Tests for a double-dummy solver

[tysen2k]

I repeat this is a function of suits needing only 10 tricks for game, so what? What is the new point?

Is that not enough? Maybe the question would be easier to answer if you asked "Why do we pass?" rather than "Why do we open?"?

 

Tysen

[cherdano]

In the Zar thread, someone gave a link to Richard Pavlicek's database of bridge deals with precomputed double dummy results, and I think Tysen has been using a similar database here.

 

Is there any reasonably big deal database that includes single dummy results, computed by one of the strong bridge programs (GIB, Jack)? I am wondering whether double dummy analysis might, for example, systematically undervalue Queens.

 

Arend

[awm]

There are some general issues about designing a hand evaluation method that perhaps we can try to answer by analyzing play in top competition. Here are some of them:

 

Is bidding double-dummy contracts good?

 

It's quite possible to construct hands where a contract makes double-dummy, but where no one would legitimately make the hand. On the other hand, it's easy to imagine contracts that should be defeated double dummy, but where the defense can easily go wrong. If either of these situations is too common, it would tend to indicate that there's not much value in trying to find the double dummy contract as often as possible (which would seem to be the goal of all the various hand evaluation analysis). So here are two tests we could try:

 

When a contract is bid in top-flight competition, how often does the result of the hand (making versus not making, ignoring extra over/under tricks) match the double dummy prediction?

 

If a pair were to somehow bid only the contracts that make double dummy, and always take the double dummy number of tricks, how many IMPs would they win (or lose!) when compared to real players?

 

Should we aim for making, or for par?

 

First let me define par. The par spot is the lowest contract such that neither side can improve their results by bidding more (assuming double dummy play/defense). So if no one is vulnerable, we can make 4♥, and the opponents can make 2♠, then the par spot is 4♠X by opponents, 300 for us. The various bidding analysis tend to assume that the goal is to find the making spot, not the par. The exception to this is rules like the Law of Total Tricks (and Lawrence's substitute in "fought the law"). These rules seem to perform poorly when compared to things like zar/bumrap/binky points, but they're really aiming for a different target. So here are the questions:

 

How often does the par result involve my side going down in a contract?

 

If a pair were to let their opponents bid and make their best contract rather than competing to the par spot (when par is negative for their side), how many imps per board would they lose?

 

If these numbers are large, it suggests that aiming for the making spot might be the wrong idea. Note that I expect these numbers to depend a lot on vulnerability -- when unfavorable it will rarely be right to bid something that will not make.

 

Is there an advantage to overcompeting par?

 

In principle there could be advantages to overcompeting par. For example, suppose I open 3♠ on a hand where the opponents cannot make a four-level contract. It may be the case that 3♠ doubled will go two down for -300. So it's a bad bid... or is it? It could well be that the opponents have great difficulty doubling the contract, and will often be pushed into a bad 4♥ game. If I had passed they would have played 3♥ making three. So how can we measure this?

 

Examine hands from top flight competition where one side bid beyond par. Compare their actual result on the board to the par result (computed using a double dummy solver). How many IMPs did they gain/lose on these hands?

[mike777]

Yes,

1) should hand evaluation measures get us to the best possible contract or the best contract possible?

2) should hand evaluation measures be an important tool in resolving conflicting goals such as getting us to the best contract or not giving the opp a free ride to theirs?

 

Please note these goals are different from getting us to the PAR spot or to the best contract.

 

I just played a TM. We lost on this hand. Ok we could have won it on another hand that involved hand evaulation but that one did not involve me. This hand I had the crucial decision to make. I decided that we could not make 3nt on any normal lead or defense so did not bid it. Our opp bid it and made it on a lead that I thought as obviously not best on the bidding. How do we factor that into hand evaluation?

 

If you set junky goals you get junky results. On the other hand a new bidding tool or theory need not be perfect, just better, whatever that means. :rolleyes:

[Echognome]

As a follow-up to this, I would find it quite interesting if there was another column. Namely if for each shape there was %chance WE make game and %chance OPPS make game. This might help when discussing whether or not it's important to get into the auction.

[PMetsch]

There are some general issues about designing a hand evaluation method that perhaps we can try to answer by analyzing play in top competition. Here are some of them:

 

Here are some conclusions from a Peter Cheung. He has done a lot of statistical work (also with OKbridge data). All the numbers can be found at:

 

http://crystalwebsite.tripod.com/index.htm

 

I am no expert, so I can not tell if all those numbers make sense.

