3 prisoner puzzle

So when you talk about a fair coin you are thinking along the lines of abiding by the rules of physics and not necessarily that the coin has an equal probability of coming up heads or tails.

No, I talk about something different then.

A fair game or roulette has a house edge. An unfair one has magnets.
 
No, I talk about something different then.

A fair game or roulette has a house edge. An unfair one has magnets.

Yes. A game is fair if the probability of winning is what it seems to be. The roulette game is fair if each the wheel has an equal chance of stopping on any particular number. The house has a clear advantage since it doesn't pay out when the wheel stops on zero.
 
If a game is fair, neither side looses more than the other. I choose to play a game to demonstrate the point that different information was available to the different characters. In particular, there was no information withheld from the warden so the probability of any prisoner being executed was either 0 or 1 from the wardens perspective.

If you go back to the scenario in the OP but instead of just the one prisoner questioning the warden, all three separately question the warden and don't communicate with each other. In all cases, we can look at the information that all three prisoners have and conclude which prisoner will be executed but each prisoner will still only know that they have a 1/3 chance of being executed.

Each prisoner will still have a 1/3 chance of being executed according to the information that they possess. Depending on the information available to you, you calculate different probabilities.

For instance, imagine that the warden has been told that Harry is to be killed. To Tom he says that Dick will be let free, to Dick he says that Tom will be let free and to Harry he says that Tom will be let free.

From the warden's point of view, the probabilities of each prisoner being executed are:
Tom 0, Dick 0 and Harry 1

From Tom's point of view, the probabilities are:
Tom 1/3, Dick 0 and Harry 2/3

Dick sees it thus:
Tom 0, Dick 1/3 and Harry 2/3

Harry calculates that the probabilities are:
Tom 0, Dick 2/3 and Harry 1/3

The prisoners assume that in the absence of any other information each of them has an equal chance of being chosen. If they have access to more specific information (one of the prisoners has friends in high places, one of them is convicted for murder and another for forging his driving licence, whatever), they will be justified in calculating the probabilities differently.
 
The problems are only superficially different. The hero in this problem can pretend that each prisoner is a door in the Monty Hall problemWP and that he himself is the door that was chosen first. And when the warden gives him the name of a prisoner that will not be executed, he can pretend that that prisoner is the door that was opened to reveal a goat. Now he can guess who will be executed (which ironically corresponds to winning the car). Is it him, or the other guy? If the probability that it's him is P, then the probability that his "first choice" was correct is also P, but that's impossible unless P=1/3.
We've already been there (several times).


If a game is fair, neither side looses more than the other.
A fair game or roulette has a house edge. An unfair one has magnets.
Yes. A game is fair if the probability of winning is what it seems to be. The roulette game is fair if each the wheel has an equal chance of stopping on any particular number. The house has a clear advantage since it doesn't pay out when the wheel stops on zero.
I disagree with Rasmus and Michael C, and agree with Dan's definition (assuming both sides calculate their bets correctly). A "fair game" is one with no edge, i.e. the stakes and payouts correspond to the probabilities. However, it's not necessary for the probability of each outcome to be the same, just the average payout on each outcome. A fair game corresponding to the one Dan proposed (Tom bets on Harry to be executed vs the warden betting on Tom to be executed) would require Tom's stake to be twice the warden's.


If you go back to the scenario in the OP but instead of just the one prisoner questioning the warden, all three separately question the warden and don't communicate with each other. In all cases, we can look at the information that all three prisoners have and conclude which prisoner will be executed but each prisoner will still only know that they have a 1/3 chance of being executed.
OK, here's a variation on the original puzzle:

Each prisoner separately questions the warder, and we eavesdrop. As Dan says, if we overhear all three conversations then we can be sure who's the unlucky one. And we already know (at least, most of us do) that if we overhear only one conversation then we can re-assign the probabilities as 0, 1/3 and 2/3.

So, what happens if we overhear two of the conversations? What's the probability that we will have enough information to identify the unlucky one?
 
So, what happens if we overhear two of the conversations? What's the probability that we will have enough information to identify the unlucky one?

2/3.

And interestingly in the other 1/3rd of cases the odds are 1/2 for which of the other two it will be.
 
Last edited:
2/3.

And interestingly in the other 1/3rd of cases the odds are 1/2 for which of the other two it will be.
Yes, in this case the answer that pops into the head turns out to be the right one when you go through the combinations – which doesn't always happen in probability problems. (My immediate thought: in 1/3 of cases the same name will be given.)


Support for Dan's definition (and mine) of a "fair game":

Casino, Sports Betting, and Lottery - A Comparison of the House Edge
A fair game or a fair bet is one in which no commission is involved; an obvious model is a game between friends. Therefore, in percentage form, the house edge is obtained by subtracting an actual game winning from a fair game winning and dividing the result by the fair game winning, finally multiplying by 100. Thus,
House Edge = [ (fair game winning - actual game winning) / fair game winning ] * 100


Merriam-Webster's Guide to Everyday Math: A Home and Business Reference
Any game in which the expectation is 0 – which means the bettor wins as much as he or she loses – is called a fair game. In a fair game, gains are offset by losses.
 
