Retention is the probability that you'd still remember a card if you reviewed it right now. It's the single most useful number in spaced repetition: it tells you whether your reviews are actually working, not just how many you've done.
flangu uses the word "retention" for three different things. They're related, but they answer different questions. Mixing them up is the easiest way to misread your own statistics.
Three numbers, three questions
| Name | Question it answers | Where it comes from |
|---|---|---|
| Target retention | "How often do I want to remember?" | A setting you choose per deck |
| Standard retention | "Across every review I've ever done, how many did I pass?" | A raw count of all reviews |
| True retention | "How many cards did I actually remember, the first time I was asked again?" | A refined count of reviews |
Target retention is a goal, set before the fact. Standard and true retention are measurements, taken after the fact. This article covers all three, but spends most of its time on the last two, since they're the ones that get confused with each other.
Target retention: the setting
Target retention is the Retention slider on a deck's settings, from 70% to 99%, default 90%. It tells FSRS how confident to be before waiting longer to show you a card again. A lower target means fewer, more widely spaced reviews and a higher chance of forgetting in between. A higher target means more frequent reviews and a lower chance of forgetting.
Target retention doesn't measure anything. It's an input to the scheduler, not an output of your performance. See Deck settings for how to change it.
Everything below is about the other two numbers: what actually happened, measured from your review history.
Standard retention
Standard retention is the share of all your reviews, ever, rated Hard, Good, or Easy rather than Again:
standard retention = (Hard + Good + Easy) / total reviews
Every review you've ever done counts toward this number, including a card you review twice in the same day.
That last part is what makes standard retention a slightly noisy measurement. If you get a card wrong, FSRS shows it to you again soon, often later the same day. When you pass it the second time, that second attempt is still a review, and it still counts toward standard retention as a pass. A single lapse can end up counted twice: once as a fail, once as a same-day recovery. The result leans a little more optimistic than your day-to-day recall actually is.
True retention
True retention answers a narrower, more honest question: of the times you were asked to recall a card on a different day than you last saw it, how many did you get right?
true retention = (Good + Easy, on different-day reviews) / (total different-day reviews)
Two refinements separate this from standard retention:
- Same-day reviews don't count. If you fail a card and see it again later that same day, neither attempt counts toward true retention. Only reviews where at least one day has passed since the last one are included. This is what "true" refers to: it isolates recall that survived an actual gap in time, which is the entire point of spaced repetition.
- Hard doesn't count as a pass. Standard retention treats Hard the same as Good or Easy: any rating above Again is a pass. True retention only counts Good and Easy. Rating a card Hard means you got there, but with real effort. True retention doesn't call that a clean pass.
Both refinements push in the same direction: true retention is always equal to or lower than standard retention for the same set of reviews, and it's the number that best reflects how well you're actually retaining vocabulary long-term.
Note
If you have no different-day reviews yet, such as a brand-new deck, true retention has nothing to measure and won't display a value until at least one review crosses a day boundary.
Which number you'll actually see
Target retention is set once per deck, on the deck's settings.
Both numbers are shown in your statistics, globally, per deck, and per card, behind a Standard/True toggle. True retention is the one worth watching over time; Standard is there mainly for contrast, so you can see how much the same-day-recovery effect is inflating it.
See How statistics work for where each statistics screen lives, and Global statistics for the toggle itself.