 

Is bidding double-dummy contracts good?

 

copied from double dummy accurate section

 

The most important general finding is that double dummy analysis is very accurate as compared to actual play from OKBridge. The overall total number of tricks taken by the declarer is 9.21 (9.22 for imp and 9.20 for mp). The double dummy analysis of the same deal produce 9.11 (9.12 for imp and 9.11 for mp). So actual play by OKBridge player takes 0.1 tricks more then the double dummy analysis result. This is from 383,000 deals and over 25 million plays.

 

Should we aim for making, or for par?

 

copied from final contract section

 

Out of 25 non-slam contracts only 8 contracts are shown to have positive results.

 

They are 1H 1S 1NT 2H 2S 3NT 4H 4S. This is the case for both imp and mp scoring. I call them "our contracts" or "Peter's 8 contracts". If it is our hand we prefer to play in "our contract". If it is their hand, we prefer them to not play in "our contracts"

 

You may also be interested in data on the folowing page: http://crystalwebsite.tripod.com/hand_patterns.htm

[cherdano]

Is bidding double-dummy contracts good?

 

copied from double dummy accurate section

 

The most important general finding is that double dummy analysis is very accurate as compared to actual play from OKBridge. The overall total number of tricks taken by the declarer is 9.21 (9.22 for imp and 9.20 for mp). The double dummy analysis of the same deal produce 9.11 (9.12 for imp and 9.11 for mp). So actual play by OKBridge player takes 0.1 tricks more then the double dummy analysis result. This is from 383,000 deals and over 25 million plays.

This information is not very important. What would be important is the typical error (e.g. by stating how often DD analysis is off by 1 or 2 or 3 tricks etc.)

By the same argument, if I throw coins and decide for 4 out of 5 deals that NS can make 9 tricks, and 10 on the other, this would be a perfect analysis for MP contracts, since the average predicted number of tricks is 9.20.

 

Arend

[PMetsch]

This information is not very important. What would be important is the typical error (e.g. by stating how often DD analysis is off by 1 or 2 or 3 tricks etc.)

Just looked at all those numbers at the website, but I could not find info about errors nor other data to calculate errors.

 

Another interesting point about the comparison of DD solver vs.OKBridge: The OKBridge declarer takes more tricks than the DD solver declarer. Now assume the DD solver is correct, then how can OKBridge declarer beat the best possible result? I think he must be playing worse than the DD solver, but the OKBridge defenders play much worse to compensate.

[Finch]

The most important general finding is that double dummy analysis is very accurate as compared to actual play from OKBridge. The overall total number of tricks taken by the declarer is 9.21 (9.22 for imp and 9.20 for mp). The double dummy analysis of the same deal produce 9.11 (9.12 for imp and 9.11 for mp). So actual play by OKBridge player takes 0.1 tricks more then the double dummy analysis result. This is from 383,000 deals and over 25 million plays.

Surely there is a big variation by contract, and particularly by level?

 

I believe that I tend on average to beat DD par when declaring at the 1- and 2-level, and I am substantially worse than DD par in slams.

 

Also, I would expect 3NT contracts to be close to DD par when bid after the defence has opened the bidding (becuase the opening lead is likely to be right, and because everyone knows all the high cards).

[PMetsch]

Surely there is a big variation by contract, and particularly by level?

copied from:

 

http://crystalwebsite.tripod.com/double_dummy_accurate.htm

 

The following table is a break down of the numbers by the level for NT contact.                       
actual play  double dummy    difference

level 1       7.34      6.93      0.41

level 2       7.86      7.56      0.30

level 3       9.29      9.10      0.19

level 4       10.25     10.10      0.15

level 5       10.34     10.25      0.09

level 6       11.40     11.48     -0.07

level 7       12.27     12.40     -0.12


The following table is a break down of the numbers by the level for spade contact.
actual play  double dummy    difference

level 1        7.83      7.64      0.19

level 1        8.18      8.06      0.13

level 1        8.61      8.52      0.09

level 1        9.86      9.87     -0.00

level 1        10.30     10.23      0.07

level 1        11.45     11.57     -0.11

level 1        12.26     12.39     -0.12

 

Obvious the last table should read level 1 untill level 7

 

I believe that I tend on average to beat DD par when declaring at the 1- and 2-level, and I am substantially worse than DD par in slams.

 

If the data above is true then you are not the only one :). At higher levels the defense make less errors/wrong leads.

