[phil_20686]

 gordontd, on 2011-March-16, 03:32, said:

Your conclusion doesn't follow from your premise (not that I accept your premise - who was that author whose players at Ann Arbor bridge club seemed to endlessly bid four-card majors with a six-card minor on the side?)

What about 1-4-4-4, 1-4-3-5?

Sure, he can have those hands. Was pretty tired last night, but the point is its essentially impossible for him to have 3 spades. If you want to compare the odds you should be including only 3/8 of all 3-1 distributions (barring singleton J) where the finesse is right, compared with all 2-2. Since originally each 3-1 break was 50%/8=12.5, and we are left with 3. Whereas we still can pick up all 2-2 breaks. Moreover, if you give lho 6 clubs to go with his two hearts, then both hands have the same number of spaces, so there is no reason to prefer the spade finesse to the drop. I think most people were being lazy and only considering the heart layout as "known" and not the bidding inferences. 4 card overcalls tend to imply a good hand, which means south would (barring an 8-2 club break) always have a t/o double if he has 3 spades.

I mean, its obviously somewhat close, but I would expect the drop to be "significantly" better, probably 8-10% or so. That is my intuition anyway.

[gordontd]

 phil_20686, on 2011-March-16, 08:45, said:

Sure, he can have those hands. Was pretty tired last night, but the point is its essentially impossible for him to have 3 spades.

That's why we're not considering taking the finesse the other way

Quote

If you want to compare the odds you should be including only 3/8 of all 3-1 distributions (barring singleton J) where the finesse is right, compared with all 2-2. Since originally each 3-1 break was 50%/8=12.5, and we are left with 3. Whereas we still can pick up all 2-2 breaks.

Shouldn't you only be considering half of 2-2 breaks, since those with Jx in North will have been revealed at the moment of decision? You can pick up all 2-2 breaks, but only if you know they're breaking 2-2. You can also pick up all 3-1 breaks with a singleton in South, but you still need to make a decision.

Quote

Moreover, if you give lho 6 clubs to go with his two hearts, then both hands have the same number of spaces, so there is no reason to prefer the spade finesse to the drop. I think most people were being lazy and only considering the heart layout as "known" and not the bidding inferences.

I thought you shouldn't take unknown suits into account to calculate vacant spaces. Whether I'm write or wrong about that, it was clearly the view shared by nearly all those I asked how they would play it, which is why I concluded that it was "likely" (ie above 50%) that the finesse would be taken.

Quote

4 card overcalls tend to imply a good hand, which means south would (barring an 8-2 club break) always have a t/o double if he has 3 spades.

I don't think anyone has posted anything in contradiction of that.

Quote

I mean, its obviously somewhat close, but I would expect the drop to be "significantly" better, probably 8-10% or so. That is my intuition anyway.

I'd have expected more than intuition for a claim of that magnitude.

[bluejak]

 phil_20686, on 2011-March-15, 17:23, said:

People generally only make 4 card over calls when they have a decent hand, so it seems right to put south with both the club honours. Certainly he must have one.

 gordontd, on 2011-March-16, 09:26, said:

I don't think anyone has posted anything in contradiction of that.

Well, I haven't because I thought the whole analysis was irrelevant, not because I agreed with it.

I make four-card overcalls when I have a good suit that I want led, and am especially likely to make them when I have nothing else I want led. So in my case I would be unlikely to have either club honour.

But I do not see its relevance to the thread.

[gordontd]

I carefully split up Phil's post and responded to each bit separately. You've taken one of my responses and put it beneath something from an entirely different post, as though I had been responding to that.

[phil_20686]

The claim for the adjustment was based on the fact that the "normal" line would lead to two off. I was merely pointing out that the analysis of the "normal" line was clearly flawed. It seems relevant if the "normal" line actually does lead to one off.

Besides that, its interesting in its own right.

@Gordon

Whether you should use "unknown suits" in vacant spaces is often somewhat grey area. The point is that from the bidding you have inferences that assign greater likely hood to some distributions more than others. An obvious example. Suppose the opposition opened 4H and you bid to 6S, and need to find a Q in a two way finesse. The preemptor turns out to be void in spades. Technically that is your only "known" suit, but it would be facile not to conjecture that he will have at least 7 hearts, so one should count that as "known". The point about unknown spaces is mostly that you should not bother conjecturing too deeply about the positions of other suits, as it automatically takes into account the a priori distribution of suits. Perhaps you are confusing it with the the fact that you should not consider discards of a suit as "known" cards.

