[pran]

No screens.

We could not exclude the strong possibility that the BIT was caused by a player considering passing or bidding 3NT rather than raising to 4♥; both alternatives demonstrably suggesting a pass over the eventual 4♥ bid.

Remember that the 6♥ bid must be ruled upon from the auction, the BIT and the cards held by East (who did bid 6♥). The cards held by West (who committed a BIT) are completely irrelevant!

[mikeh]

cherdanno, on May 12 2010, 11:55 AM, said:

Quote

Finally, don't you think that the combination of West having extra values on the actual hand, and East making a strange 6H bid, should be allowed as evidence towards the possibility of such tells/tendencies?

This is a common view and, imo, fatally flawed.

Committees only see hands on which the non-offending side got a bad result. In situations in which the non-BIT hand moved aggressively after the BIT, there will be a range of outcomes:

1. The aggression led to a horrible result: the non-offending side is happy with the ATB result, and the committee never sees it

2. The result appears normal: the non-offending side is satisfied, after seeing the aggressive bidder's hand, that the bid was not demonstrably influenced: the committee never sees it

3. The result is great for the offending side: the non-offending side gets a bad ruling and appeals

4. The result is great for the offending side, the ruling adjusts it and the offending side recognizes that the adjustment was appropriate (I and, I am sure, others who post here have called the director before taking action in situations where I was morally certain that I had my call and also morally certain that the result would be rolled back...I make the call and accept the ruling). The committees never sees it

5. The result is great for the offending side, the score is adjusted and the offending side appeals.

The committee only sees (3) and (5), so they only see hands on which the BIT hand delivers the values (or lack of values) that results in the other hand's action working out.

So almost every hand they see will have this apparent coincidence. But if they saw every hand on which a BIT occurred and further action flowed, they'd see that some hands the BIT was caused by weakness, not strength, or strain not level issues...and the aggressive move was a disaster.

It is unfair, therefore, to invoke a 'law of coincidence' from which to infer that a good result MEANS that the BIT was always correctly interpreted...that it always showed the information that happened to exist on those occasions that result in committee hearings.


Let me stress a point I have tried to make: I am NOT arguing that there should have been no adjustment. All I have ever said is that the OP didn't give us enough information on which to justify an adjustment. Had I been on the Committee, I would have asked a number of questions, and would have had some degree of scepticism when listening to the answers. But if EW were beginners or had a track record of bizarrely aggressive bidding, or could point to some other plausible explanation, I would find it difficult to adjust, because I still don't see how the BIT demonstrably suggested bidding on, let alone jumping to slam.

The BIT would suggest to me, in a vacuum, that partner may have thought about passing or 3N or raising or cue-bidding. Only 1 of those 4 possibilities implies that bidding beyond game would be good, while 3 of them suggest that passing would be good. So even if the imp scale warranted action (which I don't think it does, since we may get doubled), most of the BIT issues make bidding silly. Finally, the fact that E jumped to slam rather than cuebid....surely if you were taking advantage, but weren't sure it promised extras, you'd cue bid in order to allow partner to sign off?...suggests, slightly, that he is simply an idiot. And we don't punish idiots for being idiots...at least, not via a committee ruling and an adjusted score.

I suspect that more times than not, the evidence I heard would not satisfy me, and an adjustment would flow....I say that based on being on many committees over the years.

[Cascade]

awm, on May 13 2010, 07:45 AM, said:

I think this should be rolled back to 4♥.

However, I disagree with some of the arguments presented in favor of this position. Partner's actual hand should not have any effect on the ruling -- either the BIT suggests bidding 6♥ or it doesn't; whether partner actually holds extras or not is irrelevant. Also, I'm not sure the IMP table is relevant here. The IMP scores are already built in to each bidder's actions -- the question is not whether bidding 6♥ has a positive IMP expectation or not (doesn't matter) but whether the IMP expectation for bidding 6♥ given the BIT is greater or smaller than it would otherwise be. This is mostly a function of the success probability. Note that opener's hand is good enough that 4♥ will probably make opposite a lot of "borderline raise/pass" hands partner could hold too -- there's no reason that the only likely possibilities are "4♥ is down or 6♥ is making."