[Fluffy]

Nice data, your data base is fully randomly generated?, and how much time did it take to analyse all those deals double dummy?

[tysen2k]

Nice data, your data base is fully randomly generated?, and how much time did it take to analyse all those deals double dummy?

The database is 1 million random deals that Matt Ginsberg compiled using GIB. The database is available on his website. I don't know how long it took him to generate them all.

[tysen2k]

As a follow-up to this, I would find it quite interesting if there was another column.  Namely if for each shape there was %chance WE make game and %chance OPPS make game.  This might help when discussing whether or not it's important to get into the auction.

I was thinking about this yesterday too. On my list.

[han]

Exactly Matt, I was thinking the same thing. For example, with 2344 shape I expect that the opponents are more likely to have game than with 4432 shape. This could be a reason for opening with the 2344 shape as well.

 

Also, I expect that the more unbalanced we are, the more likely the opponents have game (since they are more likely to have distributional hands too). This could be yet another argument for opening light with highly distributional hands.

 

BTW thanks for adding the freaky hands too. Now that we have the errors available too, we can draw our own conclusions from these.

[han]

By the same argument, if I throw coins and decide for 4 out of 5 deals that NS can make 9 tricks, and 10 on the other, this would be a perfect analysis for MP contracts, since the average predicted number of tricks is 9.20.

 

Arend

Arend,

 

This is very disturbing to me. Last time I was in Germany they were using Euros, and these coins have only two sides. It seems to me that it is impossible to guarantee a 4/5 chance by throwing coins a finite number of times.

 

Apologies for an off-subject response.

[EricK]

By the same argument, if I throw coins and decide for 4 out of 5 deals that NS can make 9 tricks, and 10 on the other, this would be a perfect analysis for MP contracts, since the average predicted number of tricks is 9.20.

 

Arend

Arend,

 

This is very disturbing to me. Last time I was in Germany they were using Euros, and these coins have only two sides. It seems to me that it is impossible to guarantee a 4/5 chance by throwing coins a finite number of times.

 

Apologies for an off-subject response.

You generate any probability you like with a coin as follows:

 

Express the probability as a binary number (eg 4/5 = 0.1100110011001100....) and throw the coin until you get a head. Then count it as a success if the first head appears on a throw number indicated by a 1 in the expansion (eg for 4/5 if the first head appears on throw number 1,2,5,6,9,10 etc), and a failure otherwise.

 

This doesn't guarantee a finite number of throws, but you would be very unlucky to have to throw the coin an infinite number of times :rolleyes:

 

Eric

[awm]

Why should we be interested in the probabilities of making game?

 

I think a lot of us count points for shape when we open. For example, other than 10-12 notrumpers, I don't think many of us would open this in first seat:

 

xxx

xxx

Axx

KQJx

 

However, I know a lot of people who would open:

 

xxx

x

Axx

KQJxxx

 

Now, assuming fairly standard methods, it would be logical to open 1♣ on both hands (if we were going to open). This steals basically no space from opponents. It doesn't particularly direct a lead (in this case the clubs are good, but we have to open 1♣ on three small at times so I don't think partner can bet on good clubs). It doesn't really help us in competitive auctions much, because partner won't know we have six clubs on the second hand and can't really raise the suit very aggressively.

 

So why is it that people open the second hand and not the first? The reasoning is that the second hand is somehow "better." Suppose that partner has some random thirteen count and we end up in a game. The top hand is likely to be a disappointment, and our chances of making 3NT or 4M are probably not good when partner has a "minimum game force." The bottom hand has a nice source of tricks for any contract, and a possible ruffing value if we find a spade fit. It seems likely that we would have play for a game on most 13-counts partner could produce.

 

As for majors versus minors, I would happily open this hand playing fairly standard methods:

 

AQxxx

KJxxx

x

xx

 

I have six losers, 27 ZAR, rule of 20, blah blah blah. Most players would open this hand. It's likely that we have a fit in one major or the other, and we will often have a good chance at game when partner has a decent hand. On the other hand, switch this to:

 

x

xx

AQxxx

KJxxx

 

I'm not nearly so eager to open this hand, and would probably pass. If I open, chances are good that we will end up in 3NT when partner has a decent hand (despite a likely 5-3 or 5-4 minor fit). I don't necessarily like our chances of making 3NT. My weakness in the majors suggests partner will need many cards there.

