In one sentence
For a problem about a positive integer , the answer is often not given in the question. Then:
- compute: work out the cases honestly;
- guess: look at the data and guess the general formula, by analogy and by trial;
- prove: prove the guess by mathematical induction.
The first two steps are discovery; the third is proof. A guess you cannot prove is only a conjecture.
Example: a₁ = 1/2, aₙ₊₁ = 1/(2 − aₙ). Find the general term
Step 1 / 6- Compute the first term
. With one term you cannot see anything.
Predict first
Join points on a circle by all possible chords (no three chords through one point). For the disc is cut into regions. How many regions for ?
Not but . A way to count regions: each chord you draw is cut by the chords already there into some number of pieces, and each piece adds one region. The final total is (each crossing corresponds to points on the circle). For : .
Stop 1 · Recursive sequences: guess the general term
A recurrence tells you “how the next term comes from this one”, which is exactly what the inductive step needs. So for recursive sequences, “guess, then prove” almost always works: once the guess is right, the proof is just substituting it into the recurrence.
Stop 2 · Sums: guess the degree, then the coefficients
. Each added term is quadratic in , so guess that is a cubic polynomial — just as (adding linear terms) is quadratic.
One problem, several solutions
Find .
Guess + induction
; times that is , which is exactly . Guess .
Inductive step: .
CostYou need a guess first; with a guess, the proof is one line of algebra.
Telescoping
. Add from to : everything in the middle cancels, leaving .
CostNo guessing, but you must think of this way of splitting the term.
Stop 3 · Products: guess a fraction
, .
Once you have guessed , the inductive step is .
Guessing without proving
Break it
Counterexample lab: remove one condition
Conclusion: It holds for every ← still true?
has a base case and an inductive step: however many terms you test, it never fails.
is backed only by five data points, with no inductive step. Press “One more term”: the sixth term overturns it.
Why can it not be proved? To prove you would have to show that “one more point doubles the number of regions”. But the new chords from the new point add regions, which has nothing to do with . The step you cannot prove is exactly where the guess goes wrong.
Practice
, . Find .
Compute .
, . Find .
Find and prove your answer.