How to Add Cards to an SRS Deck Without Corrupting Difficulty

The short answer

Add new cards to your SRS deck in small, controlled batches (5 to 10 per session), and only after each word has passed a genuine first recall test outside the main schedule. Adding cards carelessly, whether too many at once or without proper encoding, corrupts the FSRS Difficulty and Stability values that drive every future interval. The result is a deck that over-tests easy items, under-tests hard ones, and generates review debt faster than you can clear it. If you want the scheduler to work for you, protect its first data point.

How FSRS uses your first rating to shape every future interval

FSRS tracks three values for every card: Difficulty (D), Stability (S), and Retrievability (R). Together they form the DSR memory model.¹ Stability is the number of days until Retrievability drops to 90%. Difficulty (a 1 to 10 scale) governs how much Stability grows after a successful review. Retrievability is the probability you can recall the item right now. For a full explainer of all three components, see how spaced repetition works.

The critical detail: your very first rating on a new card initializes both D and S. FSRS uses the first grade (Again, Hard, Good, or Easy) to set the card's starting Difficulty and its first Stability value. Every subsequent review updates those numbers incrementally. If that first data point is wrong, the entire scheduling trajectory drifts from reality, because all future D and S calculations build on the initial state.

Why careless additions corrupt your deck

False first ratings from working memory echo

When you add a brand-new word and review it seconds later, your brain is not retrieving from long-term memory. It is re-reading the sensory echo still in working memory, which decays within roughly 10 to 18 seconds when rehearsal is blocked.² Press "Good" on that phantom success and FSRS assigns a moderate Difficulty and a long initial Stability. When the card returns days later, the echo is gone. You fail, FSRS raises Difficulty, and the card enters a cycle of overcorrection. For the full science behind this failure mode, see why you should not put new words straight into spaced repetition.

Repeated failures inflate Difficulty permanently

The opposite path is just as damaging. You add a word you have never encoded, fail it three times in ten minutes, and each "Again" press ratchets Difficulty upward. FSRS interprets the repeated misses as evidence that the item is inherently hard. In reality, the item was never ready for scheduling. The card ends up with an artificially high D value and tight intervals that persist long after you have actually learned the word. Older SM-2 systems called this "ease hell." FSRS mitigates the worst cases, but even FSRS cannot fully recover from a Difficulty value built on data that reflects missing encoding rather than true item complexity.

Too many new cards overwhelm encoding

Even if each individual card gets a clean first rating, adding too many at once creates a different problem. Distributed practice research consistently shows that spacing items across sessions produces stronger retention than massing them into one block.³ Add 30 new words in one sitting and each gets less cognitive attention. Flashcard studies confirm the same pattern: students who space their new-card introductions across sessions retain more than those who front-load.⁴

There is also a compounding review-load problem. Every new card generates future reviews on an exponential schedule. Ten new cards today might mean 10 reviews tomorrow, 15 in three days, 25 in a week. Add another 10 cards each day for a week and the review queue balloons. Within two weeks you either skip days (which defeats spacing) or spend so long reviewing that you stop adding new material entirely. For more on this and related pitfalls, see common spaced repetition mistakes.

How to batch new cards without breaking your schedule

The goal is to keep first-rating data clean and review load sustainable. A few concrete rules help:

1. Cap new cards based on review time, not ambition. A common heuristic: total daily reviews (new + due) should fit inside the time window you will actually honor. If you have 15 minutes, 5 to 10 new cards per day is more sustainable than 30.
2. Clear due reviews before adding new cards. Due items are the ones closest to forgetting. Delaying them while you chase the dopamine of fresh material lets Retrievability drop and failures cluster. Reviews first, new cards second.
3. Space new-card additions across sessions. If you have 40 words to learn this week, spread them across 5 to 7 sessions rather than loading them all on Monday. This distributes the encoding work and the resulting review debt.³
4. Monitor your review queue. If due cards spike above your daily budget, pause new additions until the queue stabilizes. FSRS cannot save you from a deck that outgrows your calendar.