As to what you should consider, well its a bit deep. If you agree that south should have a good hand, and with the hands I excluded, then the Sim suggested that both work roughly half the time, of the remaining cases the drop works nearly twice as often as the finesse. Of course, if you disagree about what hands to exclude (and its not completely clear cut) then the sim will start to reflect the vacant spaces argument more and more. The key point is that for a 4 card overcall south should always have enough for a t/o double if he has 3 spades, and also that with 6 clubs he will normally have enough extra shape to bid 2C if he had enough HCP for a t/o double in a balanced hand.

Another way to look at it is just to note that the vacant spaces argument predicts 11/20=55% for the finesse, but if you exclude rho from having 3 or 4 spades, then 2-2 distributions make up 40/(40+18+5)= 8/13= 62%. (in my head so approximate). Of course, the finesse also picks up Jx onside, which but the drop also picks up stiff J.

More accurate calculation, excluding RHO having 3 or 4 spades, but using the a priori otherwise:
Assume that you start with a top one and a spade up, then at the decision time:
Finesse: Jxx/Jxx+xx=3/8*49/(3/8*49+0.5*40)= 0.48
Drop: xx/Jxx+xx=0.5*40/(3/8*49+0.5*40)= 0.52

SO the drop is better

The reason the sim probably over-represeted is that these calculations do not accurately reflect the fact that some hand pattern are more common than others. The sin gave vastly more 2425 hands than you would predict from using the a-prior suit% in this way. Its also possible that the sim results were not enough hands to be statistically significant, and hence overrepresented the drop, but I cba doing any more since I have to count them by hand.

[bluejak]

 gordontd, on 2011-March-16, 11:03, said:

I carefully split up Phil's post and responded to each bit separately. You've taken one of my responses and put it beneath something from an entirely different post, as though I had been responding to that.

That is solely because the software does not allow me to copy direct.

But it is the bit you were referring to about four card overcalls showing strength. I am quite sure it is what you were responding to.

Ok, I shall copy it using copy and paste:

=============================================
Quote

4 card overcalls tend to imply a good hand, which means south would (barring an 8-2 club break) always have a t/o double if he has 3 spades.

I don't think anyone has posted anything in contradiction of that.
=============================================


Well, I haven't because I thought the whole analysis was irrelevant, not because I agreed with it.

I make four-card overcalls when I have a good suit that I want led, and am especially likely to make them when I have nothing else I want led. So in my case I would be unlikely to have either club honour.

But I do not see its relevance to the thread.


=====================

Happy now?

[jallerton]

All this discussion about the best line for a declarer who has been paying full attention is interesting from a declarer play point of view, but is of little relevance to the ruling.

The actual declarer claimed the rest of the tricks without stating a line, (presumably) believing that the rest of the tricks were obviously his.

Let's look again at the exact wording of Law 69B.

law 69B said:

B. Director’s Decision

Agreement with a claim or concession (see A) may be withdrawn within the Correction Period established under Law 79C:

1. if a player agreed to the loss of a trick his side had, in fact, won; or

2. if a player has agreed to the loss of a trick that his side would likely have won had the play continued.

The board is rescored with such trick awarded to his side.

What does "if a player has agreed to the loss of a trick that his side would likely have won had the play continued" mean?

The "had play continued" is clear enough. If play had continued, this particular declarer would still have been at the helm; hence my earlier suggestion that the TD should attempt to ascertain the player's state of mind when he claimed.

But what is meant by the word "likely" in the context of "would likely have won"?

One problem here is that the word "likely" is normally used as an adjective often preceded by "very", "quite", "more" or "less", but in this Law 69B2 sentence it has been used as an adverb on its own. Chambers English Dictionary says that the use of "likely" in this way is slang [not good for a formal document!]. Another English dictionary says that "likely" is used as an adverb in American English, so I have delved into my foreign language dictionary collection and looked in Websters. According to Websters "likely" as an adverb means "probably"; "probably" itself is defined in Websters as "in all probability; so far as the evidence shows; presumably".

