How to force FSRS to show today's (re-)learned cards tomorrow (1d interval)

Anki FSRS
1

I switched from SM-2 to FSRS. I miss the fact that I got to review all learned cards the next day. In FSRS, a learned card that I answered correctly both times, would be scheduled for review in a few days. I understand that this may be optimal scheduling. But I really like to cement yesterday’s (re-)learned material today, not in a few days; however inefficient that may be. After this point, or if I press easy, the algorithm can do its thing. Is there a way to make FSRS to always have 1d for good for cards (re-)learned today?

Solutions that I thought of:
Raising retention. I have to raise it to 0.93 to get the 1d interval, that’s too high. It would affect the learned cards too much.
“multi day” learning phase. ie setting the last learning step 1d. Not really compatible with FSRS according to the manual:

(Re)learning steps of 1+ days are not recommended when using FSRS.

Currently, I just start decks with an SM-2 preset and change the preset to an FSRS one once a deck of say 100cards is done. But is there better way?

I’m using: Version ⁨24.04 (429bc9e1)⁩

Edit: Apparently the problem is not clear:
Problem: With FSRS the first interval after having learned a card is at 3-4 days.
Desired solution: set that interval to 1 day without affecting retention rate for old learned cards.

Edit 2 (was done after all the answers)

Superimposed
Superimposed1920×1080 111 KB

This graph shows the intervals for

  • Default parameters for 33333313 - Blue
  • Default parameters for 23333313 - Red
  • changed parameter for 33333313 - Green

The only changed parameter is the initial stability for Good. It’s set to the initial stability for Hard.

It’s a janky image overlay, but It shows Green intervals always less than one step behind Blue ones. The interval difference does not diverge further. Not much else changed. I checked for up to 10 answers, it’s the same: one step behind. Still wouldn’t use it change it unless more experienced users say it looks ok. Too many variables.

2

I didn’t really understand your problem. Do you need the first interval for new cards in 1 day?

3

Yes. That is the problem. By default (retention at 0.9), the first interval for new cars is at 3-4 days. I want it to 1d, as it was with SM2.

4

It is 3-4 if you answer the “good” card. For “again” and “hard”, you will most likely get about 1 day.

But if you want to achieve this without changing your habits, then you can try to do it through a filtered deck or through naming FSRS parameters.
It is better to use a filtered deck. Other options are not recommended.

2 Likes
5

it should look like this

(introduced:2-introduced:1) deck:name

But “-introduced:1” not supported
Therefore, use such a search query. And do it before you study the new cards of today.

introduced:2 deck:name

https://docs.ankiweb.net/searching.html#first-answered
Maybe someone has a better solution?

1 Like
6

You can have a 1d learning step, if that’s what feels comfortable to you. The only downside is that it postpones FSRS taking over scheduling – which as you pointed out is inefficient, but it’s not harmful in any way. Nothing will break.

3 Likes
7

1d learning step works for your case.

3 Likes
8

This is really funny. On the one hand, you’re stating the obvious, Sherlock! That’s what I said I wanted. Problem is I’m not getting it.
On the other hand, you could read this and interprete it as an insult. As in “dude, you’re so stupid, you really need the 1d learning steps”
Either way, thanks.

9

Sorry if I offend you. There are three reasons why I don’t recommend using (re)learning steps of 1d+ days:

  1. It may make the true retention deviate from your desired retention.
  2. Hard/Good might show longer intervals than Good/Easy. (same to Hard/good interval is longer than good/easy - Frequently Asked Questions (ankiweb.net))
  3. The next Good interval scheduled by FSRS might be shorter than the previous interval set by learning steps.

If you could accept the phenomenon above, it’s OK to use 1d+ learning steps.

@dae How about adding my explanation into the warning of 1d+ learning step?

Btw, the category of this post should be FSRS.

2 Likes
10

This is something Barbara Oakley recommends in her book. I read her books many times and they’re well researched IMO. Obviously it’s different for everybody but reviewing next day is a good practise IMO. (FSRS almost always schedules my newly (re)learned cards the next day so it’s good for me)

11

Don’t worry he wasn’t insulting you. Just stated the obvious. Use 1d learning step if you will (keeping in mind what LMSherlock said).