 

I should also note that many modern systems require that partner make immediate decisions about whether to force game opposite an opening. This includes the 2/1 game forcing method that is so popular, as well as most strong club or diamond response structures. Most relay systems also have this property, even some of the ones where the relay is not game forcing (since the main information discovered before deciding to game force is frequently whether opener's points are max or min, not opener's detailed shape). Obviously being forced to decide whether to set up a game force early in the auction has its weaknesses. But since we seem to have to do it, it will be good to understand how to evaluate distribution in the absence of detailed knowledge about fit.

 

So the relevent question would seem to be:

 

How many points do I need, with various distributions, such that the probability of game becomes roughly equal?

[Blofeld]

This doesn't guarantee a finite number of throws, but you would be very unlucky to have to throw the coin an infinite number of times :rolleyes:

To be overly pedantic, it does guarantee (i.e. probability 1) a finite number of tosses, but the # of possible numbers of tosses is infinite.

[Guest Jlall]

Why should we be interested in the probabilities of making game?

 

I think a lot of us count points for shape when we open. For example, other than 10-12 notrumpers, I don't think many of us would open this in first seat:

 

xxx

xxx

Axx

KQJx

 

However, I know a lot of people who would open:

 

xxx

x

Axx

KQJxxx

 

Now, assuming fairly standard methods, it would be logical to open 1♣ on both hands (if we were going to open). This steals basically no space from opponents. It doesn't particularly direct a lead (in this case the clubs are good, but we have to open 1♣ on three small at times so I don't think partner can bet on good clubs). It doesn't really help us in competitive auctions much, because partner won't know we have six clubs on the second hand and can't really raise the suit very aggressively.

 

So why is it that people open the second hand and not the first? The reasoning is that the second hand is somehow "better." Suppose that partner has some random thirteen count and we end up in a game. The top hand is likely to be a disappointment, and our chances of making 3NT or 4M are probably not good when partner has a "minimum game force." The bottom hand has a nice source of tricks for any contract, and a possible ruffing value if we find a spade fit. It seems likely that we would have play for a game on most 13-counts partner could produce.

 

As for majors versus minors, I would happily open this hand playing fairly standard methods:

 

AQxxx

KJxxx

x

xx

 

I have six losers, 27 ZAR, rule of 20, blah blah blah. Most players would open this hand. It's likely that we have a fit in one major or the other, and we will often have a good chance at game when partner has a decent hand. On the other hand, switch this to:

 

x

xx

AQxxx

KJxxx

 

I'm not nearly so eager to open this hand, and would probably pass. If I open, chances are good that we will end up in 3NT when partner has a decent hand (despite a likely 5-3 or 5-4 minor fit). I don't necessarily like our chances of making 3NT. My weakness in the majors suggests partner will need many cards there.

 

I should also note that many modern systems require that partner make immediate decisions about whether to force game opposite an opening. This includes the 2/1 game forcing method that is so popular, as well as most strong club or diamond response structures. Most relay systems also have this property, even some of the ones where the relay is not game forcing (since the main information discovered before deciding to game force is frequently whether opener's points are max or min, not opener's detailed shape). Obviously being forced to decide whether to set up a game force early in the auction has its weaknesses. But since we seem to have to do it, it will be good to understand how to evaluate distribution in the absence of detailed knowledge about fit.

 

So the relevent question would seem to be:

 

How many points do I need, with various distributions, such that the probability of game becomes roughly equal?

I'm not sure how we got from "5521 is more likely to make game than 2155" to "AQxxx KJxxx xx x is an opener and xx x AQxxx KJxxx is a pass." It seems a few steps in the logic chain are missing.

 

1) There are other variables. Lead direction, saves, preemption. One could argue since you are 2155 THEY are more likely to make a game thus you should open lighter with the minors in hopes that you can find a fit and save. Obviously something is missing from that logic too, but to think that likelihood of game is the only variable in opening seems wrong.

 

2) Passing does not preclude getting to game. We are more likely to make game with the majors, true, but only if we find a fit. If we pass first and then later find a fit, we can upgrade accordingly. It is not like it's now or never and we have to guess. Maybe we should wait to find a fit before doing our upgrading?

 

3) Bridge is a partnership game. Partner also knows major suit games are easier than minor suit games (or light HCP 3Ns) and can adjust his aggression according to where we are likely to play. If he has a minor suit fit, for instance, he will be less agressive than if he has a major suit fit.

