It is on our roadmap for this year 
9 Likes
Is it still only planned for vocab or grammar too?
1 Like
As it is planned right now, it will be available for all content and review types.
14 Likes
to my (very limited) knowledge this would be the first FSRS implementation working with dynamic content, right? Looking forward to what you came up with since I’ve always been wondering how (and if) you want to combine FSRS which is strictly designed for static flash cards, with multiple fill-in sentences
2 Likes
I’m also interested in this. For the grammar and so on I don’t see how this will be a benefit as its a different sentence each time
1 Like
personally I have a lot of trust. From outside it feels like they’re taking a “do it right” approach, considering how often it’s been delayed. I’m sure most of the delay is due to the complexities of combining a unique SRS-system like Bunpro, with something like FSRS which is designed for more classic SRS systems.
5 Likes
I think FSRS will just make mastering the easier points much more difficult, as it will always “optimise it” in a way so it won’t get to the automaticity. Of course it’s probably not true, and they are going under the belt in any way after some immersion, but this hole thing with FSRS makes me uncomfortable in some ways. Probably because I think it’s just beating around the bush, especially with system like bunpro where each sentence has a word you don’t know, and you don’t see the old one in like 12 circles.
2 Likes
Last I checked it’s an opt-in thing anyway, so it’s not like anyone’s forced to use it if it doesn’t end up working for them.
4 Likes
That’s the thing about FSRS - forget about “mastering” stuff. If it deems something easy, seeing it after a long time is exactly what you want, so you can learn more new stuff instead. And if you’re worried about the long intervals anyway, you can always just adjust your target retention. As it approaches 100%, the review intervals shrink drastically.
5 Likes
@Saraph
Even the difference between something like 85% and 92% is huge, at least from my own experience on Anki. I only ever put things very important to remember above 90% because having everything that high produced such short intervals (for cramming for uni tests this was great). It feels like I could get almost as good results with much less work between 80-85%.
3 Likes
So I only just recently activated FSRS for anki and…
oh my god
everything is so much better
please please please please please may the Bunpro SRS arrive soon
4 Likes
Do you know if I get the Anki mobile app that you can use FSRS on it?
Yes you can (android here but iphone should be the same presumably)
1 Like
Commenting here to support the idea of improving the current SRS, but I think there are also other alternatives that could be explored by Bunpro other than FSRS. FSRS is great, way better than the SM-2 algorithm that Anki used previously, and people noticed a huge improvement. But that doesn’t mean it’s the best nowadays.
For reference, Anki is currently using FSRS-6. But here is the benchmark maintained by the same people who develop fsrs-rs, the implementation that is currently being used by Anki. And this is how the different algorithms compare against each other:
As you can see, there are WAY better things than FSRS-6. The main reason why Anki is not using them is because the top ones are DNN based (aka AI). The Anki maintainers don’t want to add something like RWKV-P because it needs to do inference on every single card, something that would consume too much power if running in something like a smartphone, which is understandable. It was discussed here, and just last month they achieved reducing the model to a size below that of FSRS-6 and implement it on an Anki fork for anyone to try on PC. It mentions only FSRS-7, but RWKV-P is also there as an option.
Although it makes sense for Anki to not to adopt this as they’re local first, since Bunpro is a web service first, it would make more sense having something alike here than in Anki, and it would put Bunpro way ahead of other alternatives in the SRS field.
Of course I’m aware of the difficulties something like this would entail, like how expensive would it be to run (it’s not something huge like running an LLM, but still…), or how it would mess up with offline reviews in the app. But I just wanted to share this to be taken in consideration for future improvement, and in any case, just like others, I’ll be happy enough with FSRS when it’s added 
Thanks again!
9 Likes
I just switched two days ago and am working through a newly added backlog of 2400 cards 
ahhh I didnt apply changes to old cards haha so its been very chill
1 Like
Everything you guys say sounds like wizardry to me but it looks like everybody is pretty pleased with this FSRS thing so now I want it too.
7 Likes
I didn’t know about this fork! I’ve been trying it and it works well. Though I am a little confused if you are supposed to have both FSRS and RWKV active in the settings.
Sounds like hell. When you optimize the parameters in the deck options, you can leave the “Reschedule cards on change” off, so it’s only applied to new reviews. If you do reschedule and it gives you many (thousands in your case) dues right now, you can undo this (Edit → Undo or Ctrl+Z). You can also temporarily reduce the target retention rate and reschedule again, so the influx of dues isn’t so sudden. Usually no more than 90% retention rate is recommended (I personally use 85%).