I will make this brief, bunpros team should make it their next priority sincerely to add FSRS in the year of 2025 hell even this month coming off of anki to bunpro I love so much bunpro has to offer, varied sentences, different degrees of learning andい so many other features, but FSRS cant come any sooner the unreliability of standard SRS is not enough to cut it for people wanting to learn your vocab system. if FSRS were added many many anki users would no longer have a reluctancy to jump over here. please integrate it as soon as possible.
6 Likes
They’ve mentioned they are going to add a Fsrs opt-in, but not a replacement for the current system.
Sounds like overall they are going to make an opt-in option for it.
2 Likes
Just confirming this and a few other review related overhauls are all planned. Had hoped to release them earlier but our pipeline currently has a few things in it.
14 Likes
That would be really great. I’m currently planning on stopping using bunpro for vocab once I finish the N5 vocab deck (but continue using it for grammar) because of the massive drop in retention I observe in bunpro compared to anki with fsrs. Although that drop may not be entirely explained by the SRS algorithm (and may not perfectly correlate with actual progress in proficiency in the language), I would seriously reconsider my plan if bunpro allows me to use FSRS.
Bunpro’s SRS implementation is very forgiving. I think a big difference is the reset to SRS1 vs -1 by default for failed vocab. Have you tweaked this? It should lead to better retention (classic Anki style relearning on a forgotten word).
1 Like
I admit I did not (and will probably be too lazy to try it out knowing that bunpro will have FSRS in the future and that I can temporarily migrate my vocab studies to anki in the meantime). Wouldn’t this unnecessarily shorten all intervals for a missed card though? Or does the system remember that it’s not actually a knew card but a learned one that happened to have a lapse?
In the classic Leitner system, a forgotten card goes back to the initial SRS state (meaning you’re fully relearning it). This will shorten the interval of every missed card, because the assumption is, that it was forgotten.
By default, Bunpro uses an alternative system where a forgotten card goes back by one box. Hence the interval of the forgotten card is shortened, but not fully resetted. I can’t speak for the Bunpro team, but my best guess is that the assumption is, that a lapse does not actually mean that you’ve fully forgotten a term.
So on Anki (even with FSRS under most circumstances), forgetting a card will lead to a full reset, since it’s being relearned. This will obviously lead to a higher workload.
Bunpro lets you adjust this setting under Settings → Reviews → SRS strictness
If you experience higher retention on Anki compared to Bunpro, there could be a lot of reasons for it (card quality, etc). But if you use different SRS strictness settings on both systems, I would assume that tweaking this will have a large impact, where even without FSRS, your retention will improve.
Further reads (and image source for the first two):
6 Likes
How do you guys feel about the ghosts system in Bunpro? I currently have ghosts on for vocab, and left the SRS strictness at the default 1. But I am struggling with some vocab in a level that I dont see them often and get them right about 50% of the time. So the vocab keeps going up and down in SRS but never progresses. I’m not sure if the ghosts are helping me much in those cases, so I am considering removing them and making the SRS more strict for vocab.
The ghosts have been very helpful initially with new vocab that I am not familiar with
I went ahead and set it so that my SRS level decreases by half if I get it wrong, and adds a ghost. It’s a bit annoying when I get something wrong because of a simple typo, but it does strike a good balance overall. It has definitely increased my retention while still being less punishing than a complete reset. Maybe you could try that?
Edit: I’d like to add that I also changed my review schedule to have shorter intervals. It increases reviews per day but that seems to help also. Might be worth giving that a shot too
1 Like
I wish there were a system to better handle medium term memory, where you are familiar with a word/grammar point but confuse it with other things that have a similar meaning or sound the same.
After a while, ghosts turn into “what’s the last thing I got wrong that means x?” When I first run into something I haven’t seen in a while, a lot of times, I can remember 4-5 things that mean the same thing that would fit but not the correct answer.
I also wish there was a an automatic way to review similar items/nuance differences.
1 Like
What type of typo? Do you fail cards if you knew how to write it but hit a wrong key by accident? I usually just ESC and re-do in these cases. Or when multiple words are possible but they aren’t marked as alternatives, I still don’t fail the card. I only fail it if I actually make a mistake
2 Likes
I’m doing Bunpro on a phone, so sometimes what happens is I’ll be going through he vocab reviews and end up in a situation where I make a typo and then my muscle memory causes me to hit the enter button and move to the next one before I have time to react to my error. I hope that makes sense lol.
Thanks a lot for the info! Afaik for FSRS, even if it resets to the initial state, at least on my system, the next intervals after answering the card correctly are still much higher than for newly cards (could be wrong on this one).
I think I will try resetting the cards to the initial state for a trial run. Right now I am mostly reviewing grammar and I think the default settings of just reducing the level by 1 is not enough for some grammar which intervals are too long. This might lead to some massive increase in reviews - maybe then it’s beneficial to increase the intervals between levels, though that’s not great for new cards.
So +1 on the need to include FSRS in bunpro, imho it’s really needed.
Absolutely, I’m not arguing against FSRS, I was explaining what my best guess is in regard of a large part of retention discrepancy between Anki and Bunpro
Sorry, I should’ve been clearer that this +1 was towards the general thread and not your regarding your post, which I think was really informative.