 

4) Even if it were only about likelihood of game, the chance may be great enough to open the second hand. Alternatively, it may be so low even with 5-5 in the majors that we should pass the first one. All of this is moot, of course, since there are other things to consider.

 

Let's not jump to conclusions because of a priori odds that involve only 1 variable and 1 person in the partnership. Perhaps there is more to bridge than that.

[cherdano]

By the same argument, if I throw coins and decide for 4 out of 5 deals that NS can make 9 tricks, and 10 on the other, this would be a perfect analysis for MP contracts, since the average predicted number of tricks is 9.20.

 

Arend

Arend,

 

This is very disturbing to me. Last time I was in Germany they were using Euros, and these coins have only two sides. It seems to me that it is impossible to guarantee a 4/5 chance by throwing coins a finite number of times.

You have a lot of trust in the fairness of our coins in Germany :rolleyes:

[awm]

I'm not sure how we got from "5521 is more likely to make game than 2155" to "AQxxx KJxxx xx x is an opener and xx x AQxxx KJxxx is a pass." It seems a few steps in the logic chain are missing.

 

1) There are other variables. Lead direction, saves, preemption. One could argue since you are 2155 THEY are more likely to make a game thus you should open lighter with the minors in hopes that you can find a fit and save. Obviously something is missing from that logic too, but to think that likelihood of game is the only variable in opening seems wrong.

 

2) Passing does not preclude getting to game. We are more likely to make game with the majors, true, but only if we find a fit. If we pass first and then later find a fit, we can upgrade accordingly. It is not like it's now or never and we have to guess. Maybe we should wait to find a fit before doing our upgrading?

 

3) Bridge is a partnership game. Partner also knows major suit games are easier than minor suit games (or light HCP 3Ns) and can adjust his aggression according to where we are likely to play. If he has a minor suit fit, for instance, he will be less agressive than if he has a major suit fit.

 

4) Even if it were only about likelihood of game, the chance may be great enough to open the second hand. Alternatively, it may be so low even with 5-5 in the majors that we should pass the first one. All of this is moot, of course, since there are other things to consider.

 

Let's not jump to conclusions because of a priori odds that involve only 1 variable and 1 person in the partnership. Perhaps there is more to bridge than that.

(1) Yes, there are other variables. If I could open the 1-2-5-5 hand with 2NT showing 8-11 points and 5-5 in the minors I would happily do so. However, we're considering one-level "constructive" openings in a fairly standard system. I'm not convinced that a 1♦ call has a great lead directional value, or that it's even much help to partner in finding a fit since it doesn't necessarily show five cards.

 

(2) I agree that passing doesn't preclude getting to game. But if there are fairly "normal" hands which make game good but which partner wouldn't bother to open (i.e. Kxxx Qx Axx xxxx) that's a pretty good argument for bidding isn't it?

 

(3) Sure, partner knows these things. But partner often has to decide whether to force game pretty early in the bidding (i.e. her first bid over 1M in a 2/1 game forcing system, and her second bid in many other sequences). Partner will generally assume that my hand is "normal" for the bidding so far (minimum but not sub-minimum values, not much more distributional than I have shown). These five-five hands with ten points are likely to be a surprise to partner, who will probably be assuming something along the lines of a (5422) twelve-count when making her decisions. So the question is: will partner usually make the right decision anyway? Or is my hand so far from what she expects in terms of playing strength, that she will often force us to games that go down?

 

(4) Certainly, we don't have all the data about probabilities of game yet. It could be that 1-2-5-5 hands play incredibly well in 3NT because we can run the five card suits. I suspect that hand evaluation for notrumps and suit contracts are different (i.e. 5431 is better than 5422 for play in a suit, but probably the same or worse for play in notrump). But these are just my suspicions, and that's why I'm asking the question.

[Fluffy]

Also, I expect that the more unbalanced we are, the more likely the opponents have game (since they are more likely to have distributional hands too). This could be yet another argument for opening light with highly distributional hands.

I somehow think the opposite, the chances that opponents have game when you get 4333 is quite high, because everything breaks evenly. Maybe this is a weak argument.

[Guest Jlall]

(1) Yes, there are other variables. If I could open the 1-2-5-5 hand with 2NT showing 8-11 points and 5-5 in the minors I would happily do so. However, we're considering one-level "constructive" openings in a fairly standard system. I'm not convinced that a 1♦ call has a great lead directional value, or that it's even much help to partner in finding a fit since it doesn't necessarily show five cards.

