absent-sapphireā¢15mo ago
I need help with the Anki algorithm
Hello, I don't understand why the intervals between "correct" and "hard" are that far appart.
What I can tell you is that it's been a since october 2023 that I had not touched this deck.
Maybe Anki considers that if I answer correctly the card despite leaving the card covered with dust, it means that it should have a bigger interval?
Also, the card interval is "8 days" in the broswser.
Can someone help me understand?
15 Replies
rare-sapphireā¢15mo ago
I don't really understand all this (esp with FSRS), but note your Ease % drops when you grade something hard. It's now at 210% down from the default 250%.
Actually at a quick glance at the docs, card Ease might not even be a factor as far as FSRS is concerned. š¤
rare-sapphireā¢15mo ago
absent-sapphireOPā¢15mo ago
No, I put the ease factor manually back to 250% with an add-on. Don't worry about the ease factor.
absent-sapphireOPā¢15mo ago
absent-sapphireOPā¢15mo ago
@Stevi (JP 38k | KR 31 | ENDE ā) @Nekyo @Kiddo (JST) any idea?
rare-sapphireā¢15mo ago
If it's been a while since you've touched the deck, I think Q14 that I posted in the image above, very possibly explains the situation clearly...
quickest-silverā¢15mo ago
also if it's FSRS... it only takes into account the score on the first view/review of the day for the memory state
"A5: FSRS only takes into account one review per day. If you review a card multiple times per day, only the chronologically first review will be used by the optimizer. "
and yes... FSRS doesn't use "ease factor" it uses another field instead that I don't see in the second screenshot...
quickest-silverā¢15mo ago
quickest-silverā¢15mo ago
S is the card stability, D is the card difficulty
FSRS uses D instead of "Ease"
so either the 23.xx series of Anki don't show those or the deck doesn't use FSRS
so it's probably SM2 v3
for SM2 "Answering cards later than scheduled will be factored into the next interval calculation, so you receive a boost to cards that you were late in answering but still remembered."
on Anki's version of SM2, hitting Again sends it back to "relearning" status and Current Interval gets multiplied by New Interval, hitting Hard multiplies Current Interval by 1.2, hitting Good multiplies Current Interval by 2.5 (well Current Ease)
absent-sapphireOPā¢15mo ago
Thank you @MichaĆ«l P (ćć¼ćå 26) and @Daichi for taking the time to answer my question. Yes, I received a huge boost Michael.
I never used FSRS. As you can see in the screenshot, it is deactivated.
Do you guys use FSRS?
inc-lavenderā¢15mo ago
the research shows that it can reduce the number of total reviews you have after you give everything time to equalize for a few months. (lower number of reviews for the same retention.)
The way that it treats individual cards is more adaptive, and it essentially attempts to adjust to target the average retention rate that you set.
It seems to be a fairly effective algo that is an upgrade from the old algo.
I would look into it. I think its neat.
quickest-silverā¢15mo ago
Iāve been using FSRS since the early releases
Thereās a massive amount of hype around it. Some of the hype is close enough to correct, but some of it is really wrong.
Like Ā«Ā solving ease hellĀ Ā» is only semanticsā¦ it doesnāt use Ā«Ā easeĀ Ā». If you check the authorās own published comparison between fsrs and sm2, if you fail the card at first review youāre actually worse off with fsrs.
absent-sapphireOPā¢15mo ago
Thank you for your answer.
Ok, but did you personally notice less reviews? A better retention rate than with Anki SM2 algorithm?
quickest-silverā¢15mo ago
I had a pretty high retention rate with SM2 already
and I started retiring cards ealier
what I noticed is that I got like thousands of reviews added when I switched from SM2 to FSRS
and it resurfaced cards again (less than thousands, but still several hundreds) when I reoptimized
absent-sapphireOPā¢15mo ago
Ok, thank you.0