The main issue is criterion 1, does the UI demonstrably suggest bidding 6♥? I don't think this means that 6♥ has to be the "best action" given the UI -- it just means that the odds of 6♥ being right have to be substantially higher (given the UI) than they would be (in the same auction without the UI).

Since arguably no good player would ever bid 6♥ here, it's not surprising that (without the UI) the odds of 6♥ being right are quite small. It's easy to construct many totally normal 4♥ bids where 6♥ is horrible. Also, the fact that 3♥ (invitational) was bid at the previous turn suggests that even this player does not think 6♥ is odds on -- why is this a "three or six" hand where we can never play 4♥?

Of course, we cannot punish someone for just being a bad bidder. Just because 6♥ is very rarely successful (without the UI) does not mean we should penalize someone who bids it. However, the issue is whether 6♥ is much more likely to be successful (even if still less likely to be right than passing 4♥) given the UI. I think this is the case -- the odds that partner has a "borderline slam try" for his BIT are not 100%, but even if they are only 25% this is surely higher than the a priori odds of 6♥ being right opposite a "normal 4♥ bid." Thus we have Pr[6♥ is right given the BIT] >> Pr[6♥ is right given only the hand and auction]. This should be sufficient to roll back the result.

This argument seems to me to ignore the substantial possibility that partner has a 3NT (or some other action - not PASS) or 4♥ decision. While a cue or 4♥ decision might increase the chance of 6♥ being right both a PASS or 4♥ and a 3NT or 4♥ decision diminish the chance of 6♥ being right.

Its far from clear how the numbers would stack up.

[awm]

Cascade, on May 12 2010, 03:24 PM, said:

This argument seems to me to ignore the substantial possibility that partner has a 3NT (or some other action - not PASS) or 4♥ decision.  While a cue or 4♥ decision might increase the chance of 6♥ being right both a PASS or 4♥ and  a 3NT or 4♥ decision diminish the chance of 6♥ being right. 

Its far from clear how the numbers would stack up.

The thing is, I'd rate the chances something like this:

(1) Borderline 3♥/P; Low probability without BIT; Moderate with BIT; 6♥ never right
(2) Borderline 3♥/3N; Low probability without BIT; Moderate with BIT; 6♥ never right
(3) Totally normal 4♥; High probability without BIT; Never with BIT; 6♥ almost never right
(4) Borderline slam try; Low probability without BIT; Moderate with BIT; 6♥ almost always right

Regardless of the exact values of the probabilities for this table, it seems like the chance that 6♥ is right given the BIT is substantially more than the a priori chance. The reason is that the chances of 6♥ being right if partner has anything but the borderline slam try are extremely marginal.

[cherdanno]

As is often the case I think I agree mostly with MikeH (it seems to me that if we were discussing an actual pair the probability we would agree on whether to adjust or not would be pretty high), but still don't understand his main point.
Mike, when West has the good version of the tank and East makes a bid that takes advantage of the good version of this tank, that obviously increases the likelihood that EW have some "black magic" that tells East what kind of hand West has. Why shouldn't this be allowed as evidence towards the assumption that they do have such a magic understanding?
Obviously, in many cases (e.g. if EW are a very weak pair) it would be extremely weak evidence. But disregarding it completely (as you seem to be arguing for) seems non-sensical to me.

[mikeh]

cherdanno, on May 12 2010, 05:02 PM, said:

As is often the case I think I agree mostly with MikeH (it seems to me that if we were discussing an actual pair the probability we would agree on whether to adjust or not would be pretty high), but still don't understand his main point.
Mike, when West has the good version of the tank and East makes a bid that takes advantage of the good version of this tank, that obviously increases the likelihood that EW have some "black magic" that tells East what kind of hand West has. Why shouldn't this be allowed as evidence towards the assumption that they do have such a magic understanding?
Obviously, in many cases (e.g. if EW are a very weak pair) it would be extremely weak evidence. But disregarding it completely (as you seem to be arguing for) seems non-sensical to me.

