A man is stranded on an island. A benevolent genie presents three boxes, 23 white marbles, and 7 black marbles and instructs the man, "You may distribute the marbles into the boxes any way you see fit, but you must use all of the marbles. Once you finish, you will choose a box at random and then choose a marble from that box at random. If the marble is white, then I will help you escape from this place."
Assuming the man distributes the marbles in his best interest, what is the probability that he escapes the island?
This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try
refreshing the page, (b) enabling javascript if it is disabled on your browser and,
finally, (c)
loading the
non-javascript version of this page
. We're sorry about the hassle.
This gives the answer but doesn't fully explain why this choice is optimal. I present justification that this result is optimal.
Rather than maximizing the chance of escaping, we will minimize the chance of not escaping (calculating the complement).
We need to show that splitting up the black marbles into multiple boxes is not optimal. It is easy to show that for two groups, the optimal choice is to split the marbles evenly. a k + a k < a + b k + a − b k This should be fairly straightforward to prove. I leave this to the reader.
Now we need to show that combining the groups is strictly better than the best choice of splitting between two groups. This should be quite trivial to do. b a < b / 2 a / 2 + b / 2 a / 2
Finally, we generalize and say that for any amount of groups, combining any two of them is strictly more optimal, so repeatedly applying this strategy leaves us with one group. Optimizing this last group is trivial and we are done.
Log in to reply
I think there is a better way to explain about that "optimal set up".
Before we go, please note that we need to choose white pearl from three boxes. In other words, what we need to actually choose is A box which had white pearl in it .
To optimize our chance in choosing the right box, we need to do two things.
Raise the white chance to 100%, simply by distribute the white in all boxes, 3 3 .
Lower the black chance to 0% or null. Now this one is actually impossible. Why? As from the question had told us, we need to use ALL the pearls. Therefore, it leaves us with the only lowest chance we can create, by putting black pearls in one box, 3 1 .
Based on that, we can simply estimate how much pearls we need to put in each box.
Black | White | |
Box A | 0 | 1 |
Box B | 0 | 1 |
Box C | 7 | 21 |
Putting 2 or more white in box A or B is rendered useless. It won't change the value of chance since it was already a 100% white with no black.
But, if we lowered the number of white in box C, it reduces the value of chance. And therefore, we must keep the number of white as maximum as we could (remember that 2 8 2 1 > 2 7 2 0 )
Log in to reply
beautifully explained sir! Regards, Adithya
what about lowering the number of black marbles in box C? I followed your same reasoning on my own but I thought it was incomplete
i think this is a great method of proving what wasn't proven in the original answer but can you tell me what a, b, and k represent in the problem? thanks!!!
Maybe I misunderstood but I disagree with the part that says "for two groups, the optimal choice is to split the marbles evenly". For two groups, the optimal choice is still to have only one white in one box, and the rest in the other, right?
The question can be reformulated as a straightforward integer program to minimize the probability of selecting a black marble, which converges to 1/12. Consequently, the best case scenario is 11/12.
nice question
This is an interesting optimization problem! For simplicity, let b : = 7 , w : = 2 3 be the total number of black and white marbles, respectively, and define for i = 1 ; 2 ; 3 :
b i , w i : B i : E : number of black and white marbles in chest i event, that we choose chest i with probability P ( B i ) = 3 1 event, that we escape the island
We calculate the probability to escape the island using Bayes' Theorem. The black and white marbles satisfy two boundary conditions: P ( E ) = i = 1 ∑ 3 P ( E ∣ B i ) P ( B i ) = 3 1 i = 1 ∑ 3 P ( E ∣ B i ) , b 1 + b 2 + b 3 = b , w 1 + w 2 + w 3 = w
Before we start optimizing, consider two edge cases first:
Case 1: One chest does not contain any white marbles. If we pick that chest, we stay on the island for sure. The other two chests let us escape with a probability less than or equal to one: P ( E ) ≤ 3 1 ( 0 + 1 + 1 ) = 3 2
Case 2: Two chests contain exactly one white marble. If we pick either of those, we escape the island for sure. The remaining marbles determine the probability of the third chest: P ( E ) = 3 1 ( 1 + 1 + w − 2 + b w − 2 ) = 3 2 + 3 1 ⋅ w − 2 + b w − 2 = 3 2 + 3 ⋅ 2 8 2 1 = 1 2 1 1 Notice the first case has less probability - it is clearly non-optimal, so every chest must contain (at least) one white marble!
We want to show that case 2 is indeed optimal. We prove: "Any optimal solution has exactly one chest, that may contain something other than one white marble". To begin with, we examine only one pair of chests and leave the rest alone. By symmetry, we can work with chest 1 and 2: K : = P ( E ∣ B 1 ) + P ( E ∣ B 2 ) = 3 1 ( w 1 + b 1 w 1 + w 2 + b 2 w 2 ) , b 1 + b 2 = : b ′ ≥ 2 , w 1 + w 2 = : w ′ ≥ 2 Let's redistribute the marbles of only these two chests, so that one contains exactly one white marble and the other contains the rest. To shorten notation, we define x i : = w i + b i ≥ 1 with ( ∗ ) x 1 + x 2 = w ′ + b ′ . Compare the probabilities before and after redistribution: w 1 = 1 , b 1 = 0 : K ′ K ′ − K : = P ( E ∣ B 1 ) + P ( E ∣ B 2 ) = 3 1 ( 1 + w ′ − 1 + b ′ w ′ − 1 ) ( ∗ ) = 3 ( x 1 + x 2 − 1 ) 2 ( x 1 + x 2 ) − b ′ − 2 = K ′ − 3 1 ( x 1 w 1 + x 2 w 2 ) = … = 3 x 1 x 2 ( x 1 + x 2 − 1 ) b 1 x 2 ( x 2 − 1 ) + b 2 x 1 ( x 1 − 1 ) ≥ 0 Note that K ′ ≥ K - we have increased our probability by creating a chest with exactly one white marble! We can repeat this step until all but the last chest contain exactly one white marble. After a finite number of steps, we end up with case 2, and our probability is now greater than or equal to the one we began with! We conclude, that case 2 is optimal: P ( E ) ≤ 1 2 1 1
Rem.: I phrased the optimization step a bit more general than necessary - it should work for any number of chests less than or equal to the number of white marbles!
Rem.: It is important that we have only a finite number of steps in our optimization algorithm for the proof to work. Sadly, I cannot recall why (or if) a countable number of steps might lead to problems. If someone wants to elaborate, I'm more than interested in an opinion!
Problem Loading...
Note Loading...
Set Loading...
The man's best strategy is to place one white marble in the first two boxes and then the remaining marbles in the third box. This guarantees that he'll escape the island if one of the first two boxes are selected. If the man selects one of the boxes with only one white marble in it, then the probability that the man draws a white marble when selecting one of those boxes is 1. The remaining box will have 21 white marbles and 7 black marbles, so when that box is selected, the probability that the man draws a white marble from that box is 4 3 . Now we sum the probabilities that he draws a white marble over the boxes:
( 3 1 ) ( 1 ) + ( 3 1 ) ( 1 ) + ( 3 1 ) ( 4 3 ) = 1 2 1 1
Therefore, the probability that the man escapes the island is 1 2 1 1