[Question] why the seed includes review_time.getTime()?

[L-M-Sherlock]

https://github.com/open-spaced-repetition/ts-fsrs/blob/aaa154376876da9dc8176daddcd22335ba5318ec/src/fsrs/abstract_scheduler.ts#L91-L96

Does it mean I will get different interval if I review the same card in the same date?

[ishiko732]

Does it mean I will get different interval if I review the same card in the same date?

Yes, when the interval > 2.5, it is possible to get different intervals on the same day due to variations in seconds. The initial idea was based on the conditions at that moment.

To maintain the same interval within the same day, you can use time % 86400000 (1 day).

[L-M-Sherlock]

The fuzz seed implemented in Anki is generated from card id and the number of reviews of the card. The seed will not change if the user doesn't review it. It's a good design because the fuzz factor will not change over time.

fn get_fuzz_seed_for_id_and_reps(card_id: CardId, card_reps: u32) -> Option<u64> {
    if *crate::PYTHON_UNIT_TESTS || cfg!(test) {
        None
    } else {
        Some((card_id.0 as u64).wrapping_add(card_reps as u64))
    }
}

source: https://github.com/ankitects/anki/blob/f00211df359f8214e9e97688776876a724f59266/rslib/src/scheduler/answering/mod.rs#L602-L609

The current implementation of fuzz seed in ts-fsrs can cause a weird case: the interval may decrease over time. For example, if the original interval is 100 days, and the fuzz factor is 1 in the first day, the fuzzed interval would be ~110 days. If I don't review it and delay the review to the next day, based DSR model, the original interval will increase to ~105 days. But the fuzz factor may decrease to 0, then the fuzzed interval would be ~95 days.

[ishiko732]

Anki is generated from card id and the number of reviews of the card.

Since the card type in ts-fsrs does not have a card_id, making it incompatible with Anki's approach, should I remove the time variable?