My conclusion from all of this is that the TD should only transfer a trick which would probably (which implies a significantly greater than 50% chance, say 70%-99%) have been won by the non-claiming side had play continued. If there is judged to be a (say) 55% chance, that is not enough.

Interestingly, Law 69B2 would not appear to allow the TD to transfer a trick which he is 100% sure would have been won by the non-claiming side: "would definitely have won" does not appear to be a subset of "would likely have won" and 69B1 seems to refer to tricks which had been completed before the claim. But I may have misunderstood; I can't believe that this is what the lawmakers intended.

[Cascade]

 jallerton, on 2011-March-27, 11:23, said:

My conclusion from all of this is that the TD should only transfer a trick which would probably (which implies a significantly greater than 50% chance, say 70%-99%) have been won by the non-claiming side had play continued. If there is judged to be a (say) 55% chance, that is not enough.

I am not convinced. I think that something that is more than 50% is probable or likely.

[Echognome]

I'll muddy the waters even further. In statistics, the usage of "likely" is a derivative of "likelihood", which I believe to be synonymous with "probability". Here's a quote from a statistics book:

"There are six faces on a die, and on a fair die each is equally likely to come up when you throw the die."

Here, the likelihood of each event is 1/6. The term "equally likely" can be viewed as "having the same odds" or "having the same probability".

I agree that it would have been much clearer if there was a more descriptive phrase, such as "more likely than not". But there isn't and it is gray.

I will throw my own interpretation in with Wayne's. Another way to view "equally likely" is "equally probable". And if the law book read "would probably won the trick" then I associate that phrase as having a meaning of the cumulative probability of being greater than 50 percent.

[jallerton]

Imagine that you are playing a head-to-head match against team A tomorrow. You rate your opponents as significantly worse than your team. In probability terms, you rate your probability of winning the match as 75%.

Imagine that you are also playing a head-to-head match against team B the day after tomorrow. You rate these opponents as almost, but not quite, as good as your team. In probability terms, you rate your probability of winning as 51%.

If I ask you if you are going to win your match against team A tomorrow, you might well answer: "Probably".

If I ask you if you are going to win your match against team B the day after tomorrow, you might well answer: "Maybe", "Possibly" or "About 50-50". But you would not say "Probably", would you?

[Cascade]

If I had a better than even chance I would say i would probably win.

I might qualify likelihoods in that range in some way from probably and almost certain to probably but could easily lose etc

[dburn]

Idle speculation prompted by the previous three posts prompted these thoughts:

Suppose A, B, C and D enter a dice-tossing tournament consisting of two knock-out rounds.

B, C and D are armed only with regular cubes whose faces are numbered from 1 to 6, but A has got a regular octahedron with faces numbered from 1 to 8.

An octahedral dice (yes, I know what the singular is according to the dictionary, but one should never say "die") will beat a cube with probability 0.642857143 (this is equal to 9/16 + [9/16*1/7], and if you don't know why it should be equal to that, you should study the works of Jeff Rubens until you do).

Who is most likely to win the tournament? Why, A of course. Is A likely to win the tournament? No - he will do so only with the above probability squared, which is less than one half. So, even within a very small range of possibilities, that which is most likely to happen is unlikely to happen.

Should the lawmakers have used the word "likely" in Law 69B2? Probably not.

[gordontd]

 jallerton, on 2011-March-27, 16:44, said:

Imagine that you are playing a head-to-head match against team A tomorrow. You rate your opponents as significantly worse than your team. In probability terms, you rate your probability of winning the match as 75%.

Imagine that you are also playing a head-to-head match against team B the day after tomorrow. You rate these opponents as almost, but not quite, as good as your team. In probability terms, you rate your probability of winning as 51%.

If I ask you if you are going to win your match against team A tomorrow, you might well answer: "Probably".

If I ask you if you are going to win your match against team B the day after tomorrow, you might well answer: "Maybe", "Possibly" or "About 50-50". But you would not say "Probably", would you?

If I rated our chances as 55% and you asked me if we are likely to win (it's "likely" that's used in the wording, not "probably") I'd say yes.

[Cascade]

 gordontd, on 2011-March-28, 00:53, said:

If I rated our chances as 55% and you asked me if we are likely to win (it's "likely" that's used in the wording, not "probably") I'd say yes.

Exactly you are more likely than not to win.

