Stepora Blog · Anki

FSRS Anki Settings for USMLE: Should You Turn It On?

Published 30 October 2026 · Stepora

Somewhere in Anki's deck options, next to the settings you actually understand, there is a toggle for FSRS — an optional scheduling algorithm that the USMLE community argues about endlessly. Half your study group swears their reviews got smarter overnight; the other half says they saw nothing, or that it scrambled their intervals a month before the exam. This guide settles what the toggle actually does, what the famous 0.90 number means, and — most importantly — who should touch it and who should walk away.

What FSRS actually changes

Anki's default scheduler works from fixed steps: you learn a card on set intervals, graduate it after set steps, and the intervals grow by fixed multipliers that are the same for every card and every student. It is simple, predictable, and it has carried thousands of students through Step exams exactly as it is.

FSRS — the Free Spaced Repetition Scheduler — replaces those fixed steps with intervals fitted to your own review history. Every time you press Again, Hard, Good or Easy, FSRS updates its model of how quickly you forget that kind of material and sets the next interval accordingly. In practice the setup is three moves: enable FSRS in your deck options, run the optimizer on your review history so the parameters fit your data rather than defaults, and let future intervals personalize themselves.

That is the whole conceptual change: one-size-fits-all intervals become your-intervals. Whether that buys you anything depends on the target you set — which brings us to the number everyone quotes.

The 0.90 retention target, explained

FSRS has one headline knob: desired retention, and the USMLE community's usual recommendation is 0.90. That number means FSRS aims for roughly a 90% chance that you will recall a mature card when it comes due. It is a dial, not a destiny: set it higher and FSRS shortens intervals — you forget less, but your daily review count rises. Set it lower and reviews shrink while forgetting grows.

Now connect that to the rule that actually governs your Anki life: the review cap. In dedicated, reviews are capped at roughly 30–60 minutes a morning so Anki never crowds out question blocks. The 0.90 target fits most students' budgets inside that cap. A student studying twelve-hour days might push the target up a little; a working student or someone on rotations might ease it down. The target serves the cap, not the other way around — if your reviews overflow the cap, the scheduler settings are wrong for your life, whichever scheduler you use.

Why it is debated

The debate has a visible symptom: FSRS remains opt-in. Major decks — AnKing v12 ships FSRS as opt-in — offer it without making it the default, which itself tells you it is not yet the unquestioned standard.

And here is the evidence gap that matters most. What is actually verified about Anki and scores is volume: a Step 2 CK cohort study (Deng et al.) found roughly +1 point per 1,700 additional Anki cards completed. Nothing in the verified corpus ties a scheduler choice to a score. FSRS promises efficiency — the same retention with fewer reviews, or better retention with the same reviews — not points. So the debate is really about efficiency and feel, and efficiency claims are nearly impossible to verify on your own: you cannot A/B test two schedulers on the same brain in the same month.

There is one more honest caveat. FSRS learns from your ratings, so it is only as good as your grading. Students who press Easy on cards they half-knew, or who cram a deck in bursts, feed the model misleading data and get misleading intervals back. The scheduler cannot fix dishonest reviews — it amplifies them into your future schedule.

Who should turn it on

FSRS earns its keep for students who look like this:

• Months of review history on the same deck. The optimizer needs real data to fit. A fresh profile, or a deck you downloaded last week, gives it nothing to learn from.
• A mature deck and a long runway. Preclinical students and early pre-dedicated students have months for personalized intervals to pay off. Interval reshuffling compounds over time.
• Honest grading. You press Again when you did not know it. The model's output quality is a mirror of your input honesty.
• Comfort with tweaking. You are fine watching the first week, adjusting the retention target, and reverting if it misbehaves. FSRS is a tool you manage, not a button you press once.

Who should leave the default scheduler alone

And these students should not touch it:

• You are in dedicated or within weeks of the exam. Switching reshuffles your intervals — cards you expected next week may move, and your daily due count changes unpredictably right when your schedule is least flexible. Never change the machinery during the most volatile phase of prep.
• Your current setup is working. Reviews fit inside your 30–60-minute cap, new cards sit at 20–50 a day, and your question blocks are protected. A scheduler switch solves no problem you actually have.
• Your history is thin or messy. A shared deck downloaded recently, or months of cram bursts before tests, gives the optimizer bad data — and fitted-to-bad-data intervals are worse than the default's honest fixed steps.
• You do not want another settings project. Optimizing parameters and tuning retention targets is real cognitive overhead. If that overhead comes out of question time, it costs more than it could ever save.

If you switch: the safe setup

1. Sync and back up first. Export a backup before you change anything, so reverting is one click.
2. Enable FSRS, then run the optimizer on your full review history. Do not skip this step — default parameters are guesses; fitted parameters are the point of the whole exercise.
3. Set desired retention to 0.90. Adjust it only against your review cap: if reviews overflow your 30–60-minute budget, lower the target or reconsider the switch.
4. Keep your phase's new-card numbers. FSRS changes intervals, not the new-card guidance: 50–80 a day while learning systems, 20–50 in dedicated, new cards only when yesterday's reviews fit the cap.
5. Watch the first week. If due counts explode, you likely optimized on thin or messy history — lower the target, or revert to the default scheduler and try again with more data.
6. Grade honestly from now on. The algorithm learns from every press. An honest Again today buys you accurate intervals for months.

What matters more than the scheduler

Here is the unglamorous truth the toggle wars obscure: the students who finish both their deck and their question bank do not win on scheduler choice. They win on the fundamentals — doing reviews every day, pressing Again honestly, holding the review time cap, and unsuspending new cards from the system they just studied in their question bank. The Deng et al. finding rewards cards completed over months, and no algorithm completes cards for you. Showing up does.

So treat FSRS as what it is: a fine-tuning discussion for students whose fundamentals are already solid. If your reviews fit the cap, your question blocks are protected, and you have months of honest history — try it. If any of that is not true, fix the fundamentals first; the scheduler debate will still be there when you are ready.

However you schedule it, Anki belongs inside a day that protects your question blocks first — the phase-by-phase Anki schedule shows the numbers, the Step 1 study schedule shows where the block sits, and the schedule maker builds the whole day for you, review cap included.