3 prisoner puzzle

Yes, he has 50% chance to survive after he hears the answer. It doesn't matter which name the warden gives since we assumed he doesn't favor either of the names.

There is no contradiction since he has 2/3 chance to survive with 3 unknows and a 1/2 chance to survive when there are only 2 unknowns.

I'll just chime in: Do the experiment!

As far as the contradiction is concerned: No!

If there are some cases where his chances increase from 2/6 initially to 3/6 there must be other cases where his chances decrease!

Let's look at 9 cases. We agree that initially Tom will die in three of these cases.

But you give him 50% every time halfway through. So in the end he will die 4.5 times according to you.

Where do the additional 1.5 deaths come from?

Besides being wrong, which the experiment would demonstrate nicely, you are being plain illogical here.
 
There are 3 combinations to walk free before the warden reveals his information. {Tom, Dick} {Dick, Harry} {Harry, Tom} Tom has now 1/3 chance of dying.

Now let's see. 1/3 of the time the warden has to say Dick. 1/3 of the time he has to say Harry. 1/3 of the time he can say either Dick or Harry. Now the probability after the answer depends wether the warden favors either name in this last situation. But I assume a fair coin. So he says: 1/3(Dick) + 1/6(Dick) = 1/3(Harry) + 1/6(Harry). Now with a fair coin the warden is equally likely to say either Dick or Harry.

That leaves our two remaining survival groups {Tom, Dick} {Dick, Harry} equally likely. Giving Tom a 50% chance to survive after he hears the wardens answer.

Edit:
So my answer to OP is: Tom is correct if the warden acts like a fair coin. Tom is incorrect if the warden favors either name.

Question for you (leading, of course):

What makes the phrase I bolded above true?
 
No, that is your mistake.

Look -

1/3 of the time it's {Dick, Harry}. Then the guard says Dick 1/6th of the time from this possibility, and by the same token Harry 1/6th of the time (assuming he picks one randomly).

1/3 of the time it's {Tom, Harry}, so from this we get a 1/3rd that the guard says Harry.

1/3 it's {Tom, Dick}, so 1/3rd guard says Dick.

Now, IF the guard says Dick, there is a (1/3)/(1/3+1/6) = 2/3 chance it's {Tom, Dick}, and a (1/6)/(1/3+1/6) = 1/3 change it's {Dick, Harry}. So Tom still has a 1/3 chance of dying.

Get it?

Ok, I'll reply to this specifically then. Well ask a crucial question actually. Since we have agreed to a fair coin the warden is equally likely to say either Dick or Harry.

So why does the chance of {Tom, Dick} differ from {Dick, Harry}?
 
It seems a different situation than the Monty Hall problem because in this case, the knowledge doesn't change who gets executed, while in that case, the contestant has a choice of changing the outcome by choosing a different door.

What does change is the probility that it will be Tom as the field narrows and he is still in it.

It would be the Monty Hall problem if Tom were given the chance to change identies (as far as the executioner is concerned) with one of the other prisoners. If that were the case, should he change? Living is the goat and dying is the car...
 
Last edited:
Let's play a card game. Find the Ace of Spades out of three cards.

After you have a chosen a card (You have 1/3 chance of winning.) I will remove one incorrect card and show it to you. Now your chance of winning has gone up from 1/3 to 1/2.

Let's play a different game. Three cards again but now you try to avoid the Ace of Spades.

After you have chosen a card (You have 2/3 chance of winning.) I will remove one correct card and show it to you. Your chance of winning has come down from 2/3 to 1/2.

Right?
 
Let's play a card game. Find the Ace of Spades out of three cards.

After you have a chosen a card (You have 1/3 chance of winning.) I will remove one incorrect card and show it to you. Now your chance of winning has gone up from 1/3 to 1/2.

No, it hasn't.

Let's play a different game. Three cards again but now you try to avoid the Ace of Spades.

After you have chosen a card (You have 2/3 chance of winning.) I will remove one correct card and show it to you. Your chance of winning has come down from 2/3 to 1/2.

Right?

No. You haven't added any new information by showing me the card.

Before you show me anything, to the best of my knowledge I have a 1/3 chance of having the Ace of Spades. You have 0 chance of showing me the Ace of Spades. My chances, as far as I know, of having the Ace of Spades hasn't changed. If, however, you pick a card at random and show it to me, you have given me information because 1/3 of the time you will show me the Ace of Spades. That you didn't increases the chances that I have it.

But, do the experiment.
 
It seems a different situation than the Monty Hall problem because in this case, the knowledge doesn't change who gets executed, while in that case, the contestant has a choice of changing the outcome by choosing a different door.

What does change is the probility that it will be Tom as the field narrows and he is still in it.

It would be the Monty Hall problem if Tom were given the chance to change identies (as far as the executioner is concerned) with one of the other prisoners. If that were the case, should he change? Living is the goat and dying is the car...

That really doesn't matter because the question being asked is what are the chances that the contestant has the car once he knows it's not behind a particular door. The question of whether or not to switch is simply a decision based on that knowledge and can be removed without changing the puzzle.
 
Lets play a third game. Three cards again and again I try to avoid the Ace of Spades. This time I am stuck with my original card, no changes. You show me a non-ace card.