Here is, I hope, a simple explanation of the effect of selection bias.

Imagine a set of one thosuand cats, made up of every colour possible, and with several hundred sick cats. We stipulate that the cats all began as healthy...that is, they had their colour before their health became an issue (so as to rule out any possibility that health affects colour, but leaves open that colour affects health).

We decide that we will take to the vet ONLY those cats that are both black and sick.

The only cats the vet sees are black and sick, yet he knows that there are other cats out there. We then invite him to infer that because the cats he sees are black, and sick, there must be a causal connection. It can't be a coincidence that the only sick cats he sees are black.....they must be sick BECAUSE they are black

We show committees only those hands on which we think a BIT may have led to a successful action. So the committee doesn't see the BITs that led to a bad outcome, or to a neutral outcome (no action taken) and so on...it only sees sick black cats and knows that there are other cats out there but doesn't know if any of them are black or sick.

So it is tempting for the vet to infer a causal connection between hair colour and health, and for committees to infer a causal connection between extra value BITs and successful aggressive action. Neither is, on those reasons alone, valid or fair.

[kenrexford]

I still can't believe that this is not friggin' obvious. But, I cannot understand how the insanely detailed arguments that I and Adam are posing are not clearly right.

If Responder bids 4♥ opposite a non-forcing 3♥ call, he has a hand where 4♥ is the obvious call, or where 3NT was an option, or where 3♠ was an option, or where passing was an option, or where a cuebid was an option, or where something above 4♥ was an option.

If Responder breaks tempo, he does not have a hand where 4♥ was an obvious call. Therefore, he has a hand where 3NT was an option, or where 3♠ was an option, or where passing was an option, or where a cuebid was an option, or where something above 4♥ was an option.

We all agree so far.

However, if every option is equally likely, which is ridiculous, nonetheless 6♥ rates to be a reasonable call in one out of every six instances when there is no hesitation but in two out of every five instances when there is a hesitation.

If the partner hesitates frequently, then the odds of a 6♥ call working without a hesitation is 0% whereas the odds are 40% with the hesitation.

If the partner never hesitates and then bids game with the weak options, then the odds of 6♥ being right increase to 50%.

If 3♠ would have been a cuebid, then the odds of 6♥ being right increases to 66%.

Hence, because Opener's call is apparent lunacy and inconsistent with his 3♥ call, and as the odds of 6♥ go up, sometimes substantially, after the hesitation, no rational person could draw any conclusion other than that the 6♥ call was unethical and against the rules, not lunacy.

[kenrexford]

Mike, I understand your cat story as responsive to some of the arguments. I think you are missing the fact that they view the actual hand as evidence of an implicit understanding, but then maybe you are conceding this and offering a rebuttal as to why the evidence is misleading.

However, I don't think you are understanding that it is irrelevant whether Responder has a weak hand, a strong hand, or a hand that features a void in hearts and eight solid spades. The fact remains that the hesitation makes 6♥ more statistically likely to succeed, because the hesitation weeds out a huge percentage of hands where 6♥ fails.

If Opener understands his partner's hesitation, then 6♥ might be a 60% shot. If Opener has no ability to read the reason for the hesitation, 6♥ is still an X% shot, with a no-hesitation 6♥ being (X-Y)%. Because X > (X-Y), 6♥ is suggested as more likely to succeed.

[blackshoe]

Why are you assuming that opener deliberately used UI?

