The second-to-last parameter is zero. So the stability doesn’t increase when you press good.
I have never manually changed the parameters…
I haven’t optimized with the new version, since instructions warn that that will make the android client stop working.
Oops. That makes sense. In this case, FSRS-5 cannot update the stability during short-term reviews.
1 Like
I just optimized all my params in a test profile and all these problems went away…
But I need the android client…
1 Like
The Android client is not yet ready. The backend was today updated to the 24.06.03 version with the 2.19 beta5
You can’t really use FSRS 5 or short-term scheduling without optimizing the parameters. If you are using the old parameters, you are essentially making Anki schedule the cards in (roughly) the same way as FSRS 4.5.
But, I get that you can’t update the parameters because AnkiDroid isn’t ready yet. So, all you can do for now is wait.
2 Likes
I really really want this feature so I’m considering manual workarounds.
Would this work?
- Optimize using the beta.
- Before I need the android client, open my profile in the stable version, empty out the parameters, optimize and sync
- When I get back to my computer sync and optimize using the beta
I’m hesitant to just try in case I end up doing something irreversible to my collection, which would be a disaster to me.
Instead of optimizing every time, you can just save the parameters somewhere and copy them into Anki/AnkiDroid when you switch.
I believe that the stable version of Anki will show the same error as AnkiDroid, which means that opening the profile in the stable version of Anki is not necessary. You can likely just open it in AnkiDroid and modify the parameters there.
1 Like
Great so it should work. FSRS short term scheduling, here I come 
To me though, with a bunch of presets, clearing them out and reoptimizing is faster and easier.
I’m aware of the problem now. Actually, in Beta 3, FSRS-5 will take over the scheduling when the (re)learning steps run out. The stability after lapse would be lower than 0.5d, so the interval of good is < 1d.
1 Like
Ah, that sounds like really good behavior to me actually.
BUT! I just tried it, and if the user has not optimized their settings because they need the android client, the card stays at the same interval forever the same way it does with no learning steps…
Would this strategy work?
- If the parameters are invalid, including a zero value in the list, warn the user and refuse to run.
- Allow reoptimize (all) presets to run whether or not the current parameters are valid, instead of failing with an error if they are not.
That way the only thing a user would need to do to switch between clients with different FSRS versions is hit optimize all presets when they switch client, correct?
Sorry, I do not remember exactly which deck this card was from and under which setting. I believe though it had the default setting.
Same behaviour here
Wait, really 
Does this mean I can get intervals beyond what is shown here
Wait it really does, this is what have always wished for 
Sad for you, but this will be considered as a bug and “fixed” in the subsequent betas because this beta enables the short-term scheduler even for those who don’t want it.
However, I have made a request, which if implemented, will allow you to define your own learning steps and still allow FSRS to control the rest. Here’s a link to that request: FSRS 5: <1d Scheduling and Learning Steps - #54 by vaibhav
1 Like
Pretty please keep it :(((((((( 
It is working exactly like I had in mind
However, I have made a request, which if implemented, will allow you to define your own learning steps and still allow FSRS to control the rest.
Why do you have so many relearning steps, man…
1 Like
This was before me using the short term scheduler 
. And it is because some things can’t get into and stay in my head. In fact this idea has been going through my head for a very, very long time. Look at this for instance. Some 8 months back.
This update solves it all for me. I no longer have to do exorbitant amount of learning steps for ALL cards. Just allocate more steps for cards that NEED it, less for those which don’t.
Today was my first day trying this and it works magnificently 
The reason why I do need such large amounts of relearning steps, is a different thing. It has taken a great toll on my confidence.
I am hoping for FSRS 5, this scheduler and the new upcoming sorting order to help me quite a bit in my plight.
A bit unrelated. I woke up this morning to this amount of interday (re)learning cards
Those are my new learning steps.
The reason why I am using a 1s again learning step is that it shows me the card I failed, RIGHT after the next card it shows to test myself again and see if I get it right. If I get the next card failed, then I am cycling between just 2 cards until I get one of them right the next time. This helps with my cognitive overload by learning one thing at a time.
I dont know if this was always the case (I did not have any issues with this before this update) but now I am caught cycling between multiple cards without getting my previously failed learning cards shown to me in 1s interval as I wanted it.
What is the cause for this, or rather, what is the queue behavior in my case of interday learning cards, and could it be changed
Is it just because a lot of cards have been rated “again” and therefore have been rated as 1s in the night as I left it, and how could I maintain this cycling between 2 cards, even for 2 cards.
For normal learning cards, this doesnt happen.