Try an experiment
Can you spot a random sequence?
Heads. Heads. Heads. Heads. By now, tails feels overdue. But before you make that prediction, pull up a chair—we’ve got a little coin-flip game for you.
Skip to the experimentImagine two friends at a kitchen table. One flips a coin twenty times and writes down the results. The other writes heads, tails, heads, tails, all the way across the page.
Both slide their notes over to you. The names are covered. You have to guess which friend actually flipped the coin.
The tidy row might look reassuring. A long patch of heads in the other row might look like someone got lazy. But a coin has no interest in making its notebook look nicely balanced.
Let’s play that game here. We’ll make one row with a fresh 50/50 choice each time, and make the other switch back and forth. Go with your first instinct.
“Surely tails
is next?”
The coin hasn’t heard it.
Your turn to guess. Choose a row, then we’ll turn the notes over and show you how each one was made.
Which row feels random?
One friend flipped a coin. The other wrote heads, tails, heads, tails. Which row came from the flips?
Make your pick. Then we’ll show you how each row was made.
We keep whatever the flips give us—even if the result looks a bit too tidy.
The coin didn’t get the memo.
Did the streak make you hesitate? It’s easy to see four heads and feel that tails ought to get a turn. But for a fair coin, where each toss is a fresh chance, the next toss is still 50/50.
A coin doesn’t keep a little ledger of what it owes you. The previous flips don’t pull the next one toward tails.
There’s a lovely way to see this. Before you start, the chance of getting exactly HHHH in four tosses is 1 in 16. So is the chance of getting exactly HTHT. One looks much more organized, but neither exact sequence has an advantage.
Look at the little clumps.
After your guess, we outline a longest streak in the coin-flip row. Try another pair and see how it changes. Sometimes the row looks fairly mixed. Sometimes one side settles in for a while.
That doesn’t mean a neat row is impossible. Our coin-flip row could happen to alternate all the way across, just like the other one. If the rows match, we leave them that way. Tidying up the results to make them look random would spoil the game.
This is also why a short row can’t prove how it was made. Here, we know because we made the rows in two different ways; your eyes are only getting the finished notes.
What if I count heads instead of naming the order?
Then you’re asking a different question. “Three heads and one tail” can happen as HHHT, HHTH, HTHH or THHH: four exact sequences, for a total chance of 4 in 16. Each individual sequence still has a chance of 1 in 16.
Back to the kitchen table.
Suppose you get twelve heads and eight tails. There’s no need for the next four tosses to be tails to make things come out even. Twenty flips is a small slice of the story, and it doesn’t have to divide neatly in half.
So when you use a coin flip to make a choice, let each flip be a fresh start. If what you actually need is for everyone to get an equal number of turns, a shuffled list is a better fit.
Your friend’s messy notebook may be the honest one. Even when it looks a little too fond of heads.
Is there a more serious way to check?
Yes. There are tests for particular kinds of patterns, including a runs test described by NIST. This guessing game isn’t one of those tests. It’s a way to notice your expectations, and your guesses stay in your browser.
One more flip?
The coin still hasn’t heard your prediction.
Flip a coin