Practise the thinking. Keep the judgement.
What is FSRS? The flashcard scheduler explained in plain English
How FSRS estimates when you will forget a card, what desired retention changes, and the settings that matter more than the algorithm.
What FSRS does
FSRS, the Free Spaced Repetition Scheduler, estimates how likely you are to remember each card on a given day. It tracks how stable your memory of the card is and how difficult the card is for you, using your actual review history. It then schedules the next review for roughly the day your chance of recalling it falls to a target.
The older SM-2 style method grows intervals by a multiplier based on your buttons. FSRS can adapt better to cards you review late, early or after a long break, because it models forgetting rather than applying a fixed growth rule.
Desired retention is the main setting
Desired retention is the proportion of due cards you aim to remember. Anki’s default is 90%. Raising it schedules reviews sooner, so you remember more but review more. The Anki manual notes that the workload rises quickly above 90% and can become overwhelming above about 97%.
Lowering retention spaces reviews further apart, so you study less but forget more. Pick a value that matches the stakes: a high-stakes exam deck might justify a slightly higher target than a hobby deck.
FSRS or SM-2?
Anki has supported FSRS since version 23.10, including AnkiMobile and recent AnkiDroid versions. For most new learners it is worth trying. If you have a long SM-2 history and are content with your workload, switching is optional rather than urgent.
Ronri offers both FSRS and SM-2 with a default desired retention of 90%. Whichever you pick, the habits that matter most stay the same: recall before revealing, grade honestly and keep your new-card limit realistic.
Try it yourself
You raise desired retention from 90% to 97%. What happens to your daily reviews, and why?
Show the answer and explanation
They increase noticeably, because cards are scheduled sooner to keep the chance of forgetting very low.
Each extra point of retention near the top costs more reviews than the one before, which is why the manual warns about workload above 90%.