Forgetting Curve Calculator: When Will You Forget This?
See your retention decay day by day, and the day it drops below the line you set. This forgetting curve calculator runs the half-life model behind modern spaced-repetition research, so the review date is computed from your inputs, not a rule of thumb.
- Retention decays as p = 2^(-days / half-life), the model published behind Duolingo's spaced-repetition research.
- Each successful review doubles the half-life, which is why review gaps grow over time instead of staying fixed.
- Set your own threshold (50-90%) and the calculator tells you the exact day you cross it.
Your forgetting curve, right now
You're already below your 80% line. Retention is around 63% right now, so review it today.
Retention over the next 45 days
Estimate using the half-life exponential model (p = 2^(-days/half-life), the same shape behind Duolingo's Half-Life Regression). Your real curve depends on how deeply you engaged with the material.
Three starting points, and what each one is for
This forgetting curve calculator needs a starting half-life before it has any review history to work from. These three are reasonable defaults, not a measurement of your actual memory: the calculator moves off them the moment you log a review.
| Preset | Starting half-life | Best for | Not ideal for |
|---|---|---|---|
| Skimmed once | 1 day | A first pass over new material you have not tested yourself on | Anything you need to recall next week, not just today |
| Studied normally | 3 days | One focused session where you tested yourself, not just re-read | Cramming the night before with no active recall |
| Well-rehearsed | 7 days | Material you have already reviewed successfully a few times | The first time you are seeing something |
How this forgetting curve calculator estimates decay
Ebbinghaus ran the first version of this experiment on himself in the 1880s, memorizing meaningless syllables and timing how much he could still recall hours and days later. The curve he found drops fast at first and then flattens, and Murre and Dros's 2015 replication of his method found the same shape in modern subjects. The calculator above uses a specific, testable version of that shape: half-life exponential decay, p = 2^(-days since review / half-life), the same model Duolingo published under the name Half-Life Regression for scheduling language-learning reviews.
Ebbinghaus and the shape of forgetting
What made Ebbinghaus's result useful rather than just interesting is that the drop is not linear. Losing the first 20% of your recall happens within hours; losing the next 20% takes days. A flat "you forget 10% a day" model misses that entirely, which is the main reason this calculator uses half-life instead of a straight percentage drop.
Why half-life instead of a flat percentage
Half-life describes the point where recall has dropped to exactly 50%, and the decay on either side of that point follows the same curve no matter how strong the memory started. That makes it comparable across a skimmed fact with a one-day half-life and a well-rehearsed one with a seven-day half-life: both decay the same shape, just stretched or compressed in time.
What the model does not know
It has no idea how you actually encoded the material, whether you tested yourself or just re-read it, or how it overlaps with things you already know, all of which move the real half-life more than the preset can capture. Treat the output as a reasonable planning estimate, not a measurement, the way you would treat any calculator that has never actually seen you try to recall anything.
When should you actually review, before you forget it?
The honest answer is before recall drops below whatever line matters for what you're using the information for, not after. A review that happens once you've already forgotten is closer to relearning from scratch than to reinforcing something you knew. Catching it while recall is weak but still present is what lets the interval grow afterward, which is the entire mechanism the threshold setting above is built around.
How many times you need to do this depends on what you're aiming for, and running the numbers through a forgetting curve calculator is more useful than guessing. A fact you only need for tomorrow's meeting might survive on a single well-timed review. Something you want to still know in a year is going to need enough passes that the half-life has grown past the gaps in your normal routine, which for most people ends up being somewhere in the range of four to eight reviews, spaced further apart each time rather than on a fixed weekly schedule.
Does reviewing reset the forgetting curve back to where it started? No, and that's the part worth remembering: a review resets your recall to near 100% at that moment, but the half-life underneath it keeps growing with every pass. That is why the sawtooth line in the chart above gets flatter and its peaks get farther apart the longer you keep it up, and it is also why cramming the same fact five times in one sitting does far less than reviewing it five times over five weeks.
Questions people actually ask about forgetting
What is the forgetting curve?
It's the idea, first tested by Hermann Ebbinghaus in the 1880s, that memory does not fade at a steady rate. It drops fast right after you learn something and then slower and slower the longer it has held on. A 2015 replication by Murre and Dros re-ran his method on modern subjects and found the same fast-then-slow shape, so it's not just a 19th-century artifact.
How fast do you actually forget what you just learned?
Faster than most people expect, and it depends entirely on how well you learned it. A quick skim can lose half its recall within a day. Something you studied properly, tested yourself on, tends to hold near 50% recall for several days. The three presets above are rough starting points for that difference, not a universal law.
How many times do you have to review something before you stop forgetting it?
There is no fixed number, but each successful review buys you a longer gap before the next one. That's the whole mechanism behind spaced repetition: review 1 might need to happen tomorrow, review 4 might not be due for a month, because each pass makes the memory more durable. The calculator above doubles the half-life on every review to show that widening gap.
Does reviewing something reset the forgetting curve?
It resets your recall back toward 100% at that moment, yes, but it does not reset the curve to where it started. A card you have reviewed three times decays slower after the reset than a card you are seeing for the first time, because the underlying half-life has grown. That is the entire point of spacing reviews instead of cramming the same information over and over.
Is there a way to calculate my own forgetting curve instead of a generic one?
Not from a single number, no. Your actual half-life for a given fact depends on how it was encoded, how much it overlaps with things you already know, and how you are tested on it, none of which a generic calculator can see. Tools like Anki and Duolingo estimate it per item from your real pass and fail history, which is a better approach than any one-size curve, this one included.
Why do spaced-repetition apps want you to review before you forget, not after?
Because a review after you have already forgotten is really a second first exposure, not a reinforcement. Catching it while recall is still weak but present is what stretches the interval afterward. That's the reasoning behind the threshold setting above: it marks the point where you're about to lose it, not the point where it's already gone.
Related tools
QuizPaste is not affiliated with Duolingo, Anki, AnkiWeb, or SuperMemo. Half-Life Regression is a spaced-repetition model published by Duolingo's research team; SM-2 is the SuperMemo algorithm. We name them only to describe the math this estimate is based on.