The habit that makes the math work

A spaced-repetition schedule is a promise: show up roughly daily, and each card will be waiting at the moment it’s about to fade. Break the promise and the model still works — intervals keep elapsing — but your queue turns into a debt pile. Five to twenty-five minutes a day, every day, is the operating point the algorithm was designed around. One heroic two-hour Sunday session is measurably worse than seven fifteen-minute ones, because spacing between sessions is the active ingredient.

The practical shape of a good session:

1. Open the deck list. The due count is today’s obligation; the new count is today’s optional growth.
2. Study. Flip → recall → grade honestly. Repeat until the queue is empty, including re-queued lapses.
3. Read the summary. Recall rate around 85–90% is the healthy band SM-2-type schedulers aim for. Much higher and you’re over-reviewing easy material; much lower and something upstream is wrong (usually the cards, occasionally the pace of new intake).

Reviews first, new cards as capacity

FlashDeck’s queue mirrors the standard advice: due reviews come first (most overdue at the front), with new cards dealt in at a one-per-three ratio up to the deck’s per-session cap. The ordering isn’t cosmetic — due cards decay while you sit on them, new cards don’t. If you only have five minutes, do the top of the queue; the undone new cards wait patiently.

Pacing new cards is the lever most people get wrong. Each new card doesn’t add one card — it adds a future daily review. Rough compounding at a 20-new-per-day pace: within two weeks you’re carrying 100–150 reviews a day, and a missed weekend lands you with 400+ due. If your daily time budget is ten minutes, the sustainable new-card rate is probably closer to 10 than to 40. That’s why the per-deck cap exists — set it in Manage, and treat it as a throttle, not a goal.

Honest grading is the whole game

The scheduler is a feedback loop and the only sensor is your grade. Inflate it — pressing Good because the flipped answer “looks familiar” — and intervals stretch past your actual retention, converting tomorrow’s reviews into re-learning sessions. Sandbag it — pressing Again “to be safe” — and intervals compress, flooding you with reviews you didn’t need.

The discipline that works: flip, then grade what happened before the flip.

• Again — the answer wasn’t there, or it was wrong. Not “took a moment”, but actually absent.
• Hard — got it, but it was a fight: long hesitation, nearly missed a detail, low confidence.
• Good — produced it with normal effort. This should be your most common button.
• Easy — instant, bored, zero doubt. If everything is Easy, the cards are too easy.

And a lapse isn’t a failure event to dread — it’s the scheduler receiving accurate information. A card you fail at day 20 was scheduled wrong; now it isn’t.

Lapses and leech cards

One lapse is weather; three lapses on the same card is climate. When a card keeps coming back as relearning, the problem is almost always the card: it tests two facts at once, the cue is ambiguous, or you never learned the underlying idea. The fix is to open Manage, find the card, and rewrite it — split, sharpen, or delete. The card-writing guide has the patterns. Grinding a malformed card through more lapses is how people decide they “can’t memorize” — it’s the card, not the memory.

Special situations

• Exam next week. Spaced repetition is a long-horizon tool; for a near-term exam, add a Review ahead pass (it deals scheduled cards ignoring their due dates) on top of daily due work — but keep grading honestly, or you’ll corrupt the schedule you resume after the exam.
• Travel / busy stretches. Pause new cards — reviews are the cheaper half. Ten cards of due-only maintenance preserves the deck; ten days of total neglect just means a bigger, humbler comeback session.
• Multiple decks. Session math is per-deck. Rather than juggling three decks’ new-card caps, it’s usually better to push one deck’s new intake and let the others run review-only — the deck list’s due counts make the workload visible at a glance.

Frequently asked questions

How long should a session be?

Long enough to empty the queue, short enough to do daily — for most decks that's 10–25 minutes. A 50-card session at ~5–8 seconds per card is under ten minutes. Consistency matters more than length because the scheduler keeps assigning work: skipping three days doesn't pause the algorithm, it just piles the due queue higher.

Should I do reviews or new cards first?

Reviews first — that's why FlashDeck deals them ahead of new cards (interleaved 1 new per 3 reviews). Due cards are a commitment the schedule already made; new cards are optional capacity. If time runs out, an undone new card costs nothing while an undone due card gets more overdue.

I keep pressing Again on the same card — now what?

That's a leech signal, not a memory verdict. After ~3 lapses, suspend the drill and fix the card: split it into atomic facts, add a distinguishing cue, shorten the answer, or relearn the concept and re-add it. The card-writing guide has the rewrite patterns.

Is it cheating to look at the back before grading?

Flipping the card IS looking at the back — grade after you flip, based on whether the answer was already in your head. The only cheating version is grading 'Good' because the answer looks familiar once revealed. Recognizing a revealed answer is a much weaker signal than producing it — grade what happened before the flip.

What if I miss a week?

The due queue just grows — cards don't forget on a schedule you control. When you come back, work the queue normally (most-overdue-first is already the right order) and pause new cards until the backlog clears. Expect recall to dip; lapses during catch-up are the algorithm doing its job, not punishment.