"Year Retention" and "Month Retention" statistics
How is one supposed to interpret the "Year Retention" and "Month Retention" statistics for a note?
Comments are currently closed for this discussion. You can start a new one.
Keyboard shortcuts
Generic
| ? | Show this help |
|---|---|
| ESC | Blurs the current field |
Comment Form
| r | Focus the comment reply box |
|---|---|
| ^ + ↩ | Submit the comment |
You can use Command ⌘ instead of Control ^ on Mac
Support Staff 1 Posted by drewmccormack on 17 Jan, 2018 08:08 AM
The year retention is the likelihood (probability) that you remember a note in the stack at the end of a year (on average).
The month retention should be higher, and is the likelihood you will remember a note after a month, on average.
Drew
drewmccormack closed this discussion on 17 Jan, 2018 08:08 AM.
zeppo re-opened this discussion on 18 Jan, 2018 06:54 PM
2 Posted by zeppo on 18 Jan, 2018 06:54 PM
Yes, I get that, but how am I supposed to interpret these representations of green and yellow dots?
Support Staff 3 Posted by drewmccormack on 18 Jan, 2018 07:17 PM
I think there are ten dots, so each one is 10%.
4 Posted by zeppo on 18 Jan, 2018 10:59 PM
What's with the yellow dots verses green dots? And then I have some cards that have been studied many times with no dots at all.
Support Staff 5 Posted by drewmccormack on 19 Jan, 2018 07:06 AM
The color gives some indication of how well you know something. Will be red if you know if less than 40% or something like that. Then it becomes yellow, and if you know it well, dots are green. Just an extra indication. The number is what really counts.
That some have no color is suspicious. Your retention decreases with time when you haven’t studied something. Could that be it, or are they things you have recently studied?
Drew
6 Posted by zeppo on 21 Jan, 2018 04:39 PM
By no color, I mean they are grey by the way, so from a charting standpoint it would be as if there are no dots. But i find these types of cards are very prevalent if I go into the "Recently Incorrect" grouping. Something is clearly wrong that I'll have to look into correcting when i get the time.
For one thing, I don't know how long you intend the app to consider to be "recent", but I just at a quick glance I see a card that was last studied in December 2016, with "Recent Run" showing as "1 incorrect" still appearing in that stack. That's not very recent. But what is certainly problematic is that even though I got the card wrong, it has not scheduled me to study the card again until July 17, 2019. It would appear that the algorithm is flawed, though of course it could be something else got screwed up.
If you keep getting a card correct until the algorithm determines the next study date should be a year away, and then when that year passes and you get the card wrong, then the algorithm at that point needs to schedule again relatively soon to decide just how far out of your memory the card has drifted in the past year. The most recent incorrect needs to have more weight in scheduling the next date due than the previous cards. As it stands it appears the algorithm doesn't care what your response is after a certain number of consecutive correct responses has occurred. I might as well have answered that I got the card correct back in December 2016 if its going to wait until 2019 to test me again.
Support Staff 7 Posted by drewmccormack on 21 Jan, 2018 05:22 PM
Recently incorrect is simply notes where the most recent study was ‘wrong’. No time limit is in place.
That sounds like a bug that you see, because getting a note wrong should indeed cause it to become due again the next day, regardless of how often it was right before that.
Something you could try is switching the schedule for the stack to none, and save, and then switch it back to long term again. That may ‘fix’ the scheduling bug.
Kind regards,
Drew
8 Posted by zeppo on 22 Jan, 2018 01:02 AM
I tested it with one stack of 44 notes and 6 rescheduled to be immediately due, including the three that had no retention statistic "dots". I studied those and got a couple wrong which it now scheduled for tomorrow. So they would appear to be back on track. Problem is I have 554 showing up in the recently incorrect grouping. This is obviously way too many and glancing through them I see the same scheduling issues, such as one that I most recently got incorrect last November, yet has not been scheduled to become due again until 2025. I'll have to gradually go through all my stacks with this procedure and once I get it done, see if the problem starts occurring again.
I'm wondering if the decks you use for testing have notes with as long a study history as mine, a large number of which go back to Mental Case 1.
Support Staff 9 Posted by drewmccormack on 22 Jan, 2018 08:15 AM
I guess there is one way in which you can have the last study event “incorrect”, and the schedule still way in future: The schedule doesn’t get updated unless you are studying the note when it is due, and in the correct facet order that is due.
If you study a note before it is due, or in a different facet order than is due, it is not rescheduled. It is just treated as an ‘extra’ study, but not part of the scheduled study.
So if you ran through the notes without choosing the “due notes”, it is certainly possible you could mark some wrong, but that that would not influence their schedule.
Drew
10 Posted by zeppo on 22 Jan, 2018 02:17 PM
Good to keep in mind. In my case, I only study due notes, so that's not what is a work here.
I can see now that the cards without any statistic "dots" at all are all the one that had the most recent answer as incorrect. It will be interesting to see what happens to them after I study them today now that have become due after the scheduling reset. I'm also interested in seeing if it will have an effect on my sync times once I can get all these reset (which will take a while since I don't have time to have them all come due at once).
drewmccormack closed this discussion on 23 Jan, 2018 07:12 AM.