[cherdanno]

Mikeh, one of the oddities that makes some of your posts so annoying is that you don't seem to be able to have any grasp of the identities of other posters, with very few exceptions. Otherwise you would have long noted that by now there are a few other good bridge players posting besides Frances and Jlall (there might even be a few that are better than you), and you might have noticed that you don't have to explain the basics of selection bias to me.
Anyway, I am right now not in the mood to write a long explanation of a statement as simple as "the fact that West's hand was of the good tank variety makes it more likely that EW transmitted and used UI in this impossible auction". Suffice it to say that if NS had called the director, West had turned up with a minimum (Kxxxx xx xxx Kxx), but 6H made because ♠Q was on and ♥Q turned up singleton, that would make it much less likely that EW used UI.

[Btw, the logical basis for reversing "if A, then B with 100%" to "if B, then A more likely" is Bayes' law, of course.]

[mikeh]

cherdanno, on May 12 2010, 06:34 PM, said:

Mikeh, one of the oddities that makes some of your posts so annoying is that you don't seem to be able to have any grasp of the identities of other posters, with very few exceptions. Otherwise you would have long noted that by now there are a few other good bridge players posting besides Frances and Jlall (there might even be a few that are better than you), and you might have noticed that you don't have to explain the basics of selection bias to me.
Anyway, I am right now not in the mood to write a long explanation of a statement as simple as "the fact that West's hand was of the good tank variety makes it more likely that EW transmitted and used UI in this impossible auction". Suffice it to say that if NS had called the director, West had turned up with a minimum (Kxxxx xx xxx Kxx), but 6H made because ♠Q was on and ♥Q turned up singleton, that would make it much less likely that EW used UI.

[Btw, the logical basis for reversing "if A, then B with 100%" to "if B, then A more likely" is Bayes' law, of course.]

It may well be that you are more expert in the theory of biases than am I. However, you described my thinking as non-sensical, which suggested to me that one or the other of us is missing something basic. It may well be me. What I was trying to do was put my thoughts into different words, so that either you understood and agreed with me or you could understand and correct me.

I can assure you that I have read many threads on which posters have discussed matters of math and probability. You may have noticed that I never post on those threads...because my understanding of statistics was at an undergraduate engineering student level 40 years ago and i can assure you that I don't remember much of it nowadays. I read those posts with some degree of awe and, it may surprise you to learn, a sense of humility.

And I can assure you that I don't have the sense of bridge superiority that you attribute to me (I am not saying you are unfair in thinking that I do have that sense...my posts tend to create that impression and I do have a tendency towards arrogance)

I once read an anecdote about an Oxford or Cambridge physics professor who was well known for believing a certain theory or view. He was in a seminar when a visiting, younger, physicist gave a presentation utterly demolishing this man's cherished position. The older physicist was the first to congratulate the speaker and thanked him profusely for showing him that he was wrong.

Learning the correct understanding was more important to him than being right. I don't claim to be that objective, tho that story has become a favourite of mine. I do like to think that I can be persuaded that I am wrong...it may take some effort...it may take several whacks upside the head, so to speak, but it can be done.

So if I am wrong about my analogy re the cats (which was not intended solely for those with more knowledge of these matters than I have ever had or ever will have) please whack me upside of the head with an explanation couched in simple language suitable for dummies like me

BTW, has anyone else noticed how huge the E hand was???? Wouldn't most of us be tempted to insist on game over a 1♠ response? Isn't it at least plausible that a non-expert, realizing he'd grossly underbid, decided to try to catch up? I know...even if this is so, there are still issues about whether the BIT made the effort more or less or just as risky as if there were no BIT....I happen to think the BIT made the bid riskier or neutral, not better, but there is room for disagreement there.

[axman]

blackshoe, on May 12 2010, 06:22 PM, said:

Why are you assuming that opener deliberately used UI?

I would think that instead of assume the appropriate word to use is presume. The answer being that L16 requires the player to do so- as it is reasonable to presume that the player strives to satisfy L16 and to do so it is necessary to use UI.

