incomplete information, you might not even know which rule should be applied.

 

But please post a rule that you think is complete and let us see, if there is indeed no exception.

[luke warm]

i could be wrong (i often am), but i think mike means that if one makes a rule with exceptions, it was a bad rule to start with (or that it didn't fit the definition of 'rule')

[mike777]

Well i don't know if one can give complete rules for evaluation, but i do know that those rules won't be simple.

Given the fact that you have to deal with incomplete information, you might not even know which rule should be applied.

 

But please post a rule that you think is complete and let us see, if there is indeed no exception.

I would just argue that rules are not made to be broken, but must be correctly stated. You may feel differently :). If "always cover an honor with an honor" is an example of an incorrectly stated rule here are some better ones.

 

Here is some Kleinman advice:

 

"Cover a card that declarer ( or dummy) leads when you think that by doing so you can promote another card in the suit as a trick for your side. Try to anticipate what you will do before the crucial play arises. Start planning as soon as you see dummy. Stop to think before turning trick 1."

[hotShot]

I get your point about about rules:

Instead of stating:

"Don't pass a red traffic light!"

The rule should be:

"Don't pass a red traffic light, unless you are driving an ambulance, a police car or fire engine on an emergency and if the traffic light is not out of order."

 

People or computer without proper judgement, need "rules of thumb".

 

"rule of thumb

n. pl. rules of thumb

A useful principle having wide application but not intended to be strictly accurate or reliable in every situation."