Background

https://github.com/open-spaced-repetition/fsrs4anki is a modern Anki custom scheduling based on free spaced repetition scheduler algorithm. FSRS algorithm is based on the three-component model of memory proposed by Piotr Wozniak and the stochastic shortest path algorithm introduced in my paper.

Recently, someone asked me the difference between FSRS and SM-2. There are many technical differences between them that could make average users confused. So I want to introduce it simply here.

Adaption: per card vs. entire collection

Here I want to quote the words from Wozniak (the creator of SM algorithms):

SM2 uses your performance on a card to determine the next time to schedule that card SM3+ use your performance on a card to determine the next time to schedule that card, and similar cards

Like SM3+, FSRS could use your performance on the entire collection to determine the next time to schedule all cards. FSRS optimizer could be trained from your review logs and make more precise predictions for your memory. And FSRS is open-source, so that you can prove or disprove my claim.

Model: simple vs. complex

I don't mean that the more complex, the better. But SM-2 is oversimplified.

• SM-2 uses E-factor (the ease in Anki's context) to encode difficulty and stability increase, which are separated in memory. FSRS separated them. For details, please see: https://github.com/open-spaced-repetition/fsrs4anki/wiki/The-Algorithm
• In Anki, the stability (the equivalent of interval in Anki) will increase exponentially if you press good constantly when the card is due. However, the stability doesn't increase exponentially in the dataset of MaiMemo. There is a marginal effect. The multiple of stability between each review decreases when the stability increases. FSRS considered this effect, but SM-2 didn't.