FSRS-5

[joshdavham]

I've noticed that work seems to be underway for updating the various FSRS implementations to v5.

At first glance it seems like the changes are the following:

1. Updated parameters
   • FSRS v5 has now has 19 weights vs 17 weights
2. New initial difficulty
   • Found here
3. New stability formula for short term memory
   • Also found here

Are there other changes I'm missing? And do we have the greenlight to start updating the various FSRS implementation to V5?

[L-M-Sherlock]

In my opinion, the greenlight is the next Anki release.

[joshdavham]

Hmm I'm not sure if I totally understand. Could I get an example?

Suppose I have a card that I made on Day 1 and I rate it as 'Good' each time it comes up, until say Day 100 where there's a lapse and I fail it by rating 'Again'. Because the card will now show up again on Day 100, does it get the formula applied to it then?

Or rather does it simply mean that, if a given card has already been reviewed within the last 24 hours, that the next stability calculation will be the short term one? If it's this one, do new cards that haven't been reviewed yet have the short term formula applied to them?

[L-M-Sherlock]

For example, you need to rate 'Good` twice for a new card in the same day in Anki to enter in the long-term review. So the first rating is used to initialize the initial stability, and the second rating is used to update the memory state via short-term formula.

[joshdavham]

Okay I think that makes sense. But just to confirm, in your example, if that card were to 'lapse' in the future, it would have that short term formula applied again?

[L-M-Sherlock]

If a card were to 'lapse' and had one or more extra relearning stages in the same day, the short-term formula is applied again.

[joshdavham]

Alrighty that completely makes sense!