Hi onlyPlanA,Your
6-face setup is spot on:
4 heads, 2 tails, each face equally likely. The whole problem breaks right after that, at the number
1/3.
Here's the exact slip. Once you know a
head is showing, you're no longer choosing among all
6 faces - you've been told the face on top is one of the
4 heads-faces. So your new universe is just those
4, and you count
within them:
- Fair coin 1's heads -> tails on the flip
- Fair coin 2's heads -> tails on the flip
- Double coin's heads (face 1) ->
heads on the flip
- Double coin's heads (face 2) ->
heads on the flip
So P(flip is heads | a head showed) =
2/4 = 1/2, which makes P(flip is tails) =
1/2 as well. That matches C.
Where 1/3 came from: 1/3 is the probability you grabbed the double-headed coin
before you flipped anything - the plain "which coin" odds. But seeing a head is real evidence: the double-header always shows heads, so a head makes it
more likely you're holding it. That's why its share rises from
1/3 up to
2/4, and you can't reuse the
1/3.
Lock the idea in with a smaller version: one fair coin (H, T) and one double coin (H, H) -
3 heads-faces total. Given a head shows, how many of those
3 heads-faces have tails behind them? Just
1, so the answer is
1/3 - again counted only among the heads-faces, never from the raw "1 of 2 coins."
Same move every time: after the evidence appears, count
inside the faces that could have produced it.
Answer: ConlyPlanA
Though I was unaware of application of Bayes formula in this question, I used the same approach, but got different answer.
I solved it like this:
Total possibilities = 6 (since there are 4 heads and 2 tails)
Given that one heads showed up, probability that flip side has another "heads" = 1/3
So, probability that flip side has tails = 1 - 1/3 = 2/3
What am I missing here?