Why doesn’t a random sequence have to look evenly mixed?

Not necessarily. Random does not mean “neatly mixed,” especially when you look at only a few outcomes. A run is a streak of identical results; runs are a feature examined in fair-coin simulations and in examples of Bernoulli trials. [S1][S2]

Your sequence has a run of three heads, then a run of two tails, then single heads and tails. That describes what happened; it does not, by itself, show that the coin or flipping process is faulty. If you are wondering whether the sequence supports a conclusion about the coin, keep that question separate from simply describing the streak. Record what you actually observe rather than editing the sequence to make it look more mixed.

Assume these are invented outcomes from eight flips of a coin, with H for heads and T for tails:

1. Clumpy-looking: HHHHTTTT
2. More mixed-looking: HTHTHTHT

Both have four Hs and four Ts, but they arrange those outcomes differently. The first has two long runs; the second alternates. Under the stated fair-coin model, runs and clusters can appear in simulated sequences. [S1][S2] The contrast is about appearance, not proof that one sequence is random and the other is not. These examples illustrate a point; they are not reported experimental results.

Do not treat a short run as a prediction. The examples and sources here show that streaks can occur in modeled sequences; they do not establish what a particular coin will do next, or whether any real coin is fair. [S1][S2]

For a fair-coin model, state the assumption explicitly: the tutorial’s example uses Bernoulli(0.5) trials. [S2] The observation that you have seen several heads does not support the claim that tails must come next. Keep the observation separate from the forecast: “I saw four heads in a row” is an observation; “tails must come next” is an unsupported prediction.

You could say: “You noticed a streak, and that’s real. Random results don’t have to take turns. Let’s write down what happened and see what we can say without guessing the next flip.”

Then use a small record:

Flip: 1 2 3 4 5 6 7 8 Result: H H H T T H T T

Ask the child to circle runs and count the heads and tails. This separates two useful questions: “What did we observe?” and “What does that observation let us conclude?” The record is a learning aid, not a method for predicting future results. The cited materials demonstrate runs in fair-coin or Bernoulli simulations; they do not show that every sequence-producing process behaves that way. [S1][S2]