The addon is very useful, but i cant see some of my intervals
For example, in the deck, my learning steps are: 30m 4d. The add-on only shows statistics for the 30m and graduation interval
Please advise how can this be fixed?((
Also, please tell me, does the Review Retention column also display the learning intervals, or only the intervals of graduated cards?
Iād guess you have selected a deck or a time range that has no reviews after the 4d learning step. If you are sure it should show some reviews and if you are willing to share your collection, please upload it somewhere so I can have look.
The Review Retention graph shows reviews of graduated cards only.
It seems that Anki has changed the way long learning steps are stored in the review log. I will update the add-on to take those in to account. Please give me some time as I havenāt used Anki for a while and I have to get back up to speed on Anki development.
Iāve released a new version that should improve compatibility with the V3 scheduler considerably. Most of the changes affect how reviews done with the V3 scheduler are shown. Old reviews done with V1 or V2 scheduler are shown as before.
Long learning steps are shown correctly on the table instead of being lumped together with reviews of graduated cards.
The Learning Retention table will include a ā0 sā step that shows reviews of cards with no interval, i.e. new cards and cards that have been reviewed via filtered decks only. Older versions of this add-on counted those as the graduation step, which was totally bogus.
The older schedulers record the length of the first learning step for new cards. If you have used multiple schedulers, statistics of new cards will be split between the new zero-length step and whatever your first learning step has been.
As a small improvement, long steps are shown as days instead of minutes.
Thanks for the update! Do I understand correctly that if earlier I used an intervals 1d 7d, the part of the stat in the 1d column will not refer to a delay of 1 days, but to a delay before viewing a new card?
If your learning steps have been set to ā1d 7dā, the 1d row will include, in addition to reviews after the 1d step, first reviews of cards if V1 or V2 scheduler was enabled when the first reviews were done.
Just to be clear, the table counts reviews done after the corresponding learning step. When a new card is reviewed for the first time, there is no previous step. The V3 scheduler counts the first review towards a ā0 sā step. V1 and V2 schedulers count the review towards the first learning step.
Once the reviews are stored in the review log, they stay as they are. Changing the settings or scheduler has no effect on old statistics.
It looks like the maximum interval is 52 weeks. Is that correct?
I would like to see my retention for cards with even longer intervals. My suggestion is to add a diagram that goes by months and extends for a couple of years (letās say 5, i.e. 60 months) as well a setting āhide_review_monthā.
The maximum interval is 52 weeks (364 days) by default but it can be changed from the add-on settings. You can increase it but at some point the graph starts to look messy because the grouping is hardcoded to 7 days.
Iāll keep your suggestion in mind! It might take a while before I have the time to add new features.
Iām using your add-on for years now: For me itās the most valuable add-on at all, cause it seems to be the only way to optimize learning steps.
I updated to the v3 Schedular on December 12th. Since then, Anki has been rounding up the answer buttons to whole days. The card info also rounds up to whole days. So if I have Learning Step 640m, 960m, Anki makes it 1d, 1d. However, only on the answer buttons and in the card info. In fact, Anki works with the entered values 640m, 960m.
In the Learning Retention, the add-on adopts the new rounded values. Several learning steps are summarized below. Anyone who has been using the add-on for a while will notice that the old learning steps are still displayed in minutes in the table, which makes the whole thing very confusing.
Couldnāt we stick to the minutes in the representation? If Anki is based on the minutes in the background, should the value exist in the program?
Thank you for your feedback! I havenāt used Anki since the v3 scheduler was released, so I was not aware of the issue.
I took a quick look at the Anki scheduler code. Unfortunately it seems that the latest versions of Anki store the rounded values in the review history. That would mean there is no way to get the exact interval of past reviews. The add-on does no rounding, so what you are seeing is just limited by what is stored in the database.
Iām thinking if I should add an option to always round the interval to whole days. It would at least make the table a bit easier to read. What do you think? Iām open to any ideas.
I was hoping that Anki would still make the information available somewhere at the interface. But that doesnāt seem to be the case. Very unfortunate.
Ideally, Anki could be convinced to continue providing the detailed values āāat the interface. Because they are available to Anki. But Anki probably doesnāt care. So Anki probably goes the way of many programs. The quality comes with the high accuracy and variability. And goes with the simplification.
However, caution is required when trying to find a solution. You cannot link to a specific date because that depends on the individual switchover time to the v3 Schedule. You also cannot recalculate all intervals before the change using the new method. You are missing the learning steps for this. And that might be too complicated. You also canāt round later because the new values āāarenāt actually rounded. Rather, the day of the review is determined based on the learning interval and then simply the interval between the days is specified.
Or you can just round and ignore this fact. One idea might be an individual date entry. The date of switching to the v3 Schedule could be set individually there. As a result, there could be a choice between four display variants: everything before, everything after, everything together unrounded, everything together rounded. Maybe to complicated. So may youāre right. Offering the option to round could be a good compromise. What do you think?
Your add-on is still very useful. I donāt see anything comparable. It just lost its perfection. Anki shot it to pieces for you and us.
I think rounding the dates is the way to go. It is nonoptimal in a way, but it is easy to understand and consistent with the way Anki shows the intervals.
Iāve released a new version that adds an option to round down long learning steps to whole days. It is enabled by default but can be disabled from the add-on settings.
@jambamboleo The latest version also adds hide_review_month option that does exactly what you wished for.
If I open old stats on my main deck, there is nothing between Ease Factor and Answer Buttons.
If I open old stats on a subdeck, there are Hourly Breakdown, Learning Retention and Review Retention between those (and Answer Buttons are all 0), and it works on the main deck until Anki restart. FSRS stats and Review Heatmap appear as usual.
Thanks for the report! I think the issue you are having is caused by a bug in Anki that affects the way all of the old statistics behave.
When Anki starts, it selects a deck called āDefaultā. Unless there are some cards in the default deck, it is completely hidden from the user.
When you open the statistics window, Anki uses the selected default deck for some of the graphs. Because the empty deck has no reviews, some of the graphs are hidden. You have to select another deck by clicking it before opening the stats window to see anything useful. That is also the reason why Answer Buttons shows all zero.
Anki has had this quirk for a long time. Now that I paid more attention to it, I think it makes no sense and it is just a bug.
@dae What do you think? The underlying issue is caused by the fact that in stats.py _revlogLimit doesnāt follow the same deck selection as _limit.
The old stats are unlikely to receive further changes, as they have only been kept around for add-ons, and will be removed at some point in the future.