Some exercises rate themselves; others are checked against the entry. Calibration compares the two: how well you think you know an item versus how well you actually do.
The two signals
| Signal | Exercises | How it is read |
|---|---|---|
| Self-rated | Free recall: you tap Miss Slow Steady Fast after seeing the answer | Taken at face value |
| Checked | Multiple choice and typing, compared against the entry | Right or wrong, measured |
What gets compared
Only like against like. A gap means something only inside a single skill. Recognising a word and producing it from memory are different abilities. Within a skill, items are grouped by progress:
When a gap is read
A group is read only once it clears every bar below. Under that, the answer is "not enough to tell" and nothing is said.
- At least 5 self-rated observations
- At least 5 checked observations
- The gap holds across several items
- The gap holds across more than one session
The nudge
It arrives once, at the end of a session, with your real numbers.
Your Vocabulary self-ratings averaged 3.2 but quiz accuracy was 68% on Learning and Familiar items. Adding more quiz exercises to sharpen the signal.
You're rating yourself harder than your quiz scores suggest: 89% quiz accuracy on Proficient and Mastered items. You may know these better than you think.
What changes
Rating yourself higher than the checks show
Sessions pick checked exercises more often for that group, so the items get measured. Your review schedule then follows the measured result.
Rating yourself lower than the checks show
Nothing about exercise selection changes. You get the note and nothing else.