[jdonn]

Mike your argument continues to be flawed IMO, I thought I made a good demonstration earlier with my lawyer/witness analogy but apparently as usual my analogies are only good in my head. However I think your cat analogy leads you to a simply wrong conclusion. A better one would be something like, I think black cats are more likely to be sick, you think white cats are more likely to be sick. We know there is a sick cat behind door number 1. We open the door and the cat is black. That makes it to some degree more likely that I am right than that you are. It still seems very simple to me.

To put it more simply, if you think there is selection bias you should discount the evidence not ignore it. Here that would involve ignoring a number of different points. It's like building a case in court. Responder having a good hand this time that he hesitated doesn't prove anything. Opener bidding more doesn't prove anything. Opener's bidding more being unrelated to bridge logic doesn't prove anything. That slam was a good contract doesn't prove anything. That an individual pair will have a better (read - more accurate) idea of what their UI means than a committee generally will doesn't prove anything. Etc etc etc. But put it all together and you may well have a very strong case.

Bridge appeals don't need absolute proof, they need (some degree of) likelihood. When both partner's actions on the only hand I am given to consider suggest something is the case, then I think that something is probably the case. There is no need to talk myself out of that, like I'm smarter than that little slice of reality or something.

And although I don't really want to get into it, I don't think I have had my argument correctly (or completely) characterized by a single person on the other side who has tried. It's quite frustrating.

[dburn]

awm, on May 12 2010, 02:45 PM, said:

However, the issue is whether 6♥ is much more likely to be successful (even if still less likely to be right than passing 4♥) given the UI.

Well, one can say at any rate that 6♥ is much more likely to be successful in Norway, because if it makes the result will be allowed to stand whether partner had a borderline slam try or a borderline raise to game. I leave it to those with greater mathematical competence than mine (which is to say, almost everyone) to make the required modifications to the Bayesian equations.

[nige1]

awm, on May 12 2010, 02:45 PM, said:

However, the issue is whether 6♥ is much more likely to be successful (even if still less likely to be right than passing 4♥) given the UI.

dburn, on May 12 2010, 09:17 PM, said:

Well, one can say at any rate that 6♥ is much more likely to be successful in Norway, because if it makes the result will be allowed to stand whether partner had a borderline slam try or a borderline raise to game. I leave it to those with greater mathematical competence than mine (which is to say, almost everyone) to make the required modifications to the Bayesian equations.

Sven and Bluejak may be right in law because, in theory, partner is almost as likely to be dithering about passing 3♥ as trying for 6♥.

In real life, unfortunately, David Burn, Ken Rexford and Co are equally right: in an experienced partnership, players usually interpret the meaning of their partner's mannerisms correctly, even when unclear to others.

Like Codo, I wonder why the slam-bidder eventually decided the hand was worth 2 or 3 more tricks than he did initially. The only additional information he received was his partner's tank. Bluejak points out that players do daft things. In his judgement, this change of mind is such a case. Perhaps so, but I think the director should deem a player's actions to be rational until shown otherwise. The director should at least ask the player about his reasoning. The director may still decide that the player is stupid or insane; but some players would prefer to admit to inadvertantly using unauthorised information.

A problem with "Equity law" is that it encourages and rewards infraction. This could be an example. Not all infractions are reported. Directors rarely give PPs for imparting and using unathorised information. The worst punishment you are likely to suffer is the result you would have received, absent your infraction. So the would-be trasgressor is guaranteed substantial long term profits.

Perhaps the law should sanction the shooting of hesitaters?

[barmar]

I think the problem with Mike's selection bias argument is that he forgets that committee members are also experienced bridge players. While they may only see certain situations as a committee, they encounter all the other cases in their normal bridge playing. They're not ruling in a vacuum where they ignore this background knowledge.

To use the vet analogy, the vet doesn't only see cats that people bring to him, he also visits friends who are cat owners, and he gets to see the sick, white cats that are in their homes.

[dburn]

nige1, on May 12 2010, 10:56 PM, said:

Sven and Bluejak to be right in law because, in theory, partner is almost as likely to be dithering about passing 3♥ as trying for 6♥.

