What is FSRS-6?

FSRS stands for Free Spaced Repetition Scheduler. Version 6 is the current iteration, with 21 parameters fitted to predict your personal forgetting curve. It replaces the SM-2 algorithm Anki shipped with for two decades.

The key difference: SM-2 multiplies the previous interval by a fixed ease factor. FSRS-6 estimates two latent variables for each card, stability (how slowly you forget it) and difficulty (how much each lapse hurts), then schedules the next review at the moment your recall probability drops to a target threshold (default 90%).

The result is fewer reviews of cards you know cold and more frequent reviews of cards you keep missing, automatically.

Why chess needs FSRS, not flat intervals

Chess training has three properties that break naive flashcard schedulers:

  • Massive variance in difficulty. The first move of the Italian and a 14-move Najdorf line are not the same problem. SM-2 treats them similarly until you have hundreds of reviews on each.
  • Forgetting is unevenly distributed. You will see 1.e4 e5 2.Nf3 Nc6 3.Bc4 hundreds of times in your own games. You will see a sideline of the Symmetrical English maybe twice a year. SM-2 schedules them on the same curve. FSRS notices the difference within a handful of reviews.
  • Lapses are noisy. You blunder a known move because you were tired, not because you forgot it. FSRS uses your overall recall pattern to estimate stability, so a single bad review does not throw off the whole schedule.

The practical effect: with FSRS-6, you spend most of your study time on the 5 to 10% of positions that are actually slipping, instead of grinding the same well-known main lines.

How DawnChess applies FSRS-6

Every card in DawnChess uses the same FSRS-6 scheduler:

  • Opening repertoire moves. Each move you choose to play is a separate card. The trainer surfaces only the moves whose due date has arrived.
  • Tactics. Combinations and missed tactical opportunities from your own games. Multi-move puzzles complete the full continuation before scheduling, so you grade the whole sequence as one card.
  • Master game key positions. Critical moments from 200 hand-curated grandmaster games. The first three reviews include progressive hints (arrow, then highlight) before the card behaves normally.
  • Endgames and theoretical positions. Lucena, Philidor, basic checkmates, all on the same schedule.

Because the algorithm is shared, you have one unified review queue across every part of your training.

Response-time grading

DawnChess auto-grades reviews based on how long the move takes:

  • Under 5 seconds: graded Easy. The card is well-known, and the next interval grows aggressively.
  • 5 to 10 seconds: graded Good. Standard interval growth.
  • Over 10 seconds: graded Hard. The interval grows more slowly because you needed to think.
  • Wrong move: graded Again. The card resets and reappears soon.

You can override the grade if needed, but the default is to let recall speed do the grading work. This means you do not have to interrupt your practice flow to click a button on every card.

Getting started

FSRS-6 in DawnChess is on by default. There is nothing to configure. Connect your Chess.com or Lichess account, pick an opening repertoire (or build your own from your games), and start training. The algorithm will start producing personalised intervals after about 20 reviews per card type.

If you want to see the math, the FSRS-6 paper and reference implementation are open source. DawnChess uses the published default parameters until enough review data exists per user to fit personal ones.