After over 2 years of using the regular scheduler, I decided to use FSRS. I’m assuming the model has a lot of data to train on so my config should be pretty good.
The thing is, my reviews went up drastically since I did that 2 days ago. And when I look at the estimator, it seems it’ll go up another 30% slowly until end of May and then go back down until late 2026 where it should be lower than it is now.
At the same time, I see that my learning reviews are super down (9 instead of maybe 30 per day before) but my green reviews are the ones being up a lot.
Can someone help me understand that? Is it normal? Did I mess up anything? I’m using the model’s learning steps btw.
Don’t use Reschedule cards on change if you don’t want too many reviews.
If you have too many reviews, lower your Desired retention (DR). Use Compute minimum recommended retention (CMRR) to find out the lowest DR can go.
Use different presets for different material so you can optimise the DR/params for each seperately.
I also suggest you read through Deck Options - Anki Manual to understand what’s going on. Use simulator to see how many reviews you’re expected to get and accordingly change the settings.
Is this from the FSRS Helper add-on? I don’t think those LS are particularly guaranteed to work so you’re fine not using them if you want to.
If you entered a target retention that exceeds your current retention on mature cards by a lot, then your review count is going to go up, even with FSRS. What is your current retention and what target did you set?
Okay, I think my retention was like 83% and the target right now is set to 90%.
I used the estimator to see how low I can get without a drastic increase in reviews and 87% seems like a sweet spot for my tolerance. My reviews will go up for a few months but that’s fine!
On my android phone, whenever I changed the desired retention and tried to recalculate the parameters it said “the best parameters are already applied” after a few seconds of processing. I thought it was weird but anyway.
Yesterday I tried the same thing on my laptop, and every little change I made to the desired retention; it offered new settings!
I already have like 20 less reviews today and I see the forecast curb going down a lot.
For reference, since I switched to FSRS a month ago my reviews have gone up by 20% every day… although they feel much less repetitive, I tought that was weird and it kept going up.
Also, I do have like 110,000 reviews or so it says under the “Optimize Current Preset” button.
Changing the Desired Retention (DR) doesn’t affect calculating new parameters. DR is a separate factor, applied after the parameters process the review history on your card.
But the reason you’re seeing a difference between AnkiDroid and desktop is that they aren’t using the same version of the algorithm.
Has to do more with your reviews. If you do more reviews and optimise, FSRS might give you new params. Well, it’s not worthwhile to do it frequently. Once a month is fine.
But then, which version of the algorithm should I use? The one on the laptop actually giving me less reviews rather than more review like the one on the phone does? Can I read about the difference in the macos & android algorithm to make an educated choice?
You should use the most updated version on each device. There’s nothing wrong with the slightly older version of the algorithm. The FSRS developer keeps promising no-more-changes-for-a-while, but then has an idea that could improve something a little bit, and FSRS gets updated. They aren’t monumental changes.
But since you are using different versions of FSRS, how long has it been since you’ve Optimized on each device?
My reviews will go up for a few months but that’s fine!
They aren’t monumental changes.