Fair, unbiased, and memoryless: the gambler's fallacy

What to assume they know

• Experimental probability: relative frequency over many trials — experimental probability, relative frequency, and the Law of Large Numbers
• Theoretical probability: counting favourable outcomes when all are equally likely — theoretical probability, and the equally likely assumption
• What we mean by random — randomness as an unpredictable item from a known list
• Adding fractions with the same denominator, and comparing two decimals to three places
• Multiplying a fraction by itself several times, for the run-probability calculation in section 11

What they should be able to do

• State the gambler's fallacy and identify it in a described piece of reasoning
• Explain why the probability of the next toss is unchanged by any run of earlier results
• Explain how a long-run proportion can settle towards a half without any outcome being owed, using a numerical dilution argument
• State what would have to be true of a coin for the fallacy to be correct, and say what that would make the coin
• Define what makes a coin fair and unbiased, and say what a random toss adds to that
• Give a working definition of independent trials
• Apply the reasoning to a die in a board game, and to a coin experiment already recorded
• Explain what a long run of one result is evidence about, and what it is not evidence about

Where it usually goes wrong

• "Tails is due." Nothing is due. The chapter's box exists for this sentence alone.
• "The Law of Large Numbers guarantees the counts even out." It guarantees the proportion settles. The gap between the head count and the tail count typically grows as tosses pile up; it is the ratio that shrinks. Conflating the two is the fallacy in mathematical dress, and it is the reason this topic follows the Law of Large Numbers rather than preceding it.
• "A long run means the coin remembers." It means you saw a run. On a fair coin six heads happens about once in 64 six-toss attempts.
• "A fair coin cannot give six heads in a row." It can, with probability 1/64.
• "Independent means the two things have nothing to do with each other in subject matter." It means one result does not change the other's probabilities. Two tosses of the same coin are the same coin and are still independent.
• "'Fair' describes the person tossing." Unbiased describes the coin; random toss describes the throwing. The chapter separates them in the same box and the separation is the content.
• "After a long run, switch your prediction." Neither prediction is better. If you have reason to doubt the coin, doubt the coin — do not bet against it on the strength of six tosses.
• "This only matters for gambling." It is the reason a student misreads their own twenty-toss table in Exercise Set 7.2 Q3, and the reason people expect a monsoon to "make up" a dry week.

Questions to check understanding

• Given a described run of results, state the probability of the next trial and justify it
• Identify the flawed step in a piece of gambler's-fallacy reasoning
• Explain why a long-run proportion settles without any outcome being owed
• Define fair and unbiased for a coin, and say what a random toss adds
• State whether two described trials are independent, and why
• Compute the probability of a stated run on a fair coin or die, such as three sixes or six heads
• Reasoning question: after recording twenty coin tosses, a student says the next toss is more likely to be tails because tails is behind. Reply using both the definition of a fair coin and the Law of Large Numbers

Examples worth working on the board

Inputs, not answers, except where the chapter prints the result itself. Values marked Verified are worked out here; the chapter prints no answers and this volume has no appended answer key.

• The GAMBLER'S FALLACY box (p. 164). Its scenario: a fair coin comes up heads six times running, and you feel sure the next toss must be a tail because tails is overdue. Its answer: the coin keeps no record, each toss starts afresh, and the probability of tails on the next toss is still 50%, or 1/2. It names the misunderstanding.
• Section 2's move, which the chapter does not make. Ask what the coin would need in order for the intuition to be right. It would need to know its own history, and it would need to bias the next toss towards tails — that is, to stop being a fair coin. So the intuition does not merely add something to the fair-coin model; it contradicts it. Show the two claims side by side and the contradiction is visible in one beat.
• Section 4's dilution arithmetic, and this is the brief's central addition. Start from six heads in the first six tosses: the running proportion of heads is 6/6 = 1.00. Verified continuations, using the expected count of heads in the remaining tosses:
  • after 100 tosses in total: 6 + 94/2 = 53 heads out of 100 → 0.530
  • after 1000 tosses: 6 + 994/2 = 503 out of 1000 → 0.503
  • after 10 000 tosses: 6 + 9994/2 = 5003 out of 10 000 → 0.5003

Nothing has cancelled the six heads. The excess above half is 3 in every one of those lines. What changed is the denominator. That is the Law of Large Numbers doing its work, and it needs no memory anywhere in the coin.

