3 prisoner puzzle

The confusion in both the Monty Hall and 3 Prisoners riddles is the difference between possibilities and probabilities.

There are 100 prisoners, each in confined cells with no communication with the other prisoners. The warden announces that at the end of the week, 99 prisoners will be release and 1 prisoner will be hanged. The warden cannot reveal to any prisoner their fate before the end of the week. The warden randomly selects the prisoner to be hanged: Arthur Dent.

The prisoners get rowdy and demand some information. The warden gives each prisoner a list of 98 prisoners that will be released. To comply with the law, the list provided to each prisoner does not include that prisoner or the prisoner that will be hanged. So, for each prisoner, the list of prisoners that could be hanged would include their name and Arthur Dent, except for Arthur Dent which would include his name and another prisoner. There are only two names not on the list, and you are one of them. Therefore, each prisoner could conclude that his probability of being hanged has just gone from 1% up to 50%!!!!

Each prisoner is in a position where the person hanged is him or one other prisoner. It has to be one of those two. So it is 50%. But how can it be that each of 100 prisoners now has a 50% chance of being hanged when we know that only 1 out of 100 prisoners will actually be hanged?

The probabilities and possibilities have become confused. From the start, the probabilities of getting hanged is 1% (1 prisoner out of 100 prisoners). Also from the start, the possibilities of getting hanged is 50% (Hanged/Not Hanged).

From the start, each prisoner has a 50% possibility of getting Hanged/Hot hanged. However, the probability for the Hanged possibility is 1% and the probability for the Hot hanged possibility is 99%.

In order for the warden to not violate law and give anything away, he can only give enough information so that the probabilities do not exceed the possibilities. We know that at least 98 other prisoners will be released. If the warden gives us the 98 names, it doesn’t matter. From your perspective, it comes down to you and another person, so the probability is still the same at 50%.

Here’s another way to look at the significance of “you” asking the question: The Warden can’t tell you your fate. You are the last to know your fate. So, you have 100 prisoners listed from 1 to 100. You are number 100. Every hour, for 100 hours, a prisoners name is drawn to be not released. The warden then goes down the list of prisoners, and the first prisoner who has not already been released and who’s name has not been drawn, is released. So, as the prisoners are released, we get down to something like:

98. Dick
99. Harry
100. You

A number is drawn. If 98 is drawn, 99 Harry is released. If 99 or 100 is drawn, 98 Dick is released. No matter what the previous draws are, the final draw will come down to 100 You and someone else. So, from the very get-go, 100 You has a 50% chance of being the one hanged. This can be presented to each prisoner where “You” is that prisoner’s name at position 100. So each prisoner could conclude they have a 50% probability of being hanged.

But this is not the case because probability is based on a set number of circumstances. In this case, each prisoner has a DIFFERENT set of circumstances. Each prisoner has a 1% probability of being hanged, but a 50% possibility of being hanged. When only 3 prisoners are involved, the 50% possibility and the 50% probability can be easily confused.

CONCLUSION: If Tom, Dick, or Harry will be executed, Tom gains no information about his own fate by knowing that either Dick or Harry will not be executed. However, if the person to be executed has already been determined, if Tom learns that another prisoner will be not be executed means that the other prisoner who is not to be not executed is more likely to be executed. That doesn’t change Tom’s position. So if the warden told Tom that Dick would be set free, then it is 2/3 likely that Harry will be executed, and still 1/3 likely that Tom will be executed.

However, if the “prisoner to not be set free” is determined each time, then both Tom and Harry are 1/2 likely to be hanged. If Tom is in the “special” position where his fate will not be revealed until the end, then his fate has nothing to do with how many prisoners there are but only with the options for his fate, which is 50% hanged.

Don’t confuse probabilities with possibilities.
 
If nobody else is going to simulate this to show experimentally what the correct answer is, I'll do it [using the fair coin that the warden always carries to make just decisions]:

Trial 1: warden says Dick is to be free; Tom is hung in the morning
[...]
So Toms chance of being hung is 50%. Maybe Tom should have kept his mouth shut after all.

Does the Warden have a coin with three sides?

If you look through the thread, you'll find that I have already simulated the experiment here. A die is used to decide who will be executed. I only ran 16 trials, but you're welcome to continue.
 
Don’t confuse probabilities with possibilities.

I'm trying to understand how you differentiate between probability and possibility. The way you use "possibility" seems to be 1 over "number of different things that could happen". What useful information does this convey that distinguishes it from saying "a thing will happen or it won't"?

Linda
 
If nobody else is going to simulate this to show experimentally what the correct answer is, I'll do it [using the fair coin that the warden always carries to make just decisions]:
Quite clearly no you won't. How many times did you throw the coin for each trial? What exactly did you do? How can a two-sided coin model three possible outcomes?
 
When and how did you determine who would be killed?

