I’m using Anki to review math error‑correction questions and logical reasoning problems (procedural knowledge). I have several questions about FSRS vs SM‑2 for this use‑case.
For math/logic procedural cards, which scheduler works better in practice, FSRS or SM‑2?
When FSRS is enabled, all learning steps for new cards must be within one day. That means I have to grade the same new math problem multiple times in short sessions before it graduates to review. I’m worried that repeated short‑interval reviews may create an “illusion of mastery” (memorizing the answer instead of thinking through the logic). Also, will this frequent short‑term grading pollute my FSRS history data and break parameter optimization?
Are there recommended parameter tweaks, card‑making or grading strategies to make FSRS fit math‑style thinking‑heavy content?
If you’re trying to memorize what’s on the cards – FSRS. Memory is memory (at least when it comes to facts), so there’s no distinction between different subjects.
If you’re trying to practice what’s on the cards – neither. There’s no evidence that either of the algorithms, or even spaced-repetition scheduling in general, is what you need for that. It’s just the wrong tool for the job.
Practical example --
A card for 6 x 7 → 42.
The point is to memorize the connection between those so you don’t need to do the actual math to get to the answer.
If you get this card wrong, you need to study it again soon. When you correctly recall the answer, you’ll be ready to see it at longer and longer intervals.
A card for 2186 x 697 → 1523642.
The point is to practice the steps to multiply a 4-dig number by a 3-dig number and reach the correct answer. [What’s the first step? Multiply 6 x 7 – so it’s good you already memorized that!]
If you get this card wrong, you need to understand what went wrong – did you fail at the steps or did you fail at the arithmetic? But either way, it might not be necessary to ever do this exact problem again. If you failed at the steps, you’d probably benefit more from doing a completely separate problem.
A card for Explain the steps to multiply a 4-dig number by a 3-dig number …
The point is to memorize the steps.
If you get this card wrong, you need to study it again soon. When you correctly recall the answer, you’ll be ready to see it at longer and longer intervals. But this is the most boring, most difficult, and least effective way to study this. Practice problems work much better at getting the skill down.
[That said, if you’re going to try it anyway, you might as well use FSRS. Anecdotes and common sense suggest you’d need a lot fewer repetitions, spaced further apart, for any given practice problem than you would for memorization, and FSRS gives you a much easier way to adjust for that, using Desired Retention (DR).]
I think you answered your own question there. See what a bad fit it is?
Do not tweak your parameters, or let an LLM tweak them for you, or use someone else’s parameters you found on the internet.
My goal is exam‑oriented retention of math mistakes. I don’t have a better way to figure out when I should revisit these past errors, so I’m using Anki for this purpose. I am already making mistake‑based cards, and I grade them purely based on whether I can come up with a complete solving idea on my own.
Given this scenario, do you have any recommended practical approaches for this use case?
