[pescetom]

Trinidad, on 2020-February-29, 10:05, said:

But in the reverse case, it doesn't work: Now, the TD decides that the UI suggested the action that was taken and that there was an LA, so guilty. He, again, checks with 4 players. Now, none of them would chose the alternative action. The TD cannot decide that there was no LA, since the sample group is too small to establish this. The reason is that the TD is now looking for the needle in the haystack. In the previous case, he was looking for the haystack around a needle.

This is exactly what happened to me as TD a couple of weeks ago: none of my players would choose an alternative action even though it looked blindly obvious to me. I polled online and all pollees would choose an alternative action and nobody even retained the chosen action an LA.

I'm not convinced that insufficient sample size is the only factor at work here, however - instead I suspect my players were biased by knowledge of the actual outcome of the hand (the chosen action was successful) and the players involved. I don't see that problem disappearing in your hypothesis of five wise players who figure out what happened in perfect understanding of the laws either - ours tend to respect each other but not necessarily those with less experience or ability, plus they are of course the ones most likely to make or necessitate a difficult TD call.

[Trinidad]

pescetom, on 2020-February-29, 11:05, said:

This is exactly what happened to me as TD a couple of weeks ago: none of my players would choose an alternative action even though it looked blindly obvious to me. I polled online and all pollees would choose an alternative action and nobody even retained the chosen action an LA.

I'm not convinced that insufficient sample size is the only factor at work here, however - instead I suspect my players were biased by knowledge of the actual outcome of the hand (the chosen action was successful) and the players involved. I don't see that problem disappearing in your hypothesis of five wise players who figure out what happened in perfect understanding of the laws either - ours tend to respect each other but not necessarily those with less experience or ability, plus they are of course the ones most likely to make or necessitate a difficult TD call.

Of course, knowledge of the hand may make the case more difficult in either way. However, I think that in a group of 5 this problem is smaller than for 5 individuals.

The individuals will be influenced by their knowledge, they make up their mind, and give their answer. End of story. All 5 answers will be biased.

In the group of 5 there is likely to be one individual (or even more) who plays the devil's advocate and says: "I pretty much agree with you, but what if we look at it from the other side? Let's make sure that we have covered all the bases." And now, the unbiased alternative answer is on the table.

Rik

[jhenrikj]

Trinidad, on 2020-February-29, 10:05, said:

Yes, and that is exactly what is the wrong starting point. If you keep them separate, you will not get enough good votes in the poll. If you let them work together, the probability that they will get to the right answer together will be much higher.

It is not as if I am saying something that is revolutionary: Suppose you have 5 people available to solve a complex problem (Doesn't matter what kind of problem: financial, engineering, climate, you name it). Do you know any manager who would let these 5 work separately to gather votes at the end of the process? Of course, you don't. You let them get ideas, discuss them and weight the arguments. This will lead to a much better solution then 5 separate views.

This doesnot mean that polling is useless.
Suppose that a TD decides that the UI suggested the action that was taken, but that there is no LA, so no foul. Before he finalizes his decision, he checks with 4 players. It turns out that they all would have chosen an alternative action that was not suggested by the UI. Oops! The poll just corrected the TD. There is an LA.

But in the reverse case, it doesn't work: Now, the TD decides that the UI suggested the action that was taken and that there was an LA, so guilty. He, again, checks with 4 players. Now, none of them would chose the alternative action. The TD cannot decide that there was no LA, since the sample group is too small to establish this. The reason is that the TD is now looking for the needle in the haystack. In the previous case, he was looking for the haystack around a needle.

You can not reliably determine whether there is a needle in a haystack or not by sampling some parts of the haystack. You can determine whether there is hay in the haystack by sampling some parts of it.

Rik

You are supposed to put the pollee in the same situation as the player. In how many situations is a player allowed to discuss his options with 5 experts before he makes his choice?

We are not interested on knowing what the best/correct answer to the problem is. You are looking for the wrong information. We are only interested on what individuals think is the right answer, by putting them together in a group you have destroyed every possibility to find that out.

[Trinidad]

jhenrikj, on 2020-March-01, 02:50, said:

You are supposed to put the pollee in the same situation as the player. In how many situations is a player allowed to discuss his options with 5 experts before he makes his choice?

We are not interested on knowing what the best/correct answer to the problem is. You are looking for the wrong information. We are only interested on what individuals think is the right answer, by putting them together in a group you have destroyed every possibility to find that out.

You need to start by defining the TD's/AC's problem. You think that the TD's problem is to figure out what call people would make. Let me be very clear about that: The problem is not to figure out what call people would make. You couldn't care less about that.

The problem is to figure out what LAs there are.

We can solve this problem in two ways.

• A poll: We ask a large amount of players: "What would your call be?" and "What are the alternatives?". (This is what TDs do. It would already be better to ask first about alternatives and then about the action they chose, but that is a topic on its own.)
  
• We give our problem, i.e. "We need to figure out what the LAs are", to a group of "experts": experienced bridgeplayers. This group will tell us what the LAs are.