[gnasher]

Do "It is likely that I will win a trick", "I am likely to win a trick" and "I will likely win a trick" all mean the same thing?

[axman]

 dburn, on 2011-March-27, 21:16, said:

Idle speculation prompted by the previous three posts prompted these thoughts:


Should the lawmakers have used the word "likely" in Law 69B2? Probably not.

Where it comes to do-overs, the specification of 'likely' is dubious; unless, of course, the desire is an immense amount of litigation and dissatisfaction.

[suprgrover]

 dburn, on 2011-March-27, 21:16, said:

An octahedral dice (yes, I know what the singular is according to the dictionary, but one should never say "die") will beat a cube with probability 0.642857143 (this is equal to 9/16 + [9/16*1/7], and if you don't know why it should be equal to that, you should study the works of Jeff Rubens until you do).

Isn't this probability just 9/16? That's what I get when I either do the math or run a lot of trials. (Your fundamental point still holds, of course.)

[WellSpyder]

 suprgrover, on 2011-March-28, 08:07, said:

Isn't this probability just 9/16? That's what I get when I either do the math or run a lot of trials. (Your fundamental point still holds, of course.)

I think it depends on what you do if the match is a tie. dburn assumes you toss the dice again, with a further 9/16 chance of the octahedron winning and a 1/8 chance of another tie....

[lamford]

 jallerton, on 2011-March-27, 11:23, said:

Interestingly, Law 69B2 would not appear to allow the TD to transfer a trick which he is 100% sure would have been won by the non-claiming side: "would definitely have won" does not appear to be a subset of "would likely have won" and 69B1 seems to refer to tricks which had been completed before the claim. But I may have misunderstood; I can't believe that this is what the lawmakers intended.

But the claiming side cannot accept the trick either under 79A2: "A player must not knowingly accept either the score for a trick that his side did not win or the concession of a trick that his opponents could not lose." If you are right, therefore, the trick will be unallocated - indeed I could find nothing in the Laws to indicate that a trick had to be won by one side or other, or that the total number of tricks must total 13! I expect there is something there if I look hard enough.

[bluejak]

 dburn, on 2011-March-27, 21:16, said:

A has got a regular octahedron with faces numbered from 1 to 8.

I always wanted one of those for Backgammon. It always seemed silly to me that you do not have a side 1 on the doubling cube. Having played a few 64 games and one at 128 I feel eight faces for the doubling "cube" numbered 1, 2, 4, 8, 16, 32, 64, 128 far more sensible.



The problem with "likely" is simple: it means different things in different contexts, and even different things to different people. Perhaps the only real solution is to get an interpretation from a relevant authority but let us see what we can do.

Law 12C1E includes the following wording "The score assigned in place of the actual score for a nonoffending side is the most favourable result that was likely had the irregularity not occurred." This means that, in the ACBL, it is important to know what "likely" means Furthermore, this was part of Law 12C2 used worldwide under earlier Law books. How does the ACBL interpret 'likely', and did the rest of the world interpret 'likely' for this Law?

I have been told that there is an interpretation in the ACBL that likely is 1 in 3, while at all probable [Law 12C1E2 for the offenders] is 1 in 6. Many years of discussion on newsgroups and forums, talking to my American friends including top TDs, and sitting on and commenting on ACBL ACs have convinced me it may be an official interpretation but it is certainly not followed. The actual usage is probably closer to 1 in 5 for likely and 1 in 10 for at all probable. The EBU came up with an approach that tended to mirror this.

So, how about 1 in 5 for likely? I suppose it means "a reasonable likelihood". But as others have pointed out, if a player said "I am likely to win this match" he probably means he has at least a 6 in 10 chance of winning. Maybe it is the difference in wording, "that was likely" is a lower likelihood, so perhaps Law 12C is a red herring.

Perhaps we should look at logic and fairness, not always the best approach in interpreting the Law book. What would we expect this Law to mean?

Quote

Agreement with a claim or concession (see A) may be withdrawn within the Correction Period established under Law 79C .... if a player has agreed to the loss of a trick that his side would likely have won had the play continued.

Suppose we have this position where a player would have won a trick a certain percentage of the time. What does this Law feel like? Should we give him the trick if he would have won it 15% of the time? 25%? 35%? 45%? 55%? 65%? 75%? 85%?

My feeling is that 55% accords with th feel of this Law. What do you think?