Tom should be even more worried after hearing Harry or Dick will go free. Even if he were allowed to change identies, his chances are even Steven he will die.
 
Let's play a card game. Find the Ace of Spades out of three cards.

After you have a chosen a card (You have 1/3 chance of winning.) I will remove one incorrect card and show it to you. Now your chance of winning has gone up from 1/3 to 1/2.

Did you actually play this game and count the number of times you found the ace. Or are you theorizing?
 
That really doesn't matter because the question being asked is what are the chances that the contestant has the car once he knows it's not behind a particular door. The question of whether or not to switch is simply a decision based on that knowledge and can be removed without changing the puzzle.
Yes, you're right. The problem is (slightly) different because there are two cars and one goat. In this problem and that, the probability does change, but in different directions.
 
Before you show me anything, to the best of my knowledge I have a 1/3 chance of having the Ace of Spades. You have 0 chance of showing me the Ace of Spades. My chances, as far as I know, of having the Ace of Spades hasn't changed. If, however, you pick a card at random and show it to me, you have given me information because 1/3 of the time you will show me the Ace of Spades. That you didn't increases the chances that I have it.

(Bolding mine.)

Thanks. Now I see I was wrong. I didn't realise the difference bewteen those two cases.

So to anyone who has read the thread in order: Ignore what I said earlier :D

Edit: After reconsidering I'm not so sure. You might end up ignoring this post after all :D. I'll give it some thought overnight. Watch this space!
Edit2: Yep, wrong. And wrong again.
Last Edit: I'll think on this afterall. I'll post tomorrow.
 
Last edited:
Yes, you're right. The problem is (slightly) different because there are two cars and one goat. In this problem and that, the probability does change, but in different directions.

No, it doesn't for reasons previously explained.
 
There's a difference between picking a card at random and knowingly exposing a non-ace of spades.
 
Did you actually play this game and count the number of times you found the ace. Or are you theorizing?
I actually did, but with pieces of paper with "car" and "goat" written on them. It's true that you should switch.
 
Last edited:
Ok, I'll reply to this specifically then. Well ask a crucial question actually. Since we have agreed to a fair coin the warden is equally likely to say either Dick or Harry.

So why does the chance of {Tom, Dick} differ from {Dick, Harry}?

Read what I wrote before - I laid it out very carefully. Here it is again, arranged differently:

The chance of {Tom, Dick} is the initial probability of {Tom, Dick} - 1/3 - times the probability the guard says Harry in that case - 1 - divided by the total probability the guard says Harry - 1/2. So it's (1/3)*1/(1/2) = 2/3.

The chance of {Dick, Harry} is the initial probability - 1/3 - times the probability the guard says Harry - 1/2 - divided by the total probability the guard says Harry - 1/2. So it's (1/3)*(1/2)/(1/2) = 1/3.

This is Bayes' theorem, by the way, but it's also much less complicated than that makes it look. Just make a table of the possibilities.
 
Last edited:
(Bolding mine.)

Thanks. Now I see I was wrong. I didn't realise the difference bewteen those two cases.

So to anyone who has read the thread in order: Ignore what I said earlier :D

This is precisely where most people get these problems wrong.

In an earlier life, [mumble] years ago, I made the same mistake and it took actually writing a program to do the simulation for me to find out where.

The funny part was I didn't actually have to run it. Just writing the code showed me where my error was.
 
I've only got a few minutes, but I think you're wrong. See Tom has learned something, he's learned that a specific person will be hang. As soon as a person is specified, Tom's chances are reduced to 1/2.

Knowing a specific person will hang is statistically not the same as knowing that one or the other will hang.

Two possible people could have been identified - Dick & Harry. Now, two outcomes have been reduced to one. That's the extra information.

The only way Tom's chance could remain 1/3 is if the warder said, "Well, one of them will hang" because two choices still remain.

It's easier just to look at it with respect to the current situation, which is that only two people can possibly hang and Tom is one of them. Therefore, irrespective of what the chances were before, they now have to be 1/2.


But how is knowing which of Dick and Harry will be free any different than knowing that either Dick or harry will be free?

Tom could have done jsut as much without talking to the warden. He could have thought to himself "Well, one of the other two will be set free, that's for sure. So it's down to me and one of the other two, who's to hang. It's one of two of us. So I've got a 50/50 chance."

But that's certainly not right. We know that there's a 1/3 chance, at least to start. And Tom, without learning anything specific from the warden, can justify the logic of a 1/2 chance? Can't be right. Therefore, Tom hasn't learned anything from the warden.
 
Yes, you're right. The problem is (slightly) different because there are two cars and one goat. In this problem and that, the probability does change, but in different directions.


They are formally identical, as several people have explained. For example, I said here:
... getting hanged being the 'prize' (:(), getting reprieved being the goats, Tom being the door chosen by the contestant, and Dick and Harry being the other two doors (Dick corresponding to a goat).


You just have to ignore the fact that the 'prize' isn't a car but a hanging! Tom corresponds not to the contestant but to the contestant's chosen door – there actually isn't anyone corresponding to the contestant in the prisoner version. Well, except that we do, I suppose – if we're trying to guess the unlucky recipient then we should switch our guess from Tom to Harry after the warden's answer.
 

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom