[Feature request] Estimated Total Knowledge over time

Anki FSRS
43

Any chance that you will implement it in Anki natively?

44

Possibly. As I’ve said I should have done it natively from the start.

45

SSE should be updated now :+1::partying_face::sleeping:

46

I guess it’s caused by the same problem mentioned here: Set Due Date doesn't update the interval of card

So the Estimated total knowledge in Card Retrievability section is not perfectly accurate.

In addition, suspended cards are excluded in Card Retrievability section. It also induced difference.

Btw, if I understand it correctly, the last value of “Estimated total knowledge over time” shows the sum of retrievability of the yesterday, right?

47

Nice catch! It’s always the obvious stuff I miss :person_facepalming:.

Ok I’ll stop aiming for the estimated total knowledge. I feel like there is something else wrong with the addon because the stabilities of the cards the addon generates don’t line up with the stored stabilities in Anki.

const expected = last_stability[num_cid].toFixed(2)
const actual = JSON.parse(cards_by_id[num_cid].data).s.toFixed(2)
if (expected != actual) {
    inaccurate_cids.push({ cid: num_cid, expected, actual })
}

image
image1597×161 37 KB

Is that a problem?

48

Could you test it with one card in a new deck? I guess it would help us locate the problem.

49

image
image504×442 29.8 KB

image
image783×505 82.8 KB

image
image609×421 17.4 KB

image
image695×364 22.6 KB

image
image628×383 13.6 KB

image

Is this ok? You can see the difference in the forgetting curve and the displayed graph.

As the graph is an estimate anyway it may not be a problem worth solving?

50

image
image1256×766 61.1 KB

I don’t know why it didn’t reach 100% here.

51

That’s because of the way the candlestick works. It displays the beginning and end values for that period of time.

image
image597×341 15.3 KB

52

Fine. I will dive into the code and try to solve it.

53

Now I’m even more convinced that candlestick charts are not suitable for Anki

54

Ok I’ll try to make it a line graph. Never really liked the candlestick anyway it was just convenient.

55

image
image619×355 17.1 KB

Same card as earlier:
image
image613×350 14.7 KB

Much better. I’ll probably give an option to replace the load chart like this as well.

56

I just noticed that changing the parameters in the deck config doesnt affect the graph…

57

Of course, it should affect the graph. What’s the problem?

58

I was using config.fsrsWeights instead of config.fsrsParams5.

Fixing that still didnt entirely fix the stabilites though I don’t think :frowning:

59

image
image3744×606 255 KB

I guess I find the same problem you mentioned?

image
image1600×1388 188 KB

Edit: I fixed that and still found that the consistency happened after the 2nd review:

image
image1920×480 126 KB

I guess it’s problem of ts-fsrs.

Edit: I located the problem and reported it to the maintainer of ts-fsrs. It will be fixed soon.

60

pr：Fix/wrong calc elapsed days by ishiko732 · Pull Request #144 · open-spaced-repetition/ts-fsrs · GitHub

@A_Blokee would you mind testing it?

61

Before:
image
Now:

image
image738×137 14.9 KB

That fixes it for a lot of cards but some are still wrong
It may have something to do with rollover?

62

I guess so. Let’s discuss it in the PR.