Comments (17)
Should I continue to use only “Hard" and "Again"?
Generally, I recommend sticking to the old rating habits because FSRS uses your old review history to learn your memory patterns and uses them to schedule new reviews.
But, using "Hard" instead of "Good" even when you recall the card easily can confuse FSRS because when you press "Hard", FSRS assumes that the card is difficult and increases its difficulty. This causes the intervals to grow more slowly.
How will the shorter interval affect my retention?
It will tend to increase your retention but I feel that the intervals are too small. So, I think that you will mostly be wasting your time if you review them this frequently.
You have three options: (the last one is probably the most effective but also the most difficult)
Keeping using "Hard" and set a lower desired retention (say, 0.85 or 0.80)
Start using "Good" and set a desired retention of, say, 0.90
Convert your old "Hard" ratings into "Good" and then re-optimize.
- Do this ONLY IF you are sure that each time you pressed "Hard", you actually meant "Good" and you didn't use "Good" for the cards that were easier than those that you pressed "Hard" on.
- Before doing this, SYNC all your devices and create a BACKUP.
- On the Desktop version of Anki, press Ctrl+Shift+; (the control key, shift key, and semi-colon key at the same time) to open up the debug console.
- Paste the following code into it.
mw.col.db.execute("update revlog set ease=3 where ease=2") mw.col.mod_schema(False)
- Press Ctrl + Enter to run the code. The message shown by Anki gives the impression that the code failed to execute. But, it does get executed.
- Close the Debug Console. Click check database and finally upload your collection to AnkiWeb at the next sync.
- Re-optimize your parameters and start using "Good" and "Again" from now onwards.
Thank you for providing detailed answers, I followed the last option, looks like everything is going well. Here are my new parameters:
0.4265, 0.8556, 2.4316, 7.9731, 5.6787, 1.0437, 0.5065, 0.0523, 1.1659, 0.2704, 0.3670, 2.5389, 0.0767, 0.2754, 1.6863, 0.2190, 4.0000
Is that all right? I start using "Good" and "Again" from now onwards.
from fsrs4anki.
Could you provide the parameters optimized from your data? I need to check whether they are normal.
from fsrs4anki.
Sure. The parameters optimized from my data is:
0.4266, 1.4061, 3.4739, 7.9731, 5.6550, 1.2496, 0.3940, 0.0772, 1.8506, 0.1539, 0.7332, 2.2107, 0.0872, 0.3368, 1.1233, 0.2859, 4.0000
from fsrs4anki.
I preview the intervals and I find that the interval is a little short if you only press hard when you recall. You can preview it by yourself: https://huggingface.co/spaces/open-spaced-repetition/fsrs4anki_previewer
from fsrs4anki.
Does this mean that I need to repeat it more times in the same amount of time, and how does this affect the memorization effect? What's the best thing I can do?
from fsrs4anki.
Does this mean that I need to repeat it more times in the same amount of time, and how does this affect the memorization effect? What's the best thing I can do?
I recommend checking your true retention and set the desired retention at the same level.
from fsrs4anki.
Does this mean that I need to repeat it more times in the same amount of time, and how does this affect the memorization effect? What's the best thing I can do?
I recommend checking your true retention and set the desired retention at the same level.
I found my average retrievability is 91%, should I set the desired retention to 0.91?
from fsrs4anki.
I found my average retrievability is 91%, should I set the desired retention to 0.91?
It's OK.
from fsrs4anki.
I found my average retrievability is 91%
This is not what L-M-Sherlock asked you to check.
Ideally, you should check the true retention in the stats on the Desktop Anki with the FSRS Helper add-on installed by holding Shift when you click the Stats button.
If you don't use the Desktop version, you can use a value close to the values written after Young and Mature in the following graph.
![](https://private-user-images.githubusercontent.com/92206575/307308258-a6900635-0f1d-438c-bc43-aecc627fafc6.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.pKS9LBAYdj2gSgbun2Oa9rls_X9ZXXe1jSuk0_goEbE)
from fsrs4anki.
I found my average retrievability is 91%
This is not what L-M-Sherlock asked you to check.
Ideally, you should check the true retention in the stats on the Desktop Anki with the FSRS Helper add-on installed by holding Shift when you click the Stats button.
If you don't use the Desktop version, you can use a value close to the values written after Young and Mature in the following graph.
![]()
Thanks for your help. I checked the true retention in the stats on the Desktop Anki with the FSRS Helper add-on:
![true retention](https://private-user-images.githubusercontent.com/16689877/307488772-d9c98849-893d-443c-93a7-9b55fe8fab02.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.CdVNau7JQE7exdxxtqURpXoOTS8DzmXZ2Ic6lfOAZDg)
It seems to be close to 90%, it is the same as my current desired retention (0.90).
Should I continue to use only “Hard" and "Again"? Do I need to use "Good" instead of "Hard"?
How will the shorter interval affect my retention?
from fsrs4anki.
Should I continue to use only “Hard" and "Again"?
Generally, I recommend sticking to the old rating habits because FSRS uses your old review history to learn your memory patterns and uses them to schedule new reviews.