• Section 5's contrast, which is the same numbers read the other way. For the count to come back to a half by the coin's own doing, the coin would have to produce six extra tails — that is, tails at a rate better than 1/2 for a while. A coin that does that is a biased coin. So the fallacy asks for the one thing that is ruled out by the assumption it starts from. Verified: the surplus of 3 above half is unchanged in all three lines above.
• Example 6 (p. 164). You are playing Snakes and Ladders with a fair 6-sided die and have just rolled three sixes in a row. The thought the chapter puts in the player's head — that a fourth six is now out of the question — is the fallacy; each roll is an independent event and the probability of a six is always 1/6, which the chapter prints as ≈ 0.166, or 16.6%. Its Key Idea adds that each roll of the dice in games of chance is independent and that this kind of randomness has no memory. Verified: 1/6 = 0.16666…, so the correct roundings are 0.167 and 16.7%; see the flag in Notes.
• Think and Reflect, p. 163. Eight fours in succession, so the roll under discussion would be a ninth four — and its probability is unchanged at 1/6; probability speaks about the long run, not about the next trial. The box's closing sentence is the cleanest statement of the whole idea in the chapter and belongs in section 3.
• The FAIR AND UNBIASED box (p. 164). A coin is assumed fair, meaning symmetrical, so there is no reason for it to land more often on one side; the chapter calls that property being unbiased. And a random toss means the coin drops of its own accord, with nothing and nobody nudging it. Sections 8 and 9 should keep the two apart deliberately: the first is a fact about the object, the second a fact about the procedure. A perfect coin placed carefully on the table heads-up satisfies the first and fails the second.
• Where the fallacy is planted in the exercises — Exercise Set 7.2 Q3 (p. 165). Toss a coin 20 times and record every result; then, in part (iv), state the probability of getting tails on one more toss. Verified: 1/2, whatever the twenty results were. The question sits inside the experimental set, immediately after three parts that ask the student to compute frequencies from their own record, which is precisely the set-up that makes the wrong answer tempting. Point this out; it is a deliberate trap and a good one.
• Section 11's calculation, which is the honest reconciliation and is added here. A run of six heads is not evidence about the next toss, but it is weak evidence about the coin. Verified: on a fair coin the chance of six heads in six tosses is (1/2)⁶ = 1/64 ≈ 0.0156 — unusual, and nowhere near decisive; you would see it roughly once in every 64 attempts at six tosses. A hundred heads in a hundred tosses would be another matter entirely. So the correct summary is: the run tells you something about the coin, and nothing about the next toss given the coin. That distinction is what stops the section from sounding like a lecture against noticing patterns.
• The two printed values for 1/6, which section 12 should reconcile. p. 161 prints 1/6 as 0.1666… ≈ 0.167, or 16.7%. p. 164 prints it as ≈ 0.166, or 16.6%. Verified: 0.1666… rounds to 0.167 and 16.7%; p. 164 has truncated rather than rounded. Read on the printed pages. Use 0.167, and if the printed figure appears, say in one line why the two differ.
• A closing demonstration worth the time, not in the book. Show a hundred-toss simulation as a running-proportion line that starts at 1.00 after six heads and slides towards 0.5 without any compensating streak of tails. The line falling while the surplus stays constant is the entire thesis in one movement.

Figures to have open

• A running-proportion chart that begins after six heads and slides towards 0.5, with the surplus of three heads labelled and visibly constant. Not in the book, and it is the single most important image in this topic — the chapter has no figure at all in either box.
• A coin drawn in cross-section with its symmetry axis marked, for section 8. Standard schematic.
• A two-panel comparison of a free toss and a placed coin, for section 9. Not in the book.
• A board-game track with three sixes already rolled, for Example 6. Redraw; do not use the photograph of the museum board on p. 161, which belongs to Experimental probability: relative frequency over many trials.
• A 1/64 tree or grid of the 64 equally likely six-toss sequences, with the all-heads one highlighted, for section 11. Standard schematic; a 8 × 8 grid of dots reads better than a six-deep tree at this scale.

Where this sits in the book

• NCERT Ganita Manjari, Class 9 Mathematics (NCF-SE 2023), Chapter 7, the boxed panel GAMBLER'S FALLACY (p. 164), Example 6 with its Key Idea (p. 164), and the boxed panel FAIR AND UNBIASED (p. 164). All three are unnumbered blocks sitting between §7.2.3 and Exercise Set 7.2; the chapter gives them no section number.
• The Law of Large Numbers paragraph and the Think and Reflect that follows it (p. 163).
• Exercise Set 7.2 Q3 (p. 165), whose part (iv) is this topic's assessment case; the rest of the question belongs to Experimental probability: relative frequency over many trials.
• The word independent recurs in §7.4's definition of a multi-step experiment (p. 168), handled in Tree diagrams make the sample space of a two-step experiment visible.
• The Did you know? box on Snakes and Ladders and Fig. 7.4 (p. 161) sit with Experimental probability: relative frequency over many trials; Example 6 refers back to the game.
• The Chapter Summary (p. 173) defines none of this section's three ideas: there is no gambler's fallacy in it, no statement of what makes a set-up fair, and no account of what independence means. The word fair does surface once, inside the theoretical-probability bullet, but as an assumed background condition rather than as something explained. Checked on the printed page.