Skip to main content

How it works

Adaptive spaced repetition

OphthoMentor doesn't just serve questions — it schedules them. Every question becomes a memory card timed to resurface right before you'd forget it. Follow the five steps below and you'll see the whole idea, start to finish.

In one line: we model how strong each memory is, watch how you answer, and show every question again at the last useful moment — spending your minutes on your weakest topics.

1

You forget — and it's predictable

Memory fades along a known forgetting curve. Cram once and most of it is gone in days. But review an item just as it starts to fade and the curve resets a little flatter each time — so the same fact needs fewer and fewer reviews to stick. That's the spacing effect, one of the most replicated findings in learning science.

RecallTime →

Dashed: recall without review. Solid: each spaced review lifts and flattens the curve.

The catch: to hit that perfect moment, we have to know how strong each memory is. So we model it.

2

We model your memory with three numbers

Instead of a fixed ladder of intervals, every card tracks three live values. The middle one — Stability — is the one that matters most.

Difficulty (1–10)

How hard this item is for you. Harder items grow more slowly and come back sooner.

Stability (days)

The key number: how many days your memory lasts before recall fades. Every good answer makes it grow, so the gaps get longer.

Retrievability (0–100%)

Your chance of recalling it right now. It drops as time passes — we review you just before it gets too low.

Retrievability falls smoothly from your last review — R(t) = (1 + 0.2346 · t / S)-0.5 — where t is days elapsed and S is stability. Bigger stability, slower fade, longer gap. This is the DSR model behind the open-source FSRS scheduler (the same modern engine used by up-to-date Anki).

3

Every answer updates the model

Answering isn't just a score — it re-estimates your stability and difficulty, which resets the next due date. Recall it well and the gap stretches; miss it and it shrinks so you see it again soon.

The review loop
You answer a boardquestionRecall is graded: Again/ Hard / Good / EasyMemory model updatesstability & difficultyNext interval set foryour target retentionCard resurfaces on itsdue date
4

Your confidence does the grading

In most apps you rate each card yourself. Here you don't: we read three signals — right or wrong, how confident you said you were, and how fast you answered — and turn them into the grade. A lucky guess barely raises stability and returns soon; a fast, confident, correct answer means the memory is strong, so stability jumps and the card waits far longer. (In Review you can still grade by hand with Again / Hard / Good / Easy.)

How Practice grades a question
You answer an MCQAnswer correct?AgainHow sure were you?HardGoodEasyNoYesGuessedFairly sureConfident & fast
5

It aims where it matters most

Two dials make the schedule yours:

Your target retention

Pick how well you want to remember — 70–97% (default 90%). Higher means more frequent reviews and firmer recall; lower means fewer reviews and more new material.

Weakest-subspecialty weighting

The queue leans toward the subspecialties you score lowest in, so scarce study minutes go where they move your board readiness the most — not just to whatever is due.

Put together: a proven memory model (steps 1–2), fed by how confidently you answer (steps 3–4), pointed at your weak spots (step 5). That combination — not the algorithm alone — is what makes it adaptive.

This is one half of the system. The other is how we measure your exam readiness — turning your answers into a single, honest score. And it all sits on top of clinician-written, reviewed questions. See it all on the how-it-works overview.

The science behind it

The method rests on decades of cognitive-science research and a modern, open scheduling algorithm: