Tom K. answered 08/16/20
Knowledgeable and Friendly Math and Statistics Tutor
Actually, there should be a choice of whether to put it back or not depending upon whether the first ball is red or black. Let's consider all four choices (put back or don't put back when first choice is red or black); the metric to use for the decision has not been given, so let's say it is the accuracy and assume that it is equally likely to choose either of the 2 jars
if choose with replacement if black
P(jar 1, black, black) = 1/2 * 1/3 * 1/3 = 1/18
P(jar2, black, black) = 1/2 * 100/201 * 100/201 = 5000/40401
5000/40401 > 1/18; select jar 2; misclassification 1/18
P(jar 1, black, red) = 1/2 * 1/3 * 2/3 = 1/9
P(jar2, black, red) = 1/2 * 100/201 * 101/201 = 5050/40401
5050/40401 > 1/9; select jar 2; misclassification 1/9
Total misclassification: 1/18 + 1/9 = 1/6
if choose with replacement if red
P(jar 1, red, black) = 1/2 * 2/3 * 1/3 = 1/9
P(jar2, red, black) = 1/2 * 101/201 * 100/201 = 5050/40401
5050/40401 > 1/9; select jar 2; misclassification 1/9
P(jar 1, red, red) = 1/2 * 2/3 * 2/3 = 2/9
P(jar2, red, red) = 1/2 * 101/201 * 101/201 = 10201/80802
2/9 > 10201/80802; choose jar 1; misclassification 10201/80802
Total misclassification: 1/9 + 10201/80802 = 2131/8978
if choose without replacement if black
P(jar 1, black, black) = 1/2 * 1/3 * 0 = 0
P(jar2, black, black) = 1/2 * 100/201 * 99/200 = 99/804
99/804 > 0; choose jar 2; misclassification 0
P(jar 1, black, red) = 1/2 * 1/3 * 1 = 1/6
P(jar2, black, red) = 1/2 * 100/201 * 101/200 = 101/804
1/6 > 101/804; choose jar 1; misclassification 101/804
Total misclassification: 0 + 101/804 = 101/804
if choose without replacement if red
P(jar 1, red, black) = 1/2 * 2/3 * 1/2 = 1/6
P(jar2, red, black) = 1/2 * 101/201 * 100/200 = 101/804
1/6 > 101/804; choose jar 1; misclassification 101/804
P(jar 1, red, red) = 1/2 * 2/3 * 1/2 = 1/6
P(jar2, red, red) = 1/2 * 101/201 * 100/200 = 101/804
1/6 > 101/804; choose jar 1; misclassification 101/804
Total misclassification: 101/804 + 101/804 = 101/402
If we choose replacement (both with black and red first),
total misclassification = 1/6 + 2131/8978 = 5441/13467 = .4040
If we choose without replacement (both black and red first)
101/804 + 101/402 = 303/804 = .3769
If we choose without replacement with black and with replacement if red
101/804 + 2131/8978 = 19553/53868 = .3630
If we choose with replacement with black and without replacement if red
1/6 + 101/402 = 19553/53868 = 28/67 = .4179
If you have to choose sampling with or without replacement and can't consider the first selection, then the best choice is sampling without replacement, as .3769 < .4040
However, it would be even better if you could select without replacement if the first selection is black and with replacement if the first selection is red, as the probability is .3630, which is less than .3769
Note that, in this case, the second selection matters, both if the first selection is red or black; otherwise, in either the black or the red, the second selection won't matter. You might wonder then, does the first selection matter if we are choosing based on the second selection only; the answer is that it does because the sampling technique changed based upon the first selection.
In the worst case, with replacement with black and without replacement if red, your second selection does not matter, so you might as well not even make the second selection.
Xenoyr M.
Aren't you supposed to look for the greater misclassification?08/19/20
Tom K.
No, you want to minimize misclassification. Now, if we want to consider that the other person isn't going to pick 1/2 from each, you do have to change your strategy into a different strategy.08/21/20
Kalokoy E.
The color of the first ball isn't given however. Is it possible to find out what choice (put back the ball or keep the ball) allows for better winning chances? If so, what's the right move/decision, and if not, why?08/17/20