How can I reason about a probability puzzle without calculating every case?
Quick answer
Define the complete set of outcomes first, then pair them so each pair contains one result in the event and one outside it. For example, with two independent fair six-sided dice, pair each roll with the same first-die result and a second-die face of opposite parity. Each pair has one odd sum, so the probability of an odd sum is one half because all 36 ordered rolls are equally likely. [S1]
The question, the details, the next step.
I’m working on a probability puzzle and don’t want to list every outcome. What does a useful symmetry shortcut look like?
A useful shortcut groups outcomes in a way that makes the event count predictable. Pairing is one simple example: if every pair contains the same number of outcomes that satisfy the event, you can count pairs rather than tallying each successful outcome separately.
Suppose two independent fair six-sided dice are rolled, with the first die distinguished from the second. What is the probability their sum is odd? There are 36 ordered outcomes: each of the six first-die results can occur with each of the six second-die results. Under these assumptions, all 36 are equally likely, so counting the fraction in the event gives its probability. [S1]
How do I make the pairs without accidentally leaving out some rolls?
Hold the first die fixed and pair the second-die faces as (1, 2), (3, 4), and (5, 6). Do this separately for each first-die result. For example, when the first die is 1, pair (1, 1) with (1, 2); when it is 2, pair (2, 1) with (2, 2).
The six second-die faces are divided into three disjoint pairs, and each possible first-die result gets its own group. So every ordered outcome appears exactly once, producing 18 pairs. In each pair, one second-die face is odd and the other even. With the first die fixed, the two sums therefore have opposite parity: exactly one is odd.
Thus 18 of the 36 equally likely outcomes have an odd sum, and the probability is 18/36 = 1/2. [S1]
Is there a quick way to check that the symmetry argument really works?
Check three things:
- Complete coverage: Every outcome belongs to a pair. Here, six first-die results times three second-die pairs covers all 36 ordered rolls.
- No duplicates: No outcome appears in two pairs. The second-die pairs are disjoint, and each first-die result is handled separately.
- Consistent event count: Every pair has exactly one odd-sum result.
A visual resemblance between outcomes is not enough; the pairing must cover the full outcome space and make the event count clear. The equal-likelihood assumption matters too. This argument does not establish the answer for loaded dice. [S1]
Can you give me a reusable template for applying this to another puzzle?
Write down the event and complete outcome space, then specify a pairing rule that applies to every outcome. Verify that no outcome is omitted or repeated, and check how many outcomes in the event occur in each pair. If that count is constant, use it to find the fraction of outcomes in the event—provided the outcomes are equally likely. If they are not, counting outcomes alone does not give the probability.
For the dice example, the event is “odd sum,” the rule is to hold the first die fixed and pair second-die faces of opposite parity, and the check is that all 36 outcomes occur once with one success per pair. That gives probability 1/2 under the stated assumptions. [S1]
A practical checklist
State the event precisely.
Describe the complete set of possible outcomes.
Explain why the outcomes are equally likely if you plan to count them equally.
Define a pairing rule that covers every outcome exactly once.
Check how many outcomes in the event occur in each pair.
If likelihoods differ or the pairing fails, use a method that accounts for that instead.
A few more useful answers
Does every symmetry argument need pairs?
No. Pairing is one convenient way to group outcomes. The essential check is that the grouping covers the full sample space consistently and makes the event count clear.
Can I use this exact result if one die is loaded?
Not from this argument alone. It relies on equally likely outcomes; if a die is loaded, outcomes may have different probabilities, so counting half the outcomes would not establish a probability of one half. [S1]
Why distinguish the first die from the second?
It makes the outcome space explicit: (1, 2) and (2, 1) are separate ordered rolls. That clarity helps verify that every possible roll is covered once.
Sources
References used for this guide.
- [S1] MIT OpenCourseWare: Introduction to Statistical Methods in Economics, Lecture 2 ocw.mit.edu · Checked