 

(2) I agree that passing doesn't preclude getting to game. But if there are fairly "normal" hands which make game good but which partner wouldn't bother to open (i.e. Kxxx Qx Axx xxxx) that's a pretty good argument for bidding isn't it?

 

(3) Sure, partner knows these things. But partner often has to decide whether to force game pretty early in the bidding (i.e. her first bid over 1M in a 2/1 game forcing system, and her second bid in many other sequences). Partner will generally assume that my hand is "normal" for the bidding so far (minimum but not sub-minimum values, not much more distributional than I have shown). These five-five hands with ten points are likely to be a surprise to partner, who will probably be assuming something along the lines of a (5422) twelve-count when making her decisions. So the question is: will partner usually make the right decision anyway? Or is my hand so far from what she expects in terms of playing strength, that she will often force us to games that go down?

 

(4) Certainly, we don't have all the data about probabilities of game yet. It could be that 1-2-5-5 hands play incredibly well in 3NT because we can run the five card suits. I suspect that hand evaluation for notrumps and suit contracts are different (i.e. 5431 is better than 5422 for play in a suit, but probably the same or worse for play in notrump). But these are just my suspicions, and that's why I'm asking the question.

1) You are right 1D does not guarantee 5, but (and this is related to point 3) partner can still raise with 4. If it goes 1D-1H-2D you will likely be saving over something, so it's not that hard to find a fit and save. 1D may not guarantee a good suit, but it's certainly your most likely BEST suit when you open it. Partner will keep this in mind when on lead. There are also other variables, obviously.

 

2) I do not deny it's possible for the hand to be passed out, but it is unlikely with 3 cards in 2 suits. You mentioned "random 13" counts earlier. If parnter has one of these and a fit, you will get to game if you pass (even a 3 card fit you will likely bid game). If his "random 13" includes no fit, you will avoid a bad game where you would get there had you opened. Similarly, if we pass and later michaels with the hand you gave, partner will know to bid game (assuming we get a chance to bid again).

 

3) Over a MAJOR partner will often have to decide immediately whether to game force or not. That seems like an argument AGAINST opening light with both majors. If you open 1D partner rarely has to decide immediately, and if you open 1C it would be even rarer. This could be one of the "variables" mentioned in point 1.

 

4) I completely agree 100 % with you about there being a need for different evaluation schemes between suits and NT. In fact, in another ZAR post from many months ago, I argued this point vigorously and I see it as a weakness of the ZAR evaluation scheme. HCP are much more accurate for balanced hands.

[thomaso]

Some notes on my Binky Points research.

 

(1) I have specifically never made claims about how the research should guide bidding. There are some obvious points and some less-than-obvious points about how to use such 'fine' evaluations. When do you invite? When do you accept invites? These are all non-trivial problems, made even more non-trivial by the fact that the narrower you put the ranges for inviting, accepting, and rejecting, the more opponents know about your hand in these circumstances. I don't want to claim that my results have only purely theoretical value, but they were never meant to be a guide for bidding.

 

(2) In particular, obviously, you want to open different strength hands differently depending on the shape of the hand (rather than the pattern.) Clearly, a light 5-5-2-1 is a better candidate for opening than a light 2-1-5-5. That's because the number of hands partner can have where game makes is higher, because he needs less for game in a major than in a minor, and if we're gonna play a suit contract, that suit contract is probably gonna be in one of my five-card suits. It's not a flaw in BP (or any other evaluation system) if it gives the same value for equivalent 5521 and 2155 hands, it's a flaw in bidding if you use one value for determining what to open all hands. When to open light is a matter of whether you have any defense, whether you have a safe rebid, whether passing might miss an easy game.

 

Many moons ago, someone made much of the fact that my data showed that AKxxx xxx Ax xxx actually had lower playing strength than xxxxx Axx Ax Kxx - that is, that honors in your long suit were actually less valuable. It is a surprising result, considering all those expert recommendations over the years.

 

Well, the difference between these two hands in Binky Points is actually relatively miniscule. On the other hand, consider what opening 1S does to partner's hand. When you open the second hand 1S and partner has Qxx in spades, he is going to over-value that queen. That means that partner is going to push to game often on precisely the wrong hands, because, from his point of view, you are likely to have honors in the suit. Even though the second hand technically has nearly exactly the same playing strength, it doesn't mean it is just as good an opener.

[tysen2k]

Welcome to these forums, Thomas.