SuperMemo SM-18 Algorithm is Much Better Than FSRS: Practical Difference #podcast

TL;DR

SuperMemo’s latest spacing algorithm is claimed to produce much longer review intervals than FSRS-based setups, reducing total repetitions while maintaining equal or higher retention for some users.

Briefing Cornell Notes

Briefing

SuperMemo’s newer spacing algorithm is portrayed as delivering dramatically longer review intervals—and, in turn, stronger long-term recall—than FSRS-based setups for a key subset of users. The practical takeaway is simple: if someone spends lots of time in spaced-repetition flashcards and wants to minimize review workload without sacrificing retention, the algorithmic choice can change daily life, not just theory.

Both participants frame their case as experience-driven rather than a formal benchmark. They compare SuperMemo against FSRS implementations in apps like Anki and RemNote, emphasizing that SuperMemo tends to require fewer repetitions to reach long intervals (often one-year-plus) and that retention can be slightly higher even while cards are seen less often. One cited personal comparison claims SuperMemo reaches a one-year-or-longer interval with about 2.5 fewer reviews on average, along with a slightly higher retention rate. Another recurring observation is that, across their own multi-app usage, SuperMemo’s “correct-answer” intervals appear “wildly” longer, and recall “feels stronger,” with overall retention reported as equal or higher.

A central non-technical argument focuses on first-review timing. SuperMemo is described as having an “optimum interval” for newly created cards—estimated from a target recall level (around 90% in the discussion)—then adding randomness so cards don’t all reappear at exactly the same time. FSRS-based systems are contrasted as lacking an equivalent concept: newly created cards are effectively scheduled for immediate or near-immediate review (often via short learning steps such as 10 minutes in Anki/FSRS workflows). The participants call this wasteful for users who create cards they already understand, arguing that reviewing something right after creating it weakens memory and multiplies review counts without benefit.

They also argue that SuperMemo’s long-term optimization philosophy better matches how many people actually study. SuperMemo is described as optimizing for long-term retention even if short-term performance may look worse, while FSRS is said to cater to users who need high short-term recall for exams. One participant adds that if someone is studying for a limited window (e.g., medical school or MCAT-style preparation), shorter intervals can be useful; but for long-horizon learning—where the goal is to remember after months or years—SuperMemo’s approach reduces total review time.

Several practical examples are used to illustrate interval growth. The discussion claims SuperMemo can produce large “interval multipliers,” such as a card moving from roughly 10 days to 130–150 days after a correct response, and that this kind of growth is rarely seen in FSRS even when desired retention is set very low (e.g., 83% in RemNote). Another complaint targets delay handling: if a card is missed, FSRS is described as capping how much the interval can expand, whereas SuperMemo is portrayed as granting much larger intervals after late reviews—interpreted as the system recognizing that the memory held up better than expected.

The episode ends with a recommendation to test rather than trust claims. Because SuperMemo’s latest algorithms are described as closed source, the participants say direct technical comparison is limited, so they urge users to run a free trial (or a freeware version) on a subset of cards and compare first-interval timing, interval lengths after correct answers, and overall retention. The message is not that FSRS is useless, but that it is not “close” to the latest SuperMemo for the long-interval, low-review workload that many experienced users want.

Cornell Notes

The discussion argues that SuperMemo’s latest spacing algorithm can outperform FSRS-based systems in real-world use by producing much longer review intervals while maintaining equal or higher retention. A key practical difference is how newly created cards are scheduled: SuperMemo uses an “optimum interval” (estimated from a target recall level) before the first review, while FSRS-based setups lack an equivalent concept and often push near-immediate reviews via short learning steps. The participants claim this leads to fewer total repetitions and stronger long-term memory, especially for users who create coherent cards and study for long horizons rather than short exam windows. They also describe SuperMemo as showing larger interval growth and more generous interval increases after delayed reviews. Because the latest SuperMemo algorithms are described as closed source, they recommend users test with a trial to compare intervals and retention directly.

What is the most important practical difference between SuperMemo and FSRS that the participants emphasize?

They focus on review workload and interval length. SuperMemo is described as scheduling cards with much longer intervals—so users see cards less often—while still achieving equal or higher retention. They tie this to fewer repetitions needed to reach long-term intervals (e.g., one-year-plus) and to the feeling that recall is stronger when intervals are longer.

