This isn’t much of a problem for SM-2, but it is for FSRS. If someone never uses Again for some reason, this will mess FSRS up.
I think if a user’s retention is above 99.5% for the last month, a warning should be displayed, asking the user, “Are you sure that you are using the answer buttons properly?” and a text that is a shorter version of this.
This should reduce the chances that someone is misunderstanding how answer buttons work.
2 Likes
IMO more than 90% looks good to me. If I remember correctly the average learning workload increases by about 2 times from 80% to 95% retention, and by 3 to 4 times for 95% and above. So I think for beginner Anki users 80% to 90% is efficient and low learning workload, more than 90% is for advanced Anki users.
1 Like
I’m not sure what you mean. I’m talking about abnormally high retention, like 99% or even 99.5% over a month or longer. I’m talking about the kind of retention that makes me suspect that the user is just mindlessly mashing Good.
3 Likes
Yup, I just mean which number is considered abnormally high.
I guess there are some cases where retention would not be 99% to 100% if users were pressing Good (or Hard) instead of Again. e.g. only half of the cards press Good instead of Again, or some decks are reviewed properly but others are incorrectly pressed Good only. If a beginner with 70% to 80% actual retention incorrectly presses Good on the cards the retention could be 90% to 95%.
1 Like
Yeah, choosing the exact cutoff is tricky
3 Likes
Is there a way to find a good cutoff value, e.g. with simulations, some maths magic or similar?
I’m not sure. We can’t infer whether the user actually has insane memory or is using buttons in some weird way based on the review history alone.
1 Like