12

Thank you! I’ve heard this recommended a lot. But I do it because it’s so powerful in getting a portion of the learned material to “cement” seemingly forever. Forgetting it on day 2 and re-learning on day 3 removes that advantage for me. On top of that, there’s no “cementing” on day 4 either, because at that point the material has lost its “novelty;” I’m not as interested as I was on day 1-2, and it has become disconnected from the related material.

13

I think this way we can exclude the new studied cards of today.
introduced:2 prop:r<1 deck:name

14

I do not recommend that you do this. Also, I do not know which is better than learning stages of more than one day or this.

You can change the first intervals by changing the FSRS weights.

image
image749×134 38.7 KB

https://www.reddit.com/r/Anki/comments/18tnp22/a_technical_explanation_of_the_fsrs_algorithm/
https://open-spaced-repetition.github.io/anki_fsrs_visualizer/

15

Keks, danke dir a million for your time.

  1. But I don’t know where to input the
    introduced:2 prop:r<1 deck:name

In custom scheduling? Would that work with AnkiMobile as well seamlessly? Or only on desktop?

  1. Do I simply change the first four values to the values in the picture?

These are my current values:

**0.4784, 1.4436, 2.3776, 10.9355,** 5.3451, 1.4545, 0.8369, 0.1517, 1.5553, 0.2441, 0.9311, 2.2088, 0.0563, 0.3169, 1.6666, 0.0214, 3.6002

Do I change the first four, so it would be. So I just copy it in?

**0.1000, 0.2352, 0.9387, 13.6999**, 5.3451, 1.4545, 0.8369, 0.1517, 1.5553, 0.2441, 0.9311, 2.2088, 0.0563, 0.3169, 1.6666, 0.0214, 3.6002
  1. I didn’t get what you mean by “I do not recommend that you do this.” If you mean reviewing material after 1 day; I don’t see a disadvantage to it beyond minimal one-time inefficiency. If you mean meddling with the parameters, I’m definitely afraid
16

If you are sure that it is better for you, then do it. But before interfering with the FSRS parameters, figure out how it works.

I don’t have a device or a program to check. But I think it should work.
https://docs.ankiweb.net/filtered-decks.html

deck:name
name - You must change the name of your deck or delete it if you want the cards to be collected from all decks.

That’s what you’re doing.
again=0.1000
hard=0.2352
good=0.9387
easy=13.6999
What if “Desired retention=0.9” should give an interval of 1 day for the “good” button. I do not know how the effect will affect the interval in this case:
https://docs.ankiweb.net/studying.html?highlight=fuzz#fuzz-factor

If you are going to use a filtered deck, then you do not need to change the FSRS weights.

17

I don’t recommend tuning the parameters manually. The initial stability also affects subsequent intervals.

18

I’ve already given you links so you can see what happens.

0.1000, 0.2352, 0.9387, 13.6999, 5.3451, 1.4545, 0.8369, 0.1517, 1.5553, 0.2441, 0.9311, 2.2088, 0.0563, 0.3169, 1.6666, 0.0214, 3.6002

rating history: 3,3,3,3,1,3,3
interval history: 0d,1d,4d,11d,27d,4d,10d,22d

3333	0.94	3.43	10.07	25.04
3332	0.94	3.43	10.07	10.39
3331	0.94	3.43	10.07	2.71
2333	0.24	0.89	2.75	7.32
1333	0.10	0.32	0.95	2.53
4331	13.70	37.46	86.72	7.52

0.4784, 1.4436, 2.3776, 10.9355,5.3451, 1.4545, 0.8369, 0.1517, 1.5553, 0.2441, 0.9311, 2.2088, 0.0563, 0.3169, 1.6666, 0.0214, 3.6002

rating history: 3,3,3,3,1,3,3
interval history: 0d,2d,7d,18d,1.4m,5d,12d,25d

3333	2.38	5.57	8.96	12.47
3332	2.38	5.57	8.96	21.58
3331	2.38	5.57	8.96	2.26
2333	1.44	4.26	7.75	11.12
1333	0.48	2.84	6.23	9.56
4331	10.94	14.66	18.18	7.06

https://open-spaced-repetition.github.io/anki_fsrs_visualizer/

19

There is no one here who would recommend such a thing. But endlessly dissuading the user from what he wants seems rude to me.

1 Like
20

It works, I just forgot the space bar.

(introduced:2 -introduced:1) deck:name