Your routine
Cards you see for the first time each day. Every one of them comes back for review, so this is what drives the load.
How likely you want to be to still remember a card at the moment it comes back. Noos uses 90% unless you change it in Settings.
Your average time to read a card, recall it and answer. Only the minutes-per-day figures use it.
The daily load keeps climbing for months after you start, so a year shows what the routine really costs.
Answers per day
204
Average over the last 30 days
Minutes per day
27
At 8 s per answer
Busiest day
224
Answers on the heaviest day
Cards learned
7,300
20 a day for 365 days
Cards remembered
6,943
95% of what you learned
Every card you answer in a day, first sightings included. The load rises for months and then levels off as older cards spread out.
Learned counts every card you have seen. Remembered is how many of them you could recall today, adding up each card's own probability.
The same routine at other retention targets
20 new cards a day for 365 days, at 8 s per answer. Your target is highlighted.
| Retention | Answers / day | Minutes / day | Remembered | Total answers |
|---|---|---|---|---|
| 80% | 133 | 18 | 6,537 (90%) | 36,744 |
| 85% | 161 | 21 | 6,753 (93%) | 44,101 |
| 90% | 204 | 27 | 6,943 (95%) | 55,085 |
| 95% | 308 | 41 | 7,124 (98%) | 80,943 |
How to read the numbers
- Answers per day is every card you answer in a day, the new ones included, averaged over the last month of the run. That is the figure to compare with the time you actually have.
- Cards remembered adds up, for every card you have seen, the probability that you would recall it today. It is the honest measure of what you know: a card at 60% counts as 0.6, not as 1.
- The table reruns your exact routine at four retention targets. Reading down the "Answers / day" column shows what each extra percent of recall costs.
What FSRS and the retention target are
FSRS (Free Spaced Repetition Scheduler) is the algorithm that decides when a flashcard comes back. For each card it keeps a stability, the number of days after which your chance of recalling it drops to 90%, and a difficulty. Every answer updates both: a correct answer after a long gap raises stability a lot, a lapse resets it.
The retention target is the one number you choose. It is the recall probability at which a card is due again. At 90%, a card comes back when FSRS predicts you have a 90% chance of still knowing it; at 95%, it comes back sooner, so you forget less and review more. The relationship is not linear: the last few percent are by far the most expensive, which is exactly what the table shows.
In Noos the target lives in Settings, under Review, and defaults to 90%. This page is the way to see what moving it would do before you move it.
How the simulation works
The simulator runs the same FSRS-6 scheduler as the app, with default parameters, whole-day intervals, no intra-day learning steps and no interval fuzz. Each day it introduces the new cards, answers every card that is due, and lets FSRS schedule the next review.
Whether you get a due card right is drawn at random from the card's own retrievability, the probability FSRS assigns to it that day, so a target of 90% really does mean about one lapse in ten. A recalled card is graded Hard, Good or Easy with probabilities of 22%, 63% and 15%; a card seen for the first time is graded Again, Hard, Good or Easy with probabilities of 24%, 9%, 50% and 17%. These are the defaults of the Anki FSRS simulator, fitted on real review histories.
The random draws use a fixed seed, so the same inputs always give the same chart. The simulation assumes you study every day and has no daily review cap; a missed day in real life carries its reviews over to the next.
Questions
What is a good retention target?
90% is the default in Noos and in Anki, and the simulator shows why: it keeps most of what you learn for a moderate daily load. Going to 95% roughly doubles the load for a few more percent remembered; 80% to 85% is a reasonable choice if you have a lot of material and little time.
Why does the review load keep growing?
Every new card comes back, and the interval between reviews grows more slowly than the number of cards you have seen. The load climbs quickly in the first months and then flattens as old, well-known cards spread out to intervals of months. Stop adding new cards and it falls fast.
Does the simulator use my real parameters?
It uses the default FSRS-6 parameters, which are what Noos uses until it has enough of your reviews to fit your own. Fitted parameters usually predict a slightly lighter load for the same target, because the defaults are cautious.
Does this apply to Anki as well?
Yes. The scheduling is FSRS-6 with no learning steps, which is close to Anki with FSRS enabled; Anki adds short intra-day learning steps, so its first day per card is a little busier. The answer probabilities are the defaults of the Anki FSRS simulator.
Let the algorithm do the scheduling
Noos schedules every card with FSRS, fits the parameters to your own answers once it has enough of them, and lets you set the retention target you just tried out. Free official stacks for every language pair are ready to add.