Professors tell you what's on the exam — just not in the syllabus. They tell you in the verbs they use, the slides they linger on, the diagrams they draw twice, and the asides they drop mid-lecture. Most students transcribe all of it and use none of it.

Predicting exam questions isn't guessing. It's reading the signals your professor already broadcast and converting them into the operations they'll actually test. The slides are the raw material; the prediction is knowing what kind of question each one wants to become.

The core move: extract the operation the professor will ask for — define, compare, justify, calculate, apply — not the content they'll ask about. Content is what's on the slide; the operation is what the exam does with it. A bold term wants a definition plus a counterexample. A repeated diagram wants you to redraw it from memory. A spoken aside wants a justification. Write the stem that forces that operation, in a context slightly different from the slide's, and you've built a predicted exam question that actually predicts.

Where the signals hide

Every slide carries a signal about the question it wants to become — if you know what to look for:

The pattern: the professor's emphasis is the exam's outline. A term they bothered to bold is a definitional question. A diagram they drew on the board instead of leaving in the deck is a reproduction question. An aside — "this will matter later," "students always get this wrong" — is a justification or error-analysis question. And any worked example is a variant waiting to happen: change one input and you have a new problem that tests whether you understood the method or memorized the arithmetic.

The slide-to-stem workflow

Harvest the verbs first. The syllabus and past quizzes tell you the operations this professor tests — "explain," "compare," "apply," "justify." That's the exam's fingerprint. Before writing a single question, know which verbs you're aiming at.

Mark the emphasis. Go through the deck and flag what got extra attention: repeated diagrams, long dwell times, spoken asides, anything the professor circled back to. Emphasis is the signal; unremarked slides are usually background.

Write six stems per lecture. Across the operations — a couple of recall items, a couple of "why" or "justify" items, a couple of apply-or-transfer items. Six is enough to cover the lecture's signals without turning prediction into transcription.

Hide the slide and score. Answer your own stems with the deck closed, then check against a one-line rubric ("must include X, Y"). If you can't answer it, that's a prediction that found a real gap — better now than on the exam.

Variant every miss. A miss isn't done when you read the answer — rewrite the stem with a new number, new setting, or new angle, and answer it again. The variant is what turns "I saw the answer" into "I can produce it."

What prediction looks like

The weak answer recites the distinction — it knows the slide. The strong answer uses it — it applies the difference to a case the slide never mentioned. That's the operation the exam will test, and the weak answer isn't practicing it.

What this isn't

Prediction isn't transcription. Copying "Smoot-Hawley Tariff (1930)" into a flashcard doesn't predict a question — it predicts the slide. The exam won't ask you to recite the title; it'll ask you to deploy the concept in a context you haven't seen. The stem has to force the operation, not just name the topic.

It also isn't mind-reading. You're not guessing which specific question appears — you're matching the operations the professor tests. You might not predict the exact scenario, but you'll have practiced the exact move, and the move is what transfers.