The rule in one line
Either an event happens or it does not, and those two possibilities cover everything. So their probabilities add to 1:
P(not A) = 1 − P(A)
Illustration: the chance of rolling a 6 on one die is 1/6, so the chance of not rolling a 6 is 1 − 1/6 = 5/6. The same works for cards: the chance a drawn card is not a heart is 1 − 13/52 = 39/52 = 3/4.
On its own this is almost trivial. Its real value shows up when the direct route is long and the opposite route is short.
Why “at least one” is hard to do directly
“At least one” means one, or two, or three, and so on, up to all of them. Counting each of those cases separately and adding them is slow and easy to get wrong.
But the opposite of “at least one” is just “none”. There is only one way for none to happen, and for independent tries it is a single multiplication. So:
P(at least one) = 1 − P(none)
A common slip is to treat the opposite of “at least one” as “all”. It is not. The opposite of “at least one success” is “zero successes”. Keep that straight and the method works every time.
Worked example: dice
How likely is at least one six in four rolls of a die? This is the same shape as the classic de Mere problem, and a closer look at the history sits elsewhere. Here is the calculation.
- Chance of no six on one roll: 5/6.
- Rolls are independent, so no six in four rolls: (5/6)^4 = 625/1296.
- Chance of at least one six: 1 − 625/1296 = 671/1296, which is about 0.518, or 51.8%.
Check the numbers: 5 × 5 × 5 × 5 = 625 and 6 × 6 × 6 × 6 = 1296, and 1296 − 625 = 671. The result is a little above one half, which fits the intuition that four rolls give a modest edge toward seeing a six, though nothing is guaranteed.
Notice what the direct method would require: the chance of exactly one six, exactly two, exactly three and exactly four, all added. The complement does it in a single line.
Worked example: two dice
What is the chance of at least one six when rolling two dice? No six on either die is 5/6 × 5/6 = 25/36. So at least one six is 1 − 25/36 = 11/36, about 30.6%.
This also shows why adding 1/6 + 1/6 = 1/3 overshoots: it counts the double-six case twice. The complement avoids that mistake because you never add overlapping cases at all.
Worked example: cards without replacement
The rule works when draws are dependent too, as long as you compute “none” correctly. Suppose you draw three cards from a standard 52-card deck and want at least one heart.
- No heart on the first card: 39/52.
- No heart on the second, given the first was not a heart: 38/51.
- No heart on the third: 37/50.
- All three together: (39 × 38 × 37) / (52 × 51 × 50) = 54,834 / 132,600, which is about 0.4135.
- At least one heart: 1 − 0.4135 = 0.5865, or about 58.6%.
Here the fractions change from draw to draw because cards are not returned. That is the difference between dependent events and independent ones. Using 3/4 three times would give (3/4)^3 = 27/64, about 0.422, and so about 57.8% for at least one heart. That is close but not equal to the correct 58.6%, and with larger draws from smaller decks the gap can be much bigger.
Worked example: a loot drop
A game item drops with a 2% chance per kill, with each kill independent. What is the chance of at least one drop in 50 kills? This is an illustration, not a claim about any real game.
- No drop on one kill: 0.98.
- No drop in 50 kills: 0.98^50, which is about 0.364.
- At least one drop: 1 − 0.364 = 0.636, or about 63.6%.
So even after 50 attempts at a 2% chance, there is still roughly a one in three chance of having seen nothing. A 2% drop is not “due” after 50 misses either, since each kill starts fresh. Drop rates are a good place to apply this rule because players naturally ask “how likely is it that I get at least one by now?”.
Checking your answer
A few quick sanity checks help catch mistakes:
- The “none” probability and the “at least one” probability must add to exactly 1.
- More tries should raise “at least one” and lower “none”.
- If the single-try chance is small, “at least one” grows slowly at first, then more steadily, but it never reaches 100% with a finite number of tries unless the single chance is already 1.
That last point guards against a frequent misconception: ten tries at a 10% chance does not make success certain. It gives 1 − 0.9^10, which is about 65%, not 100%.
Where the rule is used beyond games
The same trick appears in reliability and risk questions: the chance that at least one component fails, or at least one match shows up in a group. The well-known birthday problem uses precisely this method, and it is also the engine behind many quick estimates of low-probability events accumulating over many tries.
A caution on interpretation
The complement rule is a calculation tool, not a prediction. Saying there is a 63.6% chance of at least one drop in 50 kills does not mean you will get one. And in gambling, being shown that something “is likely over many tries” does not change the fact that casino games carry a house edge and that losses can accumulate quickly. If gambling stops being fun or starts to cause difficulty, a local support service can help.
Summary of the method
- Define the event you want (“at least one”).
- Define its opposite (“none”).
- Calculate the opposite, multiplying across tries and adjusting fractions if draws are dependent.
- Subtract from 1.
That is the entire technique, and it is one of the most time-saving moves in probability.