How Smart Practice Works

Three sciences.
One adaptive workout.

Most apps drill random flashcards. Times Machine borrows from decades of cognitive science to put your kid's brain on a schedule it can actually keep.

The Problem with Memorization

When a kid says they “know” their times tables, what they often mean is they can compute one — slowly, while you wait. That's the wrong target. The actual goal is automaticity: instant recall without calculation. That's what frees up working memory for the harder math down the road (long division, fractions, algebra). And it can't be reached by drilling random equations on flashcards.

What works instead — what the cognitive-science literature has been telling teachers for fifty years — is the combination of spaced repetition, timed retrieval, and immediate correction. Times Machine packages those three into a single adaptive practice loop.

One.

Spaced Repetition

Memory has a forgetting curve. A fact you learned this morning is crystal-clear; the same fact next week is hazy. Re-encountering it right before you'd actually forget — at the inflection point of the curve — is when re-learning locks in. Encounter it too soon and you're wasting practice; too late and you've relearned it from scratch.

FSRS (Free Spaced Repetition Scheduler)is the modern algorithm for predicting that inflection point per fact, per learner. It models difficulty, stability, and retrievability as three live parameters. Anki researchers refined it over 30 years and it measurably outperforms the older SM-2 approach. Every fact in Times Machine — “6 × 7”, “13²”, “8 × 4” — has its own FSRS schedule, computed for the specific learner.

Want the deep dive? The FSRS algorithm is open-source. Read the wiki →

Two.

Speed Matters

Curriculum-Based Measurement of Math fluency (CBM-M) has shown, across hundreds of studies, that speed of recall— not just accuracy — predicts long-term math success. Two kids who both get “7 × 8” right look identical on a worksheet. But if one answered in 0.6 seconds and the other in 5 seconds, they're not at the same place. The 5-second kid is calculating. The sub-second kid is recognizing.

Times Machine times every answer. Each fact gets a three-way grade:Mastered (correct AND under the threshold), Practiced (correct but slow), or Missed (wrong). The fast-threshold is age-tuned — younger learners get more time, adults get tighter — and you can override it per-profile. The 3-tier feedback makes the difference between “I'm getting it” and “I know it” legible to the kid in the moment.

Three.

Cover-Copy-Compare

When a learner misses a question, the obvious fix is to show them the correct answer. But passive correction (“the answer was 42”) has weak retention compared to active retrieval: the learner sees the answer briefly, it disappears, they retype it from memory.

That protocol is called Cover-Copy-Compare, and the CBM literature consistently shows it produces 40–60% better retention than passive correction. Times Machine triggers it automatically on every Missed answer: brief flash of the answer, then the same equation re-presents for the learner to type the answer themselves before continuing.

Adaptive Difficulty

FSRS handles the spacing; we add two more controls on top.

Push Mode.If the learner is hitting ≥85% Mastered on recent attempts, the system surfaces new facts faster. They don't get stuck waiting on items they already know.

Floor Protection. If accuracy on the last 10 attempts drops below 60%, the system pauses introducing new facts and surfaces only reviews until the learner stabilizes. Prevents the spiral where a kid gets discouraged and gives up.

Both kick in automatically. No knobs for the kid to fiddle with.

The Result

On the published research, the combination of spaced repetition + timed retrieval + Cover-Copy-Compare produces 4–5× faster mastery than equivalent time spent on flashcards alone — and the learning sticks(months later, not days later). That's the science. The product translates it into a loop a six-year-old will actually keep doing.

References

Wozniak, P. SuperMemo algorithm SM-2. The original spaced-repetition algorithm, and the lineage FSRS extends.
Shinn, M. R. Curriculum-Based Measurement of Math Fluency. Research lineage establishing speed of recall as predictor of math success.
Skinner, C. H., Beatty, K. L., Turco, T. L., & Rasavage, C. Cover-Copy-Compare: A Method for Increasing Multiplication Performance. School Psychology Review, 1989.
Roediger, H. L., & Karpicke, J. D. The Testing Effect. Cognitive Psychology research on active retrieval vs. passive review.
FSRS for Anki — open-source implementation and wiki

Back to Times Machine

Three sciences. One adaptive workout.