Why the Same Name Keeps Winning: The Real Math of Random Draws
"It picked Marcus again? This thing is rigged." If you run random draws regularly — in a classroom, a stream, a weekly raffle — you will hear this, and you'll probably feel it yourself. Here's the uncomfortable truth: repeats and streaks aren't a sign that randomness is broken. They're what randomness looks like.
By the FairPick team · Published June 11, 2026
The Randomizer Has No Memory
An independent random draw starts from zero every time. When you pick from 30 names, each name has a 1-in-30 chance — and when you pick again, each name has a 1-in-30 chance again, including the one that just won. The generator doesn't know who won last time. It can't, by design: that's exactly the property that makes it fair.
Our intuition fights this hard. The feeling that last week's winner should be less likely to win this week is called the gambler's fallacy, named for roulette players who bet on black because red has come up five times in a row. The wheel doesn't know about the five reds. The coin doesn't remember landing heads. Each independent event resets the odds completely.
What our intuition actually expects from "randomness" is something closer to even distribution — everyone gets a turn, results spread out tidily. But even distribution is the opposite of independence. A draw that guaranteed no repeats would be one where previous results do change future odds, which is precisely the thing a fair independent draw must never do.
The Numbers: Repeats Are Almost Guaranteed
Let's make it concrete. Suppose a teacher cold-calls a random student from a class of 30, once per lesson, ten lessons in a row, with every draw independent. What's the chance that someone gets called twice?
This is the same calculation as the famous birthday problem. The probability that all ten picks land on ten different students is:
(30/30) × (29/30) × (28/30) × … × (21/30) ≈ 0.19
So there's only about a 19% chance of no repeats — and about an 81% chance that at least one student is picked twice within just ten draws. By lesson fifteen, a repeat is a near certainty (over 98%). If your picker has never produced a repeat across many independent draws, that would be the suspicious result.
10 draws, 30 names
≈ 81% chance at least one name repeats. Repeats are the expected outcome, not the exception.
5 draws, 30 names
≈ 29% chance of a repeat — even in five quick picks, it happens almost a third of the time.
Same person twice in a row
1-in-30 after any given pick — about 3.3%. Run a weekly draw for a year and back-to-back wins become likely at some point.
The crowd's verdict
Audiences notice the one repeat, never the dozens of non-repeats. Anecdotes cluster around the surprising outcome.
Why Streaks Feel Rigged: The Clustering Illusion
Humans are pattern-detection machines. For most of history, spotting patterns fast (rustling grass → predator) was worth a high false-positive rate. The cost is that we see meaning in noise — a tendency psychologists call the clustering illusion.
Ask someone to write down a "random" sequence of 50 coin flips and you can usually tell it's fake: people avoid long runs because runs don't feel random. Real flips, meanwhile, produce a run of five or more heads-or-tails somewhere in 50 tosses far more often than not. Genuine randomness is streakier than the randomness we imagine.
There's a famous product-design example: early iPod users complained that shuffle mode kept playing the same artist back-to-back, so it "couldn't be random." It was random — that was the problem. Apple eventually made shuffle less random, deliberately spreading artists apart, so that it would feel more random. When a result needs to be trusted by an audience, perception is part of the design problem, which is why transparency (showing the list, showing the draw) matters as much as the algorithm.
Independent Draws vs No-Repeat Mode: Pick the Right Rule
Neither mode is "more fair" — they're fair at different things. The mistake is not announcing which one you're using before you draw.
Use independent draws when…
Every pick should have identical odds: weekly giveaways open to all, picking who answers each question where everyone should stay alert. Accept that repeats will happen, and say so up front.
Use no-repeat mode when…
The goal is coverage: every student reads once, three different people win three prizes, every team member takes a turn. Remove each winner from the pool until it resets — the name picker has this as a toggle.
Hybrid: backup winners
For giveaways, draw the winner plus backups in one no-repeat session. You get distinct people in a verifiable order, in a single recorded draw.
Announce the rule first
"Repeats are possible — every draw starts fresh" or "winners come out of the pool" takes five seconds to say and removes the whole argument later.
How to Sanity-Check a Randomizer Yourself
You don't need statistics software to test a picker — you need a tally sheet and patience. Enter the names A through E, run 100 picks with repeats allowed, and count. Each name should land near 20 picks, but not exactly: anywhere from roughly 12 to 28 is unremarkable on a sample that small. Run 1,000 picks and the percentages should tighten toward 20% each. Bias looks like a persistent lean that gets relatively stronger as the sample grows; randomness looks like noise that gets relatively smaller.
Two more things worth checking in any tool you rely on: whether it states its randomness source (ours is crypto.getRandomValues() — see how random number generators work for why that matters), and whether the pick happens on your device or on a server you can't inspect. A draw you can fully observe is a draw you don't have to take on faith.
Repeat-winner FAQs.
The same student got picked three lessons in a row. Rigged?
Three in a row from a class of 30 is a 1-in-900 event for any specific pair of consecutive repeats — rare, but across a school year of daily picks and many classrooms, somewhere it will happen. If it bothers the class, switch on no-repeat mode; that's a rule change, not a fairness fix.
Should a giveaway exclude previous winners?
That's a rules decision, not a math one. Excluding past winners spreads prizes and goodwill; allowing them keeps every draw equal. Either is legitimate — write it into the giveaway rules before entries open, not after the draw.
Does picking again after an absent student is chosen break fairness?
No. Discarding an invalid result and redrawing is equivalent to never having had that name in the pool — every remaining name keeps equal odds. This is the same rejection principle generators use internally.
Is a wheel more or less repeat-prone than a list picker?
Identical, if both are independent draws from the same entries. The wheel just makes the result more theatrical. Repeats on a wheel feel more dramatic because everyone watched it happen — the odds don't change.
Try It With Repeats On — Then Off
The name picker has a one-click no-repeat toggle. Run both modes and watch the difference.
🎯 Open Name Picker →Related reading: How Random Number Generators Actually Work · How to Pick a Random Giveaway Winner