You should be using a die and use two numbers per person e.g. *first* to decide who would get killed and then determine who's named by the warden.

If you use a coin to chose if either Dick or Harry goes free, and then use a coin to determine who of the remaining two goes free or dies you are playing a different game.

Are you making up the rules as you go or does it actually say that in the OP?
 
Are you making up the rules as you go or does it actually say that in the OP?

It should be crystal clear from both the OP and the past seven pages of discussion how the game should be played.

It should be equally clear that implementing a coint toss later in the game makes the puzzle extremely pointless since even the most dimwitted reader should be able to see that a toin coss equals a fifty percent chance.
 
The OP says absolutely nothing about how or when the decision of which prisoner will be hanged is made:
There are 3 prisoners, Tom, Dick and Harry. One day, the warden comes up to their cell and says, "One of you will be hanged tomorrow morning, the other 2 will be set free". Naturally, all 3 prisoners become very anxious.

At this point, we may use the symmetry of the problem to say that each prisoner has an equal chance of being hanged because there is nothing that differentiates the three except for their labels. However, when we add the rest of the puzzle:
Tom who's the most anxious of all, managed to pull the warden to one corner and asked him privately, "Please tell me if I'm the one who will be executed tomorrow?". The warden refused to tell him the answer. Tom, who does not give up easily, asked again, "If you won't tell me if I'm the one, then please at least give me the name of one of the other 2 who will be set free tomorrow?" The warden thought for a while and said, "Tomorrow, Dick's life will be spared".

The symmetry is broken because Tom has interacted with the warden so it makes a difference wether the warden had already chosen the one to hang. If the warden had already made the choice, Toms chance was and still is 1/3. If the warden had not already made a choice but had to pick a name to answer Toms second question, then the warden still has a choice of how to pick the one to hang: he could just pick Tom for being the most anxious, he could release Tom for being the first to come forward, he could flip a fair coin to choose between hanging Tom of Harry or he could roll a die and assign Tom a probability of 1/3 and Harry a probability of 2/3 to be hung.

There is nothing stated in the puzzle that says how the choice is made and without symmetry we cannot arbitrarily assign probabilities to the various events so this puzzle is simply unsolvable. There is insufficient information.
 
If the warden had not already made a choice but had to pick a name to answer Toms second question, then the warden still has a choice of how to pick the one to hang: he could just pick Tom for being the most anxious, he could release Tom for being the first to come forward, he could flip a fair coin to choose between hanging Tom of Harry or he could roll a die and assign Tom a probability of 1/3 and Harry a probability of 2/3 to be hung.

There is nothing stated in the puzzle that says how the choice is made and without symmetry we cannot arbitrarily assign probabilities to the various events so this puzzle is simply unsolvable. There is insufficient information.

And maybe the warden will always pick Tom because a guy named Tom stole his lunch money in grammar school. And maybe the probability of drawing the Ace of Spades from a shuffled deck is 100%, because in that particular deck, the Ace of Spades is a giant-sized novelty card while the others are standard size. Maybe the odds of a coin toss landing heads-up side is 0% because the tails side has a spring that keeps it from landing on that side down.

I can make up all kinds of extraneous crap to throw into a basic probability question, just like you can!
 
And maybe the warden will always pick Tom because a guy named Tom stole his lunch money in grammar school. And maybe the probability of drawing the Ace of Spades from a shuffled deck is 100%, because in that particular deck, the Ace of Spades is a giant-sized novelty card while the others are standard size. Maybe the odds of a coin toss landing heads-up side is 0% because the tails side has a spring that keeps it from landing on that side down.

I can make up all kinds of extraneous crap to throw into a basic probability question, just like you can!

What you made up is just irrelevant crap. It could have been Dick or Harry that stole the wardens lunch money so the symmetry still holds. The puzzle however states that it was Tom that pulled the warden aside and then asks a question about Toms chances. If the warden had already chosen the man to be hanged, Toms chances are 1 in 3. If the warden will choose the man to be hanged in the morning, Toms chances are 1 in 2. There is no information presented to resolve which of these (if either) is the actual scenario. If you believe otherwise, please show your case. But don't just throw crap.
 
But...

Tom didn't gain any new information from the warden.

He already knew that at least one of the other two guys would be set free. Knowing which guy changes nothing.

Right?

The probability stays exactly the same whether the Warden says:

"Dick will be set free."

"Harry will be set free."

"I'm not telling you who, but it will be two of you."

All three of those answers provides the same information on Tom's fate, which is none.

Right? If not, why not?

THANK YOU! After reading through several pages, this post finally made it clear to me why Tom's odds of death do not go up, THANKS!

Except now I am wondering why his odds weren't 1/2 from the start...
 
Ok, I thought about this some more, and now I'm back to not understanding it.