If you realize what the TD's question is (again: "What are the LAs?"), it becomes obvious that asking several people a different question ("What would you call?") is actually taking the long (and, hence, inaccurate) road. This problem is a separate issue on top of the statistical one.

So, we are definitely not asking this group what they would bid. (That would be useless, if not simply silly.)

Rik

[jhenrikj]

If you choose the second option the result has no value what so ever since that does not simulate the position the player is in. The pollees must be put in the same situation. What several individual players think is something you can use, what a group of players think together is of no use at all.

[barmar]

Trinidad, on 2020-February-28, 11:51, said:

Yes. And this is exactly what I mean with the last solution: An interactive discussion with different inputs will be more accurate than n individual opinions, that are simply tallied.

If the TD discusses the case with a few players this will be more accurate than asking some of the available players individually what they would call and some others what the UI suggests.
And, of course, the same holds for the AC.

Rik

Every time I've been polled, I've tried to think out loud so the TD gets my reasoning, not just the conclusion.

[Trinidad]

jhenrikj, on 2020-March-01, 11:21, said:

If you choose the second option the result has no value what so ever since that does not simulate the position the player is in. The pollees must be put in the same situation. What several individual players think is something you can use, what a group of players think together is of no use at all.

You are taking your eye off the ball. You are focussed on polling instread of answering the question what the LAs are.

1. What several players say is hardly useful, because in the cases where it matters, you won't get the numbers you need to say anything accurate. So, forget it.
2. What a group of players think is far more useful, because many more of the possible answers will have been discussed. The same goes for a better understanding of the actual problem. Five people working together are much more thorough than five individuals.

Who says that you need to simulate the situation that the player is in? You need to figure out what the LAs are, and you need to use the best method to do that. In principle, it is a very good idea to simulate the situation the player is in. (I actually think that running a Computer sim might be helpful.) But once you realize that the statistics show that polls often are useless, you should let go of the idea.

And just to be precise. I have never participated in a poll that simulated the situation that the player is in. It never happens. In real life, a TD is approaching a player and typically gives him a piece of paper with a hand and an auction. Then the player is asked what he would do now.

The pollee knows that this is a TD case and that there is a problem and it is either UI or MI. He needs to focus on this question only. He did not see the auction as it proceeeded, so that the idea of what is going in "grows" during the auction. He doesn't get any AI from the opponents' behavior and he is very aware that this is a crucial point in the auction. This is not remotely close to simulating the situation that the player at the table was in.

Rik

[Trinidad]

barmar, on 2020-March-01, 12:22, said:

Every time I've been polled, I've tried to think out loud so the TD gets my reasoning, not just the conclusion.

That is excellent (and I seriously mean that, no sarcasm). But how is the TD supposed to quantify your reasoning?
So, the TD knows, e.g., that Barmar passes because [he thinks the ♦K is poorly placed / he dislikes the honor structure of his hand / bidding may push the opponents into a good game]. How does that help him to establish whether 3NT / 4 ♣ / 4♠ was an LA?

The really useful information for the TD is:
"Pass and 4♠ are both sensible calls. 3NT is not. I would pass." [I think pass and 4♠ are LAs, 3NT is not. It never occurred to me to bid 4♣. The TD should ask about 4♣ if that is relevant.]

or

"Obviously, the choice seems to be between Pass, 3NT and 4♠. I pass and the other calls are simply wrong." [Only pass is a logical action. Again, the TD should ask about 4♣.]

But the above useful information is exactly what "the improved poll" would ask for, because you need less pollees to obtain a somewhat accurate result than in a poll that asks "what would you call?".

Rik

[barmar]

Trinidad, on 2020-March-02, 12:22, said:

Who says that you need to simulate the situation that the player is in? You need to figure out what the LAs are, and you need to use the best method to do that. In principle, it is a very good idea to simulate the situation the player is in. (I actually think that running a Computer sim might be helpful.) But once you realize that the statistics show that polls often are useless, you should let go of the idea.

How would a computer sim accomodate "the class of player"? Do you think we can really write software to simulate the thinking of beginners versus experts?

[Trinidad]

barmar, on 2020-March-03, 10:06, said:

How would a computer sim accomodate "the class of player"? Do you think we can really write software to simulate the thinking of beginners versus experts?

I am not stating that a computer sim would accomodate a class of player, but I think that there are situations where a sim might be useful. Suppose a player choses an action because he reasons that it has a much better expected value than the alternative his opponents claim he had. A sim could test whether he was correct about his reasoning and might help decide whether the claimed alternative is actually "logical" or so inferior that it is no longer logical.

Rik

[pran]

barmar, on 2020-March-03, 10:06, said:

How would a computer sim accomodate "the class of player"? Do you think we can really write software to simulate the thinking of beginners versus experts?

No problem - provided that you can formalize the (complete) characteristics of beginners versus experts in every relevant situation.

But i believe this would be a real problem - probably impossible to accomplish?