How does “optimum interval” for new cards create a scheduling difference?

In SuperMemo, newly created cards receive an “optimum interval” based on a target recall level (described as around 90% in the discussion), then randomness disperses first-review timing so not every card returns the same day. FSRS-based apps are contrasted as lacking an equivalent first-interval concept, effectively treating new cards as needing immediate or near-immediate review (often via short learning steps like 10 minutes). The participants argue that for cards created from material the learner already understands, immediate review is unnecessary and increases repetitions.

Why do longer intervals matter for memory quality, according to the discussion?

The participants claim that reviewing more frequently can weaken memory strength. They use an analogy: binge-watching a show and then testing later differs from spacing viewing across months, even if total exposure is similar. Their claim is that cramming-style repetition (many reviews in a short time) produces weaker long-term recall than spreading reviews over longer periods.

What role does study goal (exam-focused vs long-term learning) play?

SuperMemo is described as optimizing for long-term retention even if short-term performance may be worse. FSRS is said to fit users who need high short-term recall for school or medical exams, where short intervals are desirable. The participants argue that if someone doesn’t need high recall immediately for an exam window—and instead wants retention years later—SuperMemo’s long-interval strategy can reduce time without sacrificing retention.

How do the participants describe interval growth and delay handling?

They claim SuperMemo shows large interval multipliers after correct answers (examples include moving from ~10 days to ~130–150 days). They also argue that FSRS caps how much intervals can grow when a user is late: if a card was scheduled for 10 days but reviewed after 30, FSRS allegedly increases the next interval far less than SuperMemo. Their interpretation is that SuperMemo better reflects the learner’s actual retained strength.

What practical test do they recommend to resolve skepticism?

They recommend using SuperMemo’s free trial (or a freeware version) on a subset of cards and comparing: (1) first-review intervals for newly created cards, (2) how long intervals become after correct answers, and (3) retention as measured by the app. They argue that users can observe interval differences quickly without needing technical algorithm comparisons.

Review Questions

Which scheduling mechanism for newly created cards is presented as the biggest non-technical reason SuperMemo can reduce review counts compared with FSRS-based systems?
What evidence types do the participants rely on (and what do they avoid), and why does that matter for interpreting their claims?
How does the discussion connect interval length to memory strength, and what real-world study goal does it say can change which algorithm feels better?

Key Points

1
SuperMemo’s latest spacing algorithm is claimed to produce much longer review intervals than FSRS-based setups, reducing total repetitions while maintaining equal or higher retention for some users.
2
A major practical difference is first-review scheduling: SuperMemo uses an “optimum interval” for newly created cards, while FSRS-based systems are described as lacking an equivalent and often using near-immediate learning steps.
3
The participants argue that more frequent reviews (cramming-style) can weaken long-term memory quality, so longer intervals can improve recall strength even with fewer reviews.
4
Study goals matter: SuperMemo is framed as optimized for long-horizon retention, while FSRS is framed as better aligned with exam-driven short-term recall needs.
5
The discussion claims SuperMemo shows larger interval growth after correct answers and more generous interval increases after delayed reviews than FSRS.
6
Because SuperMemo’s latest algorithms are described as closed source, the recommended way to verify claims is to run a trial and compare intervals and retention on a personal card subset.
7
The participants caution against treating FSRS and SuperMemo as equivalent, arguing the practical differences can be substantial for power users with large, long-term collections.

Highlights

SuperMemo is described as assigning newly created cards an “optimum interval” (based on a target recall level) before the first review, while FSRS-based systems are described as lacking this concept and often pushing immediate reviews.

The discussion claims SuperMemo can generate large interval jumps—e.g., from around 10 days to 130–150 days after a correct response—more often than FSRS.

A recurring theme is that longer spacing can strengthen long-term recall: reviewing the same number of times over shorter periods is said to produce weaker memory than spacing those reviews across months or years.

The episode repeatedly urges skepticism-proofing through personal testing: compare first-interval timing, post-correct intervals, and retention metrics using a trial rather than relying on community claims.

Topics

SuperMemo vs FSRS
Optimum Interval
Interval Growth
Long-Term Retention
Spaced Repetition Trials

Mentioned

FSRS
SM-18
SM2