If that were actually the case, Sven and Bluejak would be wrong in law; if the probability that partner is thinking about bidding six outweighs the probability that he is thinking of passing three, then bidding six is demonstrably suggested.

However, even if it were the case that the two probabilities are equal, Sven and Bluejak would still be wrong in law; as jdonn and I and others have pointed out, the mere fact that partner is thinking of doing one thing or the other makes bidding six more likely to win IMPs (or avoid losing IMPs) than passing four, and is therefore also demonstrably suggested.

The actual expectation for bidding six is of course [expected IMP gain for bidding six] * P[the score will not be adjusted to making four with overtricks]. The latter probability is of course 1 if you cannot make six, but at least in Norway it is now known also to be 1 if you can.

[mjj29]

Out of interest, are there any instances where a BIT occurred, an option with a less successful LA was selected and the result was bad for the NOS where dburn and co would _not_ adjust? (i.e., where you would use the 'demonstratably suggested' criteria in order to leave the table result as it stood)

[Cascade]

Assumption 1 - A normal contract is reached at the other table 980/480 or -50/450 if this is a slam decision hand, 420/450 etc if it is a game hand, 140/-50 or 420/170 if it is a game decision hand

Bidding 6H improves IMP score if we can make 6H say probability p

Expectation is 11p 980-480 = 11 IMPs (we will either win or save 11 IMPs)

and loses IMPs if we cannot make 6H probability 1-p

Assumption 2 either slam makes or it goes down P(>=12) + P(<=12) = 1 (safety plays etc make this not necessarily so)

Expectation 11(1-p) -50-450 = -11 IMPs -100-420 = -11 IMPs -150--50 = -3 IMPs (but doubled -500--50 = -10 IMPs) or -150-140 = -7 IMPs (but doubled -500-140 = -12 IMPs)

Lets say this is close enough to 11 IMPs (it may be a little less but the math will be easier if we say 11 IMPs and we can revisit a lesser amount later)

So we win IMPs on average by bidding 6H if p>1-p which means that p>50%

Now for p>50% we need P(partner thinking of slam try) * P(12 tricks given that partner was thinking of making a slam try) > 50%

Now when partner has a marginal slam try it is unrealistic to think that slam will always be cold. On the actual hand for example slam could easily go down.

If we assume P(12 tricks given that partner was thinking of making a slam try) = 75%

then for P(partner thinking of slam try) * 75% > 50% requires P(partner thinking of slam try) > 67%

For other values of the P(12 tricks given that partner was thinking of making a slam try) we need

P(12 tricks given ...) P(Partner thinking of slam try)

<50% not possible
50% 100%
60% 83%
70% 71%
75% 67%
80% 63%
90% 56%
100% 50%

For practical likelihoods of slam making given partner has a slam try we need significantly greater than 50% probability that partner was thinking of moving towards slam.

As noted above the IMP expectation when slam does not make maybe lower than the IMP gain from when slam does make. This is complicated because the IMP expectation is dependent upon the probabilities that partner was thinking of the various options which is what we need to estimate.

If the IMP expectation lowers to 10 IMPs then we have 11p > 10(1-p) => p > 10/21 = 48%

If the IMP expectation lowers to 9 IMPs then we have 11p > 9(1-p) => p > 9/20 = 45%

If the IMP expectation lowers to 8 IMPs then we have 11p > 8(1-p) => p > 8/19 = 42%

Even in this last case we would need there to be (8/19) / (3/4) = 56% chance that partner's BIT was based on a mild slam try if slam is on average as good as 75% to meet the threashold.

[iviehoff]

mjj29, on May 13 2010, 01:52 PM, said:

Out of interest, are there any instances where a BIT occurred, an option with a less successful LA was selected and the result was bad for the NOS where dburn and co would _not_ adjust? (i.e., where you would use the 'demonstratably suggested' criteria in order to leave the table result as it stood)

If the Davids would comment on this current thread, we might know: "A shade too ethical" http://forums.bridge...showtopic=39046