(My opinion based on experience with computer systems and programming since January 1963)

[barmar]

Trinidad, on 2020-March-03, 11:11, said:

A sim could test whether he was correct about his reasoning

Why does that matter? Players make judgement mistakes all the time, and actions that result from such mistakes are still logical alternatives.

Beginners especially. If you're trying to figure out what the LAs are for a beginner, you need to ask other beginners -- experienced players, TDs, and computers can't put themselves into the mind of a beginner.

[Trinidad]

barmar, on 2020-March-04, 09:21, said:

Why does that matter? Players make judgement mistakes all the time, and actions that result from such mistakes are still logical alternatives.

Beginners especially. If you're trying to figure out what the LAs are for a beginner, you need to ask other beginners -- experienced players, TDs, and computers can't put themselves into the mind of a beginner.

I agree. Sims are not for beginners.

On the other hand, if an expert (much better than the TD) says: "Given the fact that my partner has shown xyz, I will be able to take 10+ tricks about 75 % of the time, and, therefore, bidding this game at IMPs is the only LA.", a sim could help the TD. It could test whether there are indeed 10+ tricks about 75 % of the time, something the TD might not be able to envision.

If the sim shows that there are 10+ tricks only 50 % of the time, the ruling is simple: The true value is 50 %, the expert was off by 25 %, the error in his estimate (or the "self serving bias"). In a pessimistic mood, this expert could well have erred on the other side and judged that game was only 25%, making bidding it a poor proposition. Staying in a partscore is an LA.

If, OTOH, the sim showed that there were 10+ tricks in 74.78% of the cases, the expert was correct. Staying in a partscore is not an LA for a peer of this player.

Rik

[blackshoe]

A sim. What, exactly, are we supposed to be simulating?

[pran]

blackshoe, on 2020-March-04, 21:40, said:

A sim. What, exactly, are we supposed to be simulating?

Who cares?
"The information comes from a computer so you don't need worrying about understanding. Just accept it"

[blackshoe]

Heh.

[Trinidad]

blackshoe, on 2020-March-04, 21:40, said:

A sim. What, exactly, are we supposed to be simulating?

Many good bridge players run simulations to check their bidding judgement, typically on hands that they have played. I am absolutely not the greatest fan of sims, since often, the underlying assumptions (double dummy play) are questionnable. But sims certainly have a function in testing your judgement after the fact.

If sims have a function in testing your own judgement, they can also have a function in testing someone else's judgement. So, if an expert makes a statement (e.g. "4♠ will make in 75% of the cases.") that a sim could test with a reasonable reliability, then I do not see a problem in using a sim, despite my apprehension in general.

Rik

[barmar]

Trinidad, on 2020-March-04, 10:31, said:

On the other hand, if an expert (much better than the TD) says: "Given the fact that my partner has shown xyz, I will be able to take 10+ tricks about 75 % of the time, and, therefore, bidding this game at IMPs is the only LA.", a sim could help the TD. It could test whether there are indeed 10+ tricks about 75 % of the time, something the TD might not be able to envision.

I think this is just wrong.

The definition of an LA is in terms of what a class of players would consider and do. It doesn't matter whether their judgement is correct or not. If a significant number of those players might take an action, it's an LA, even if a simulation shows that it's a mistake.

We're judging humans by human standards, not an oracle.

[pescetom]

Trinidad, on 2020-March-04, 10:31, said:

On the other hand, if an expert (much better than the TD) says: "Given the fact that my partner has shown xyz, I will be able to take 10+ tricks about 75 % of the time, and, therefore, bidding this game at IMPs is the only LA.", a sim could help the TD. It could test whether there are indeed 10+ tricks about 75 % of the time, something the TD might not be able to envision.

With everyone promoted to his due level of competence (hopefully not beyond) it is inevitable that players will often be better than TD, but not "much better".
Even if so, only the players polled by TD need to have the same vision as this expert, not the TD himself.

Not that I am against doing a sim, if sufficiently flexible and easy to use software was available to TD.

[Trinidad]

barmar, on 2020-March-05, 11:40, said:

I think this is just wrong.

The definition of an LA is in terms of what a class of players would consider and do. It doesn't matter whether their judgement is correct or not. If a significant number of those players might take an action, it's an LA, even if a simulation shows that it's a mistake.

We're judging humans by human standards, not an oracle.

Let's backtrack a little.

I do not advocate that sims are taking over humans. Humans will rule whether an action was an LA or not.

The term simulation came into the discussion because jhenrikj insisted that you can only get an answer to the question what LAs there are, by simulating the position that a player is in. The problem there is that your sample size is simply too small. But if you continue on the theme of simulations and sample size, you will see quickly that computer sims do not have problems with sample size. That's how the computer sim came into the discussion.

In most cases, a computer sim will be utterly useless. I have said that I can envision situations where a sim may be useful to assist the TD. If a player's judgement is correct than only players with the same judgement can be considered his peers. I a player's judgement was wrong, this helps to dismiss the case.

Rik