The core distinction
Two events are independent when learning that one has happened tells you nothing new about the other. They are dependent when it does. Almost every mistake in basic game probability comes from treating a dependent situation as independent, or the reverse.
A fair die has no memory. Whatever it showed last time, the next roll has the same chances. A deck of cards has a fixed set of contents, so each card you remove changes what is left. These two situations produce different arithmetic, and the right choice depends on the game’s physical or digital mechanism, not on how the numbers feel.
Independent events: dice, coins and spinners
Roll a fair die twice. The chance of a six on the first roll is 1/6. The chance of a six on the second roll is also 1/6, regardless of the first result. The chance of two sixes in a row is 1/6 × 1/6 = 1/36, about 2.8 percent.
The same holds for coin flips, a spinner, or any well-designed random number generator with independent draws. The defining feature is that the mechanism is reset or unaffected between trials. This is the setting where the multiplication rule applies with no adjustments.
Dependent events: drawing without replacement
Now draw two cards from a standard 52-card deck, keeping the first one out of the deck. The chance the first card is an ace is 4/52. If it was an ace, the deck has 51 cards left and 3 aces, so the chance the second is also an ace is 3/51. The chance of two aces is:
4/52 × 3/51 = 12/2,652 = 1/221, about 0.45 percent.
Compare that with the case where you return the first card and reshuffle. Then both draws have a 4/52 = 1/13 chance, and two aces has probability 1/13 × 1/13 = 1/169, about 0.59 percent. The difference between 1/221 and 1/169 is the effect of dependence, and it comes entirely from the removal of the first ace.
Drawing without replacement is dependent in a structured way, and a family of calculations built on it gets named the hypergeometric distribution. When people discuss deck probability, they are almost always dealing with this setting.
A quick test
To decide which case you have, ask: if I knew the result of the first event, would I change my estimate for the second?
- Two dice rolled one after another: no change. Independent.
- Two cards dealt from the same deck: yes, change. Dependent.
- Two cards drawn with the first returned and reshuffled: no change. Independent.
- A random loot drop on two different kills, where the game states each has the same fixed rate: no change. Independent, if the rate really is fixed.
- A loot system where the rate rises after each failure: yes, change. Dependent.
That last pair matters in video games. A pity timer is a deliberate way to introduce dependence between attempts. The history of earlier tries changes the odds on the next one. Not every game that feels streaky works that way, so check whether the rules actually describe such a system.
Large decks are almost independent
The effect of removal shrinks as the pool grows. Draw two cards from a deck of 52 and the second probability moves from 4/52, about 7.69 percent, to 3/51, about 5.88 percent: a visible change. In a pool of 10,000 items with 1,000 targets, removing one target changes the odds on the next from 1,000/10,000 = 10 percent to 999/9,999, about 9.99 percent, which is hardly different at all. Treating the second case as independent is a reasonable approximation, but the same shortcut applied to a small deck, such as a 40-card deck with 4 copies of a card, would be noticeably off.
Mixing the two
Some games combine both. Rolling a die to decide how many cards you draw, then drawing them, involves an independent step followed by a dependent one. You can handle each stage separately and combine the results.
The same applies to sequences. The chance that a die roll succeeds and a later card draw succeeds can be multiplied, because the die does not affect the deck. But within the card draw, successive cards are dependent on each other.
Common mistakes
- Applying the gamblers fallacy to a deck. In a genuinely independent process, past results do not influence the future. In a deck, they do, but in the opposite direction people usually expect: if many low cards have been dealt, the remaining deck is slightly richer in high cards. This is real dependence, and it is tiny with a freshly shuffled large deck, but it is not the same as a die being “due”.
- Multiplying the original odds on dependent draws. Using 4/52 × 4/52 for two aces from one deck overstates the answer.
- Assuming a software system is independent. Some systems are designed to adjust probabilities, so the history can matter. Read the rules rather than assume.
- Confusing independence with mutual exclusivity. Mutually exclusive events cannot both occur, which actually makes them dependent: if one happens, the other is impossible. Rolling a 3 and rolling a 5 on a single die are exclusive, not independent.
Why the distinction matters in practice
Getting the independence wrong changes the answer by amounts that can matter in a decision. The correct treatment lets you judge the real chance of completing a combination in a card game, or recognise that a streak of misses on a fair die says nothing about the next roll. The idea also underlies the question of whether long streaks count as evidence of anything, which is the focus of discussions about the hot hand.
When in doubt, work out one case in both ways and compare. If the numbers differ, the events are dependent, and the more careful calculation is the one to trust.