I don't see how this is the same as the Monty Hall question. In the Monty Hall version, you are given 3 doors, A B and C, and told that a prize is behind one of them. You pick a door A. You are then told that the prize is not behind door C. Would you like to change your choice to door B?

The answer is that door A and door B have the same probability as each other of holding the prize. But that does not translate to 'there is no difference in the probabilities before and after door C is opened'.

If door C were opened and there was a prize behind it, then the probability of the prize being behind Door A would still be the same as the probability of the prize being behind door B, but they would both be zero.

It's the same as the TV show Deal or No Deal. For every case that you open which does not contain the million dollars, the odds of your case containing the million dollars goes up.

I do not see how it is any different in the prisoner case.

Say the guard has two 'survival tokens'. The odds that Tom will get one of these tokens is 2/3.

Now the guard gives one of the tokens to Dick. There is now one token left, and two prisoners. The odds that the guard will give this token to Tom is 1/2, because he cannot give two tokens to Dick.

If you were betting on which prisoner would live, as in the Monty Hall case, there would be no reason to pick either or. But the odds of getting a survival token have gone down from 2/3 to 1/2.

With two prisoners and one token, I don't see how the odds can be anything other than 1/2.
 
With two prisoners and one token, I don't see how the odds can be anything other than 1/2.

Sure: obviously if one token has already been given away, each of the remaining two has a 50% chance of getting it. But unless you know to whom the first token was given, you don't know who the remaining two are. And that's NOT the information Tom gets from the warden.

Please, stop tormenting yourself and JUST DO THE EXPERIMENT. Recruit a friend, or use a random number generator.
 
If the warden will choose the man to be hanged in the morning, Toms chances are 1 in 2. There is no information presented to resolve which of these (if either) is the actual scenario.
I understand that you're saying that, it is within the wording of the problem, to imagine that the warden had not yet made his choice and the guard just made up something on the spot, then informs the warden after this conversation that he can't pick Dick (heh) so that makes Tom's probability be 1/2.

However, it is certainly implied in the problem's wording that the guard knew at the time that Dick really was in the clear.

Hey, *I* am the one who always points out that the wording of the Monty Hall problem, as it's typically stated, is ambiguous and can't be solved (most of the time when it's told, there is absolutely no constraint, expressed or implied, that Monty must offer the choice to switch). But in the case of the prisoners, the ambiguity you're hanging your hat on is an ambiguity only to you. The fact that it's implied in the problem is enough for most of us. You've made your pedantic point.
 
Here's what I've been chewing on. If Tom comes up to Harry and relates what happened in his discussion with the warden, and Harry says, "No way! I just had the same conversation with him and he told me that Dick will be set free, too." Would the odds then be 50/50?
 
Would the odds then be 50/50?

Just because of symmetry, it would have to be. I'll have to think about how to explain it though. Apparently Tom got more information after talking to Harry, but... what extra info was that?
 
I understand that you're saying that, it is within the wording of the problem, to imagine that the warden had not yet made his choice and the guard just made up something on the spot, then informs the warden after this conversation that he can't pick Dick (heh) so that makes Tom's probability be 1/2.

However, it is certainly implied in the problem's wording that the guard knew at the time that Dick really was in the clear.

Did you get here from a different thread? There is no guard in the OP, only the warden. And since "The warden thought for a while..." it could be implied that the warden had not yet made a decision.

The past tense of "One of you will be hanged tomorrow morning..." clearly implies that the sentence will be carried out during the night possibly by the other prisoners to insure their own release. This would imply that the warden must be psychic to know Dick's fate (and to have predicted the hanging in the first place).

Does this change Tom's chances?
 
Here's what I've been chewing on. If Tom comes up to Harry and relates what happened in his discussion with the warden, and Harry says, "No way! I just had the same conversation with him and he told me that Dick will be set free, too." Would the odds then be 50/50?

As Curt says, symmetry makes it 50/50. More interesting is what happens if Tom, Dick, Harry and Mo are prisoners, and only 1 is going to die, and the warden says "if you come to me privately and ask, I'll randomly choose among those prisoners not including you who will live and tell you one of their names", and then Tom and Harry both independently did, and then talked to each other, and they both were told Dick would be set free (note the only difference is the addition of Mo and a specified generalization).

Symmetry again makes Tom and Harry's chances of dying equal, but how do they stack up against Mo?

Tom, Harry and Mo dying is T, H, and M. The warden saying that Dick will live to Tom is A, and to Harry is B. The background info about someone dying, the warden's plan and honesty, etc. is X.

