Stealing SuperMemo ideas yet again, here:
In order to plot this, all retrievabilities across all cards on a given day need to be summed. If it ends up being as slow as a snail, perhaps adding a button for the user to click coudld solve it. That way the graph will only be plotted if the users wants to.
@A_Blokee Jarrett told me to ask you
I like this idea. Itād be useful for visualising Ankiās effectiveness.
On a tangent, I had an idea of ārise in average stability over timeā, wdyt?
We could just plot average stability over time
I think Iād also like to see how total knowledge drops if I stopped doing any more cards.
@A_Blokee If you end up doing this can you have two different lines plotted? Basically, one would be āThis is you with Ankiā and another would be āThis is you without Ankiā (lol). In all seriousness though, knowing that would be useful for many people.
I wanted to do this before but didnāt because of the retrievabilities not easy to get from the revlogs and the fact itās fsrs specific.
Search Stats Extended already has one āThis might take a whileā confirmation box so I donāt see why I couldnāt have a second one asking users if they want their data processed more.
I might try to do it with ts-fsrs at some pointā¦ maybe.
Dae is already considering not having SM-2 as the default scheduler so IMO wonāt be a problem on that front.
Also, I think I mightāve misunderstood Expertiumās proposal. I was thinking about a graph that simulates the future reviews and finds out how much total knowledge youāll have over time. As the simulator itself was developed by Sherlock, can @L.M.Sherlock do this?
The simulator does actually already do the future ones, its just not displayed.
As far as I understand it tries to maximize retrievability/time whenever you hit ācalculate recommended retentionā
Is a future one what @Expertium wanted?
I donāt think @Expertium wanted this (from the graph he shared) but itād be awesome to have one, can you do that possibly?
If you can do this in the simulator itād be killing two stones with one bird, as now you can also set new card limit/review limit to zero and get the graph for how total knowledge comes down without doing Anki (IMO it gives you a much better sense of how āstableā the knowledge is).
No simulations for this one, guys. @A_Blokee I meant only historical retrievabilities, from the day of first ever review to today.
Hmmā¦ okay, but why not?
They wouldnāt have to be on the graph at the same time, I thought turn the ātimeā checkbox into a radio button between ātimeā āreviewsā āmemorised totalā or something.
Iād still think the one in the past is better its just a lot harder to implement.
Iāll stick to āIāll add it to Search Stats Extended at some point if I canā.
Unrelated, but would you perhaps consider āprettifyingā the True Retention table? The native one looks ugly.
Yeah those double borders on the table have been bothering me a while XD.
Pie chart? 2 sets of radio buttons for
Day, Month, Year
Young, Mature, Total?
Maybe thatās a little too radical for the native one, If you have any ideas on how to make it look less awful thatād be great.
No, pie chart, please! Itās better to be able to see everything at a glance.
Canāt you make it look like the green-and-red one in the Helper add-on (in Expertiumās picture)?
I agree that the table is better because it is much more information dense.
The main problem that I can see is making it work on devices with small screens (it is mentioned in a previous commit that this is the reason for the current table layout).
The old table is very wide. If we want something like the old table, it also needs to be able to reflow into something that does not look terrible on phones.
Iām having a little play with it to see what I can do, but Iām not sure I can pull off the necessary CSS trickery.
Maybe this thread should be split into a separate topic?
The āRetentionā chartās data partly overlap with āAnswer buttonsā. I think we should think of a way to combine them into one. Wdyt?
New topic, everyone
Turns out ts-fsrs is cool as hell
The addon is a buggy mess right now but this is what you wanted right?
Right now it re-simulates difficulty, which given the current parameters are going to be most accurate is a good thing? Reviews are still actually reviewed at the same time so if anything the difficulty should be more accurate and I shouldent update it to the historical difficulty on every review (?!?)
Relevent code:
Self pr:
The graph is at the very bottom right of revlog reviews if you build that branch.
Dark blue on dark is not a great idea. Other than that itās good. I do wish that you implemented it natively, though.
@L.M.Sherlock take a look at the code once you have some free time