2/3.

And interestingly in the other 1/3rd of cases the odds are 1/2 for which of the other two it will be.


Same answer, probably same reasoning:

WLOG, first we overhear tom; guard says "dick" lives:
2/3 chance tom lives,
1/3 chance harry lives.

1/2 chance to overhear dick next; guard says "tom" or "harry" lives:
always know.

1/2 chance to overhear harry:
2/3 chance tom lives;
assume mental 'coin flip' : 1/2 chance guard will say "dick", (don't know), 1/2 chance guard will say "tom", (know).
1/3 chance harry lives;
guard must say "dick" (which tells us dick -- don't know).

So, 1/2 + 1/2 x 2/3 x 1/2 = 2/3.
Nice twist, Lucky. :)
 
Last edited:
Without meaning to blow my own trumpet, there was never any doubt in my mind that Tom's probability of death was still 1/3 in the original problem and I couldn't really see why anyone would think otherwise. Then it was pointed out that this problem is identical to the Monty Hall problem. The latter took me a good hour to get my head around. So I was wondering why such a difference might occur? Does it say something about my ability (or lack of) at problem solving? I first learned about the Monty Hall problem about 5 years ago but I'm struggling to believe its just experience of statistics that caused the difference.
 
Without meaning to blow my own trumpet, there was never any doubt in my mind that Tom's probability of death was still 1/3 in the original problem and I couldn't really see why anyone would think otherwise. Then it was pointed out that this problem is identical to the Monty Hall problem. The latter took me a good hour to get my head around. So I was wondering why such a difference might occur? Does it say something about my ability (or lack of) at problem solving? I first learned about the Monty Hall problem about 5 years ago but I'm struggling to believe its just experience of statistics that caused the difference.

I think the most interesting thing about these two puzzles is why people find them so non-intuitive. Personally I wasn't confused by either, but I work with probabilities (and numbers in general) every day as a profession, so I have lots of practice thinking about these things (and am quite wary of the mental traps you can fall into).

One thing they demonstrates quite cleanly is that human thought does not proceed according to the rules of logic... and that it's sometimes very difficult to force oneself to think that way if you're not used to it.

It reminds me of the following famous cog psy experiment:

The experimenter says: "Here are four cards, each with a number on one side and a letter on the other" (you look down and see four cards on a table, labeled A, F, 3, 6 respectively). "The claim is, if there's a vowel on one side of a card then there's an even number on the other. Question - which of those cards must you flip over to check whether that is true?"

This is an utterly trivial problem in formal logic (in a sense it's the most simple question you can possibly ask), and yet only a tiny fraction of people answer correctly when you do the experiment.
 
Last edited:
"A" won't do it.
You also have to turn over the "3"
... and check any cards that aren't on the table
because the claim didn't limit itself to only those 4 cards
 
Ah. Of course. Misunderstood the question. I'd reword it slightly.

All but the six.

And I'd reword it to say "prove" instead of "check". I also assumed you were limited to one card, but that isn't implied by your wording, just my mistake.
 
Last edited:
Ah. Of course. Misunderstood the question. I'd reword it slightly.

All but the six

No, that's also wrong. But you've illustrated the point perfectly - if the human brain worked logically, you couldn't possibly make a mistake answering this (let alone two).

Don't worry - if I remember correctly, the percentage of people that answered correctly (when the experiment was done in person) was less than 2%.
 
Last edited:
OK. l I guess until I hear the correct answer I won't know whether this is actually testing cognition or the occasional vagueness of the English language.
 
OK. I'll take another shot, but if this is right I think this is more a matter of vagueness of languages.
Only the A and the 3.

That's assuming the initial statement about the nature of the cards can be trusted. And also what the word "that" may be referring to. Does "that" refer to just the claim or the entire statement of the situation?
 
OK. l I guess until I hear the correct answer I won't know whether this is actually testing cognition or the occasional vagueness of the English language.

Dan O.'s answer in his post above is correct.

I don't think the question is vague, but that doesn't mean it isn't. If you do an experiment like this it's true you have to be very, very careful with the wording, and I wasn't particularly.

EDIT: Correct!

That's assuming the initial statement about the nature of the cards can be trusted. And also what the word "that" may be referring to. Does "that" refer to just the claim or the entire statement of the situation?

Yes, the "that" should be replaced with "that claim" (which is what I originally intended to type, but didn't for some reason). However if you thought the claim to be tested included the entire statement, the correct answer is none of the three you gave :).
 
Last edited:
However if you thought the claim to be tested included the entire statement, the correct answer is none of the three you gave :).
I don't think this reply needs a spoiler. Yes, if you need to prove the initial assumption also then you need to turn over all the cards.

Do they ask the subjects about their thinking in arriving at their answer? The key mistake I made was a matter of poor reading: assuming one card was implied. The second was taking the word "check" at face value.
 
Back to the OP, suppose after the warden leaves, Tom spills the beans and tells Dick and Harry what he learned.

Tom's estimation of his chance of dying remains the same at 1/3. Dick feels elated. Poor Harry now suffers with the knowledge that he has a 2/3 chance of dying.
 

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