Independence: Why Each Draw Forgets the Last One

If you understand only one mathematical idea about lotteries, make it independence. Independence is what makes the games fair. It is what makes "systems" impossible. And it is what almost every wrong intuition about lottery play quietly assumes is false.

What you'll learn

What independence means in probability, in one sentence.
Three different ways to test whether a system is actually independent.
Why every "edge" you can imagine for lottery play assumes independence is broken.
How to catch yourself reasoning as if independence were false.
The single sentence that resolves the gambler's fallacy permanently.

The one-sentence definition

Two events A and B are independent if the probability of A is the same whether or not B happened. Symbolically: P(A | B) = P(A). The vertical bar means "given that." If knowing B happened doesn't change the probability of A, they're independent.

For lotteries, this becomes: the probability of any specific outcome on the next draw is the same regardless of any pattern in the previous draws. The next draw doesn't know what the last one was. It doesn't know that digit 7 has appeared 12 times in 30 days. It doesn't know that combination 1-2-3 has never come up. The mechanism doesn't have memory, by design.

Three ways to test for independence

How do we know a state lottery is actually independent? Three lines of evidence:

Mechanism. The drawing apparatus — gravity-pick machines or certified RNGs — has no physical mechanism by which past draws influence future ones. The balls are returned and shuffled. The RNG entropy source is reset and re-seeded. There is no memory in the system because there is nowhere for memory to live.
Statistical tests. Independence has measurable consequences. The runs test, the chi-squared test for serial correlation, the Wald-Wolfowitz test, autocorrelation analyses — all of these check for hidden dependence. State lottery data passes them. Our Is This Normal? tool runs a subset of these tests on live data and shows the results.
Audit logs. State lotteries publish independent audit reports certifying that draws have been verified random. These reports are public, archived, and required by state contract.

You don't have to take any single line of evidence on faith. They corroborate each other.

Why independence makes "systems" impossible

Take any "system" you can imagine. It will fall into one of these categories, and each one quietly assumes independence is broken.

Hot-number systems. "Play digits that have been hot recently." This assumes hot streaks continue — i.e., that recent appearances make future appearances more likely. That's a violation of independence.
Cold-number systems. "Play digits that are due." This assumes that absence makes future appearances more likely — again, a violation.
Pattern systems. "Avoid combinations that have already won." This assumes past winners are less likely than other combinations going forward — a violation.
Sum-range systems. "Play sums in the most-frequent range." This is true that some sums are more likely than others (e.g., a sum of 13 in Pick 3 has more combinations than a sum of 27), but the per-combination probability is uniform. The sum-range frequency is a property of combinations, not predictive of individual draws.
Statistical systems. "Use regression analysis on 30 years of data." The data is independent and identically distributed. Any pattern the regression finds is the pattern of variance you'd expect to see in random data. There is nothing to fit.

Every system reduces, on inspection, to a claim that independence is broken in some specific way. Independence is exactly what's been independently verified, by mechanism and by test, to hold. So every system reduces to a claim that contradicts what's measurable.

How to catch yourself

The gambler's fallacy is sticky because independence is counterintuitive. Even people who know the math intellectually find themselves slipping into intuition that violates it. Here are the linguistic markers that should make you pause:

"Due to hit"
"On a streak"
"Hot lately"
"Cooling off"
"Looks like it's been avoiding..."
"The pattern suggests..."
"Has to come back to even out"
"Smart money would play..."

Each of these phrases assumes that past draws constrain or inform the next draw. They don't. Notice the phrase, then notice the assumption underneath it, then ask: does the mechanism have any way to know about this? It doesn't. The phrase is wrong even if it feels right.

The sentence that resolves the fallacy

Memorize this: The mechanism has no memory.

The gravity-pick machine does not know which balls came out last week. The RNG does not know which numbers it produced last month. There is no register in the system that stores past outcomes and adjusts future probabilities. Build any "system" you want — the system runs in your head, not in the mechanism. The mechanism doesn't change.

Whenever you find yourself reaching for a "due" or "hot" intuition, replay this sentence. It works. It is the single most useful sentence anyone analyzing a fair random process can learn.

Try it yourself

Go to Hot & Cold and look at the digit with the longest current gap. Notice that your brain wants to assign meaning to that gap — "it's about due." Now run the sentence: the mechanism has no memory. The next draw cannot know about this gap. The probability of this digit appearing in the next draw is exactly its baseline, regardless of the gap.

Then look at the digit appearing most frequently. Run the sentence again. Same result. The mechanism has no memory in either direction.

Common pitfalls

"But over time, it has to even out." The Law of Large Numbers does say that long-run frequencies converge to expected values. It does not say that future draws compensate for past ones. The convergence happens because future draws are also random, not because they correct the past. This is the most subtle point in the gambler's fallacy and the one most people miss.
"What about hot dice / hot hand in basketball?" The "hot hand" in basketball is a real but tiny effect, mostly attributable to player adjustments and shot selection — it doesn't apply to mechanical processes. Lottery balls don't get tired. They don't have a hot hand.
"What if the machine is biased?" Then it isn't independent, and audits would catch it. They occasionally have, in pre-modern lottery history. Modern audits make this vanishingly unlikely. If you genuinely suspect a state lottery is rigged, the right action is to report it to the state inspector general, not to play it differently.