P(M|ABX) = P(AB|MX)*P(M|X)/P(AB|X) = P(A|BMX)*P(B|MX)*P(M|X)/P(A|BX)/P(B|X)
= (1/2)*(1/2)*(1/4)/P(A|BX)/(1/3)

What's P(A|BX)? It's a little harder... P(A|BX) = P(AB|X)/P(B|X) = [P(ABM|X)+P(AB~M|X)]/(1/3) = [P(A|BMX)*P(B|MX)*P(M|X)+P(A|BTX)*P(B|TX)*P(T|X)+P(A|BHX)*P(B|HX)*P(H|X)]/(1/3) =
[(1/2)*(1/2)*(1/4)+(1/3)*(1/2)*(1/4)+(1/2)*(1/3)*(1/4)]/(1/3) = 7/16

So I get that P(M|ABX) = 3/7, which sets Tom and Harry's chances at 2/7 apiece.
 
Sure: obviously if one token has already been given away, each of the remaining two has a 50% chance of getting it. But unless you know to whom the first token was given, you don't know who the remaining two are. And that's NOT the information Tom gets from the warden.

Please, stop tormenting yourself and JUST DO THE EXPERIMENT. Recruit a friend, or use a random number generator.

The information Tom gets from the Warden is that Dicks life will be spared. That is one 'spared' token used up and one 'spared' token remaining.
 
What if there is a deputy warden that also knows who is to be hung but doesn't know that Tom already talked to the warden. If the deputy tells Tom that Harry will be spared Tom will know his fate. But if the deputy also says that Dick will be spared, what does Tom then know of his chances? (assume that both the warden and the deputy chose randomly and independently when there is a choice).
 
The information Tom gets from the Warden is that Dicks life will be spared. That is one 'spared' token used up and one 'spared' token remaining.

Have you done the experiment yet? Sounds like no... but the reason you are confused is that you cannot seem to put yourself in Tom's shoes, only the warden's. If you refuse to do the experiment, at least do the math. Correctly. Tell me where you think the following breaks down.

Define the following variables:

X = The warden has 2 'spared' tokens and has already determined who will get them. Tom thinks about asking the warden who doesn't get a token, and it is known that if he does ask, once, the warden will tell him exactly one (not Tom) prisoner who will get a 'spared' token, but Tom has no idea which if any the warden will pick if both Dick and Harry will get a 'spared' token.
T = Tom gets a 'spared' token.
D = Dick gets a 'spared' token.
H = Harry gets a 'spared' token.
M = Tom asks the question and is told "Dick will get a 'spared' token."

Now here's what we definitely know:

Equal, independent chances of who will get the 'spared' tokens:
Everyone has a 2/3 chance of living:
P(T|X) = 2/3
P(D|X) = 2/3
P(H|X) = 2/3
Every pair of people has a 1/3 chance of both living (i.e. a 1/3 chance of the third dying):
P(TD|X) = 1/3
P(TH|X) = 1/3
P(DH|X) = 1/3
Not everyone will live:
P(TDH|X) = 0
If we know one person who will live, then the other two are equally likely:
P(T|DX) = 1/2
P(T|HX) = 1/2
P(D|TX) = 1/2
P(D|HX) = 1/2
P(H|TX) = 1/2
P(H|DX) = 1/2

Given knowledge of who is spared, we know how likely the warden's "Dick" answer is:
P(M|TDX) = 1
P(M|THX) = 0
P(M|DHX) = 1/2

Now we know how likely it is that the warden answers "Dick" if we don't know who is spared, it has to be 50/50, but let's calculate it to be sure:
P(M|X) = P(MTD|X) + P(MTH|X) + P(MDH|X)
= P(M|TDX)*P(TD|X) + P(M|THX)*P(TH|X) + P(M|DHX)*P(DH|X)
= 1*1/3 + 0*1/3 + 1/2*1/3 = 1/2

And if we know of just one person who is spared:
P(M|TX) = P(MD|TX) + P(MH|TX) = P(M|TDX)*P(D|TX) + P(M|THX)*P(T|HX) = 1*1/2+0*1/2 = 1/2
P(M|DX) = P(MT|DX) + P(MH|DX) = P(M|TDX)*P(T|DX) + P(M|DHX)*P(H|DX) = 1*1/2+1/2*1/2 = 3/4
P(M|HX) = P(MT|HX) + P(MD|HX) = P(M|THX)*P(T|HX) + P(M|DHX)*P(D|HX) = 0*1/2+1/2*1/2 = 1/4

So finally we can see what Tom's chances are if he asks and the warden says "Dick":
P(T|MX) = P(M|TX)*P(T|X)/P(M|X) = (1/2)*(2/3)/(1/2) = 2/3

Making sure, let's check Dick's chances and Harry's chances:
P(D|MX) = P(M|DX)*P(D|X)/P(M|X) = (3/4)*(2/3)/(1/2) = 1
P(H|MX) = P(M|HX)*P(H|X)/P(M|X) = (1/4)*(2/3)/(1/2) = 1/3

Do you see how P(T|MX) is definitely not the same thing as P(T|X)?
 

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