Spaced repetition: the science of remembering

The forgetting curve — why you forget after studying

In 1885, Hermann Ebbinghaus conducted the first systematic study of human memory by memorising lists of nonsense syllables and testing himself at different intervals. He discovered that memory decays in a predictable pattern — fast at first, then slower — now called the forgetting curve.

After one hour, roughly 50% of new information is forgotten. After 24 hours, 70%. After a week, 90% — unless you review it. The insight Ebbinghaus also discovered: reviewing material at the right moment — just before you forget it — dramatically slows the decay and strengthens the memory trace each time.

This is the mechanism behind spaced repetition. It's not a study technique in the sense of a method. It's a scheduling system that answers one question: when is the optimal moment to review this again?

SM-2 — the algorithm most platforms still use

In 1987, Piotr Woźniak developed SuperMemo 2 (SM-2), the first practical spaced repetition algorithm. SM-2 was simple: after each review, rate the item 0–5. The algorithm adjusts the next review interval using a fixed formula based on an "ease factor" that modifies a base interval of roughly 1 day → 6 days → 6 × ease factor days → …

SM-2 was a massive improvement over no scheduling at all, and Anki's default algorithm is built on it. But SM-2 has known systematic errors identified through decades of real-world data:

• It ignores the specific shape of each person's forgetting curve. Everyone forgets at slightly different rates. SM-2 uses the same decay assumption for everyone.
• It over-spaces easy material and under-spaces hard material at scale, leading to wasted reviews and missed retention opportunities.
• It doesn't model retrievability directly. The ease factor is a proxy for stability, not a direct measure of current recall probability.
• It penalises cards harshly after a lapse — a wrong answer resets the interval to 1 day — even when you nearly remembered it.

FSRS — why it outperforms SM-2

FSRS (Free Spaced Repetition Scheduler) was developed by Jarrett Ye and adopted by Anki in late 2023 after large-scale empirical testing showed it significantly outperforms SM-2 on real user data. The key innovation: FSRS models memory with four parameters instead of one ease factor:

FSRS calculates the next optimal review interval by solving for when retrievability will drop to a target (default: 90%) given the item's current stability. After a successful review, stability increases — but by a different amount depending on whether the review was easy (large increase) or difficult (small increase). After a lapse, stability drops based on the item's difficulty and prior stability, and the interval is recalculated from that new lower stability, not reset to 1 day.

In practice, this produces more precisely timed reviews, fewer wasted reviews, and better retention per hour of study time.

Why spaced repetition matters specifically for CFA and FRM

Professional finance exams test a large volume of technical material — formulas, definitions, conceptual distinctions — over a 3–6 month study period. The forgetting problem is acute: material studied in Month 1 is genuinely forgotten by Month 4 unless it's been systematically reviewed.

The alternative most candidates rely on — re-reading notes, watching lectures again — is among the least effective study methods for durable retention. Re-reading creates an illusion of fluency without strengthening the memory trace. The technical term is the fluency illusion: material feels familiar after re-reading, which candidates interpret as learned, but familiarity and retrievability are different cognitive processes.

Retrieval practice — trying to recall information before seeing the answer — is consistently shown in the research to produce 2–3× more durable retention than re-reading (Roediger & Karpicke, 2006). Spaced retrieval practice (practising recall at increasing intervals) produces the best outcomes of any single study method measured in the literature.

The 85% challenge point

Research by Bjork, Kornell, and colleagues identified that learning is most efficient when you're right roughly 85% of the time. If you're consistently getting 95%+, the material is too easy and you're gaining little new retention. If you're getting under 70%, the material is too hard and errors are demoralising without producing proportional learning gains.

Pinnacle targets this 85% challenge point dynamically. FSRS schedules reviews before retrievability drops below 90%, and Pinnacle's adaptive question selection biases sessions toward subtopics where your accuracy is below 80% — keeping you in the productive difficulty zone rather than the frustration zone or the boredom zone.

How Pinnacle implements this for exam prep

Pure FSRS is designed for flashcard review. Pinnacle extends it for professional exam prep in a few important ways:

• One card per topic, not one card per question. FSRS tracks mastery at the topic level across all questions you've answered, building a stable picture of your performance without requiring you to rate individual flashcards.
• Fresh questions at each review. Instead of seeing the same flashcard again, you see a new question on the same concept. This forces actual retrieval rather than pattern recognition.
• Interleaving on top of spaced repetition. Within each session, questions are mixed across related topics — not blocked — which produces additional retention benefits beyond spaced repetition alone.
• Error pattern analysis. Wrong answers are classified by subtopic so the system can distinguish between "weak on duration calculation" and "weak on yield spreads" — and adjust scheduling accordingly.