The missing step: acquisition before scheduling

Batching solves the volume problem, but it does not fix the data-quality problem. Even a single new card per day will corrupt FSRS if the first rating comes from a working-memory echo instead of real retrieval.

The fix is a pre-SRS acquisition step: a short pipeline between first exposure and the moment the card enters the main schedule. Effective acquisition includes:

1. Meaningful first exposure. See the word in context, connect it to meaning, and produce it (type or say it) rather than just reading it passively. Production during initial exposure strengthens encoding.⁵
2. A short interleaved delay. Complete an unrelated task for 5 to 10 seconds before the first blind test. This clears working memory and forces retrieval from a real trace, not an echo.⁶
3. Active recall, not recognition. Type the word from a minimal cue. Multiple-choice tests measure recognition and can encode false associations from wrong answer options.⁷
4. Graduation into SRS with attempt history. Only after a successful blind recall should the card enter your FSRS queue. Pass the full attempt log (including any failures) to the scheduler so initial D and S values reflect real difficulty.

If you use Anki, you can approximate this with learning or intraday steps before cards graduate to the main interval queue. The key is: do not treat the first seconds after seeing a word as equivalent to a day-3 or day-30 review. For a detailed walkthrough of priming, a short interleaved distractor, and blind recall before FSRS takes over, see Fixing the First-Turn Bottleneck.

How LinGoat handles this automatically

LinGoat builds both batch control and pre-SRS acquisition into its core workflow, so you do not have to manage either manually.

New vocabulary enters a dedicated laddering system before it ever touches the main FSRS schedule. The ladder has three stages: priming (you type the word while it is visible), an interleaved micro-buffer (5 to 10 seconds of unrelated practice that flushes working memory), and blind active recall (you produce the word from a translation cue). Every attempt, correct and incorrect, is logged. When the word graduates, that full telemetry initializes FSRS with honest Difficulty and Stability values, not noise from a lucky guess or a string of failures on an unencoded item.

Card flow is automatic. LinGoat controls how many new words enter the ladder based on your current review load and study pace. You never manually set a "new cards per day" limit or worry about overwhelming the queue. Due reviews always take priority over fresh introductions. The result: clean first-rating data, sustainable review volume, and a deck that stays calibrated over months of study.

LinGoat uses written sentence practice, not isolated flashcards, so every review doubles as production practice. You type full sentences in your target language, get word- and grammar-level feedback, and each graded piece enters FSRS when it is ready. See how LinGoat works on the homepage or open the app to try it.

References

1. Ye, J. (2024). FSRS: The Algorithm. Open Spaced Repetition Wiki. https://github.com/open-spaced-repetition/awesome-fsrs/wiki/The-Algorithm
2. Peterson, L. R., & Peterson, M. J. (1959). Short-term retention of individual verbal items. Journal of Experimental Psychology, 58(3), 193-198. https://doi.org/10.1037/h0049234
3. Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354-380. https://doi.org/10.1037/0033-2909.132.3.354
4. Kornell, N. (2009). Optimising learning using flashcards: Spacing is more effective than cramming. Applied Cognitive Psychology, 23(9), 1297-1317. https://doi.org/10.1002/acp.1537
5. McDaniel, M. A., Howard, D. C., & Einstein, G. O. (2009). The read-recite-review study strategy: Effective and portable. Psychological Science, 20(4), 516-522. https://doi.org/10.1111/j.1467-9280.2009.02325.x
6. Karpicke, J. D., & Roediger, H. L. (2007). Expanding retrieval practice promotes short-term retention, but equally spaced retrieval enhances long-term retention. Journal of Experimental Psychology: Learning, Memory, and Cognition, 33(4), 704-719. https://doi.org/10.1037/0278-7393.33.4.704
7. Roediger, H. L., & Marsh, E. J. (2005). The positive and negative consequences of multiple-choice testing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(5), 1155-1159. https://doi.org/10.1037/0278-7393.31.5.1155