Guide

The SM-2 algorithm in plain English

SM-2 is the scheduling algorithm behind SuperMemo and Anki's default. Here is how its ease factor and growing intervals work, without the math.

Part of the Spaced repetition: the complete guide guide.

If you have used Anki, you have used SM-2, probably without knowing its name. It is the algorithm that decides when each card comes back, and for almost forty years it has been the default way software schedules spaced repetition. It is also simple enough to explain without any math.

This guide covers what SM-2 does and where its age shows.

A flat editorial illustration of a single card on a soft blue desk with three review marks beneath it spaced at growing distances, each gap wider than the last

Where SM-2 comes from

Piotr Woźniak designed the SM-2 algorithm for SuperMemo in the late 1980s, one of the earliest pieces of software built around spaced repetition. Anki adopted a variant of it, and through Anki it became the most widely used review algorithm in the world. When people say “the spaced repetition algorithm,” they usually mean SM-2 or one of its descendants.

How it works, without the math

SM-2 tracks two things for every card: an interval (how many days until the next review) and an ease factor (a multiplier that controls how fast the interval grows).

The loop is:

  • You review a card and grade how well it went. In Anki this is the Again / Hard / Good / Easy buttons; in the original it was a 0–5 scale.
  • A good grade multiplies the interval by the ease factor, pushing the next review further out. A card seen at one day might jump to six days, then to two weeks, then a month, then months.
  • The grade also nudges the ease factor itself. Cards you find easy grow their intervals faster; cards you find hard grow them more slowly.
  • A failed card lapses: the interval collapses back toward the start, and the card returns to frequent review until you relearn it.

That is the entire idea. Easy cards drift toward long intervals and stop bothering you; hard cards stay close until they stick. It is the Leitner system made continuous, with a per-card multiplier instead of fixed boxes.

Why it worked so well for so long

SM-2 is simple, transparent, and robust. It needs almost no data, runs anywhere, and produces a sensible expanding schedule that rides the forgetting curve reasonably well. For a 1980s algorithm running on a 1980s computer, it was a remarkable piece of work, and “good enough” turned out to last decades.

Where its age shows

SM-2’s limits come from what it does not model. It never asks the most useful question directly: what is the probability you will remember this card right now?

  • The ease factor is a crude proxy for difficulty, and it can drift in unhelpful ways (Anki users know the “ease hell” complaint, where cards spiral to punishingly short intervals).
  • It does not target a specific retention level. You cannot tell it “I want to remember 90% of these” and have it schedule accordingly.
  • It treats your grades as more precise than they really are, since self-graded recall is noisy.

These are exactly the gaps that FSRS, the modern successor now built into Anki, was designed to close by modelling memory directly and scheduling to a target retention. SM-2 is still perfectly usable, and for many learners the difference is modest, but it is no longer the state of the art.

What it means for you

If you use Anki with default settings, SM-2 (or Anki’s variant) is quietly doing its job and you do not need to think about it. If you want more accuracy or fewer reviews for the same retention, switching to FSRS is the upgrade.

And if your material is not really flashcards, the algorithm question is beside the point: SM-2, FSRS, and Leitner all schedule cards. Scheduling whole sources you study (a chapter, a course, your notes) is a different job, which is the one Memset handles. See the complete guide for where each approach fits.