But, using "Hard" instead of "Good" even when you recall the card easily can confuse FSRS because when you press "Hard", FSRS assumes that the card is difficult and increases its difficulty. This causes the intervals to grow more slowly.
How will the shorter interval affect my retention?
It will tend to increase your retention but I feel that the intervals are too small. So, I think that you will mostly be wasting your time if you review them this frequently.
You have three options: (the last one is probably the most effective but also the most difficult)
- Keeping using "Hard" and set a lower desired retention (say, 0.85 or 0.80)
- Start using "Good" and set a desired retention of, say, 0.90
- Convert your old "Hard" ratings into "Good" and then re-optimize.
- Do this ONLY IF you are sure that each time you pressed "Hard", you actually meant "Good" and you didn't use "Good" for the cards that were easier than those that you pressed "Hard" on.
- Before doing this, SYNC all your devices and create a BACKUP.
- On the Desktop version of Anki, press Ctrl+Shift+; (the control key, shift key, and semi-colon key at the same time) to open up the debug console.
- Paste the following code into it.
mw.col.db.execute("update revlog set ease=3 where ease=2") mw.col.mod_schema(False)
- Press Ctrl + Enter to run the code. The message shown by Anki gives the impression that the code failed to execute. But, it does get executed.
- Close the Debug Console. Click check database and finally upload your collection to AnkiWeb at the next sync.
- Re-optimize your parameters and start using "Good" and "Again" from now onwards.
from fsrs4anki.
According to my testing, your intervals grow faster now (with new parameters and Good) than previously (with the older parameters and Hard) irrespective of whether the first rating is Again or Pass (Hard or Good).
But, I think that the intervals still grow too slowly. A probable reason is that your true retention earlier was less than 90%.
To get a more accurate value of True Retention, select All history or 1 year in the Stats screen (after opening it while holding Shift).
According to my guess, it would be close to 80%.
from fsrs4anki.
I think that your true retention earlier was less than 90%.
To get a more accurate value of True Retention, select All history or 1 year in the Stats screen (after opening it while holding Shift).
According to my guess, it would be close to 80%.
Here is my Stats screen, I select All history:
![true retention1](https://private-user-images.githubusercontent.com/16689877/307508518-b1c27a73-3559-4d13-b63b-66e673f2e269.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.fR9OYMRUZbqV8cB2hs0OdbMKKHUv6CoYqJ-ySlQlifc)
![true retention2](https://private-user-images.githubusercontent.com/16689877/307508525-ab5f4aed-5504-44d6-9907-4450046f8f84.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.3e9vK0ecR6azqTVHgVkUhZOQ3KlWmkpBX7NvKbnpmSI)
Are the parameters correct?
from fsrs4anki.
There is no easy way to see if the parameters are good or not. What's the RMSE, by the way? You can check it by clicking Evaluate, next to Optimize
in the Deck Options.
But, they don't look strange to me. You have to decide for yourself: Are you comfortable with such short intervals (as seen in the previewer)? If not, you can decrease the desired retention slightly (to say, 0.88 or 0.86).
from fsrs4anki.
RMSE is 2.10%, Log loss is 0.4049.
Maybe I should try it for a while.
from fsrs4anki.
RMSE is 2.10%
This is in the Good range. So, I think that the parameters should be fine.
Now, you just need to decide a value of desired retention based on your needs / expectations. So, I think that you can close this issue now.
from fsrs4anki.
OK, thanks for your help.
from fsrs4anki.
Related Issues (20)
- [Question] Do optimized FSRS parameters synchronize across multiple devices? HOT 2
- [Question] Pressing `Again` on a card causes a reschedule where `Good` becomes 100y, after upgrade to anki 23.12 and reoptimizing HOT 4
- [Question] Initial Parameters for new Deck HOT 1
- [Feature Request] Produce workload graph for an individual HOT 31
- [Feature Request] new card limit in optimizer.find_optimal_retention HOT 10
- [Feature Request] Make it easier to disable SM2 ("anki") simulation in Simulator HOT 1
- [BUG] Give Simulator's graphs titles to distinguish them HOT 4
- [Question] What is the best practice for reschedule cards and update parameters button? HOT 6
- [Feature Request] Improve Learning/Relearning Scheduling HOT 10
- [BUG] Hard interval is longer than Good interval, and longer than Max interval HOT 1
- [BUG] optimizer.find_optimal_retention: ValueError: Axis limits cannot be NaN or Inf HOT 4
- FSRS Simulator File Not a Zip File [BUG] HOT 7
- [Feature Request] Change scheduling based on the type of card used HOT 8
- [BUG] List of compatible add-ons (update)
- [Question] Does review records affecte the parameters when I reset old cards to new cards HOT 2
- [Feature Request] Change to difficulty on new cards HOT 3
- [Question] Not graduating cards the same day affects scheduling? HOT 3
- [Question] Who should NOT use fsrs4anki / the limits to generalization are in the training set HOT 3
- [Question] I haven't used Anki for a while HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from fsrs4anki.