FSRS vs SM-2: Which Algorithm Actually Works Better?

If you use Anki, Mochi, or any serious spaced repetition tool, there's an algorithm deciding when you see each card. For decades, that algorithm was SM-2. Now there's FSRS — and the difference is bigger than most people realize.

SM-2: the algorithm from 1987

SM-2 was created by Piotr Wozniak for SuperMemo in 1987. It's elegant in its simplicity:

Each card has an "ease factor" starting at 2.5
When you get a card right, the interval grows by that factor
When you get it wrong, the interval resets and the ease drops
That's basically it

SM-2 worked well enough to become the default algorithm in Anki, which made it the most widely-used spaced repetition scheduler in the world. Millions of medical students, language learners, and knowledge workers have built their study habits on it.

But SM-2 has fundamental problems.

The ease factor death spiral

SM-2's biggest flaw is that ease factors only go down easily — and rarely recover. Press "Again" a few times on a hard card and its ease drops to 1.3, meaning intervals barely grow. The card shows up constantly, wasting your time on something you actually know but once struggled with.

Anki users have spent years developing workarounds. The "Straight Reward" add-on, manual ease resets, custom scheduling code. The fact that an entire ecosystem of patches exists tells you something about the underlying algorithm.

One-size-fits-all intervals

SM-2 uses the same interval multipliers for everyone. But people learn differently. A medical student reviewing anatomy cards has different forgetting patterns than a language learner reviewing vocabulary. SM-2 can't adapt to these differences.

FSRS: machine learning meets spaced repetition

FSRS (Free Spaced Repetition Scheduler) was developed by Jarrett Ye and integrated into Anki in October 2023. Instead of fixed rules, it uses a machine learning model trained on over 700 million reviews from real Anki users. As Ye described the motivation: "SM-2 treats every learner the same. FSRS learns how you forget."

The core insight: instead of a single "ease factor," FSRS tracks three parameters for each card:

Stability — how long until you'll forget this card (measured in days)
Difficulty — how inherently hard this card is for you
Retrievability — the probability you can recall this card right now

These parameters update with every review, and the model learns your personal forgetting patterns over time.

The results

Based on testing across 700 million Anki reviews, the results are clear. Compared to SM-2 at equivalent retention rates:

20-30% fewer daily reviews — you maintain the same retention with significantly less work
Better predictions — FSRS predicts when you'll forget a card more accurately than SM-2
Personalized scheduling — after about 200 reviews, FSRS calibrates to your individual memory patterns

For a student doing 200 reviews per day on SM-2, switching to FSRS could mean 140-160 reviews for the same retention. That's 20-30 minutes saved daily.

The configurable retention target

FSRS lets you set a target retention rate — say 90%. The algorithm then optimizes all intervals to hit that target as efficiently as possible. Want to relax and accept 85% retention for fewer reviews? Slide it down. Preparing for boards and need 95%? Slide it up and accept more reviews.

SM-2 has no equivalent. You're stuck guessing whether your intervals are too long or too short.

So which should you use?

Use FSRS if:

You do 50+ reviews per day (the time savings compound)
You've been using SRS for a while and have review history to optimize from
You want your scheduling to adapt to you, not the other way around
You're tired of the ease factor death spiral

SM-2 is fine if:

You're just starting out with SRS and have fewer than 200 reviews
You use a tool that doesn't support FSRS yet
You don't want to think about algorithms at all (SM-2's defaults are okay for casual use)

The honest answer: FSRS is better for almost everyone. The 20-30% efficiency gain is real and well-documented. The only reason to stick with SM-2 is if your tool doesn't offer FSRS.

What we're building

Memstride ships with FSRS as the default algorithm — zero configuration required. Smart defaults that work out of the box, with a one-click optimize button once you have enough reviews for personalization.

No ease factor death spirals. No manual tweaking. Just better scheduling from day one.

But a great algorithm alone isn't enough — if you quit because of review debt, the algorithm never gets a chance to help. That's why Memstride also includes review debt forgiveness to keep you studying through real life.

See how Memstride compares to other tools in our Anki alternatives guide.

Join the waitlist — get 1 week of Pro free and 3 months of early-bird pricing when we launch.