几何分布解题疑问：何时运用教材方法求解首正面概率问题

Hey there! Let's unpack this problem and figure out when to use the textbook method vs. your own approach—this is a common confusion with geometric distribution variants, so you’re in good company.

First, Clarify the Problem's Exact Meaning #

The key here is pinning down what "the first person to flip heads does so on their nth toss" actually means—there are two distinct interpretations that lead to the two different methods:

• Interpretation 1 (Textbook's Likely Angle): We’re looking for the probability that no one gets heads in the first n-1 rounds (each "round" is all three flipping once), and at least one person gets heads on the nth round.
• Interpretation 2 (Your Approach): We’re looking for the probability that one specific person gets heads for the first time on their nth toss, and the other two never got heads in their first n tosses—then sum this probability across all three people.

When to Use the Textbook Method #

The textbook’s concise approach relies on framing the problem as a single geometric distribution for a "combined trial". It works when:

• You can group repeated independent attempts into a single "success/failure" trial. In this case, a "failure" is "all three flip tails in one round" (probability $q = (1-p_1)(1-p_2)(1-p_3)$), and a "success" is "at least one flips heads in a round" (probability $1-q$).
• The event you care about is the first success occurs on the nth combined trial. For this scenario, the probability follows the standard geometric distribution formula:

  P(\text{first success on nth round}) = q^{n-1} \times (1-q) = [(1-p_1)(1-p_2)(1-p_3)]^{n-1} \times [1 - (1-p_1)(1-p_2)(1-p_3)]
  

• This method is ideal when you don’t care who gets the first heads—only when the first heads occurs (at the round level).

When to Use Your Custom Approach #

Your method is necessary when the problem asks for something more specific:

• You need to track which individual is the first to succeed, and that their first success is on their nth attempt (while others haven’t succeeded at all in their first n attempts).
• For this, you calculate the probability for each person separately and sum them up:

  \begin{align*}
  P &= P(\text{Bill gets first heads on nth toss, Mary/Tom got no heads in n tosses}) \\
  &+ P(\text{Mary gets first heads on nth toss, Bill/Tom got no heads in n tosses}) \\
  &+ P(\text{Tom gets first heads on nth toss, Bill/Mary got no heads in n tosses}) \\
  &= (1-p_1)^{n-1}p_1(1-p_2)^n(1-p_3)^n + (1-p_2)^{n-1}p_2(1-p_1)^n(1-p_3)^n + (1-p_3)^{n-1}p_3(1-p_1)^n(1-p_2)^n
  \end{align*}
  

• Use this when the problem explicitly focuses on the individual’s toss count for their first success, not just the round when the first heads appears.

How to Tell Which Method to Use #

Ask yourself these two quick questions to decide:

1. Does the problem care about who gets the first heads, or only when the first heads occurs (regardless of who)?
   • If only "when" → textbook method.
   • If "who" + their exact toss count → your approach.
2. Can you model the problem as a single geometric trial (combining all individuals’ attempts into one success/failure event)?
   • If yes → textbook method simplifies things drastically.
   • If no (because you need to track individuals separately) → use your custom breakdown.

Quick Note: Are the Two Methods Equivalent? #

Surprisingly, no—they answer different questions! The textbook method includes cases where multiple people get heads on the nth round (e.g., Bill and Mary both get heads on round n, which counts as "first success on round n"), while your method only counts cases where exactly one person gets heads on their nth toss and the others got nothing. That’s why the formulas look different—they’re solving distinct (but easily confused) probability questions.

内容的提问来源于stack exchange，提问作者A_for_ Abacus