How Many New Anki Cards a Day? The Review Load Maths

Anki ships with a default of 20 new cards per day and a cap of 200 reviews per day (you can see both values in the default deck configuration in Anki's source code). Those two numbers are related. The Anki manual's deck options page says it plainly: "If you are consistently learning 20 new cards a day, you can expect your daily reviews to be roughly about 200 cards/day."

That roughly 1:10 ratio is the most useful number in spaced repetition, and most people meet it the hard way. This post walks through where it comes from, how to turn it into minutes, and how to choose a new-card number you can still sustain in month six.

On this page

Why one new card becomes many reviews

A new card is not one piece of work. It's a promise to review that card again and again, at growing intervals, for as long as the deck lives. With Anki's default learning steps of 1 minute and 10 minutes, a new card is seen at least twice on its first day. Answer Good on the last step and it graduates with a 1-day interval. After that, each Good answer multiplies the interval by the card's ease, which starts at 2.5 (the manual describes this as increasing "the delay by approximately 2.5x").

To see what that means for workload, I built a simplified model of that schedule: a card you never fail, with each interval about 2.5 times the last (1, 2, 5, 12, 30, 75 and 188 days, rounded). Added up, that brings the card back 1, 3, 8, 20, 50, 125 and 313 days after you first learned it. That's seven reviews in its first year, front-loaded into the first few weeks.

Now add 20 new cards every day. On any given day you're reviewing the cards you added 1 day ago, 3 days ago, 8 days ago, and so on. So in this model the daily review count is roughly 20 multiplied by the number of those checkpoints you've passed:

• Day 7: checkpoints at 1 and 3 days, so about 2 × 20 = 40 reviews
• Day 30: 1, 3, 8 and 20 days, so about 4 × 20 = 80 reviews
• Day 90: add the 50-day checkpoint, so about 100 reviews
• Day 365: add 125 and 313, so about 140 reviews

That's the best case, because it assumes you never forget anything. You will. Anki's own studying guide says you'll typically use the Again button "about 5-20% of the time". A failed card drops back to a short interval and has to climb the ladder again, which adds reviews on top of the ideal schedule.

What forgetting does to the numbers

I ran the same simplified model with a 10% failure rate per review: a failed card restarts at a 1-day interval, and its future intervals grow more slowly, which is roughly how Anki's older SM-2 scheduler treats a lapse. With 20 new cards a day, the average daily review count in my model (these are my simulation's figures, not numbers from Anki or its manual) came out at about:

• Day 30: about 100 reviews (5 per daily new card)
• Day 90: about 140 reviews (7 per new card)
• Day 180: about 180 reviews (9 per new card)
• Day 365: about 220 reviews (11 per new card)

This is a toy model, not Anki's real scheduler. It ignores the Hard and Easy buttons, the small random fuzz Anki adds to intervals, and FSRS, the newer optional scheduler. But it lands close to the manual's figure, and it shows the thing that catches people out: the load keeps growing for months after you settle on a new-card number. Week one at 20 new cards feels easy. In this model, month six at the same setting means more than four times the daily reviews of day seven.

It works in reverse, too. If you stop adding cards, reviews fall off quickly, because most of the load comes from young cards with short intervals. That's why the manual's advice for an overwhelming pile is simply "introducing fewer new cards each day until your review burden decreases."

Turning reviews into minutes

Card counts are abstract. Minutes are what you actually have to find in your day. The conversion needs one number you have to measure yourself: your average seconds per review. Anki's statistics screen shows how many cards you studied today and how long it took, so a week of normal sessions gives you a reliable figure. As a rough guide rather than a measured figure, simple word cards might take you 6 to 10 seconds, and sentence cards with audio can easily take 15 or more.

With the 1:10 rule as a planning ratio, the steady-state formula is short:

• Daily review minutes ≈ new cards per day × 10 × seconds per review ÷ 60

Which simplifies to new cards × seconds ÷ 6. Some worked examples at 9 seconds per review:

• 10 new cards a day: 10 × 9 ÷ 6 = 15 minutes of reviews
• 20 new cards a day: 20 × 9 ÷ 6 = 30 minutes
• 30 new cards a day: 30 × 9 ÷ 6 = 45 minutes

Those are review minutes only. New cards cost extra, since each one passes through the learning steps on its first day and often needs a few more repetitions if you fail one. Budget an extra few minutes on top for every 10 new cards until you've measured your own pace.

Pick the number from your time budget, not your ambition

Rearrange the formula and you can work backwards from the time you really have:

• New cards per day ≈ 6 × review minutes you can sustain ÷ seconds per review

If you can give flashcards 20 minutes a day and you average 8 seconds per review, that's 6 × 20 ÷ 8 = 15 new cards. If you only have 10 minutes and your cards take 12 seconds each, it's 6 × 10 ÷ 12 = 5 new cards. Five a day sounds tiny, but it's 1,825 cards in a year, which is a decent chunk of a beginner's core vocabulary. (If you're studying Japanese, how many kanji you need runs the same kind of pace maths for characters.)

Whatever number you pick, judge it on your worst realistic day, not your best. The schedule doesn't care that you had a deadline. If you can only find 10 minutes on a bad Thursday, a setting that needs 30 minutes a day is going to produce a backlog sooner or later.

The settings that change the ratio

Desired retention (FSRS). If you've switched to the FSRS scheduler, the manual calls desired retention "the most important setting in FSRS." The default is 90%, which it describes as "a good balance of retention and workload", and it warns that above 90% "the workload increases very quickly, and above 97% the workload can be overwhelming." Recent versions include a simulator in the FSRS section that estimates your workload, in reviews or minutes per day, at different retention values. That's far better than any rule of thumb.

Maximum reviews per day. The default cap of 200 is a safety valve, not a target. By default, Anki also stops showing new cards once that limit is reached, unless you turn on "New cards ignore review limit". If you hit the cap regularly, the cap isn't the problem. Your new-card setting is.

Your own failure rate. Cards you keep failing (Anki calls them leeches) generate reviews out of proportion to their value. A card that's ambiguous or tests two things at once will cost you far more than ten reviews. Rewriting or suspending a handful of bad cards can noticeably lighten the load.

Check the forecast, not just today's count

Anki's Future Due graph shows what's coming, and the stats page also reports a "daily load" figure. The statistics section of the manual explains how it's calculated: add up 1 divided by the interval of every card. A card on a 10-day interval contributes 0.1 reviews a day, a card on a 50-day interval contributes 0.02. It's a direct way to see that young cards, not your thousands of mature ones, drive most of the work.

A practical routine: look at that daily load once a week. If it's still climbing and your sessions are already at the edge of your time budget, cut new cards now, before the backlog appears. Increasing new cards is easy to do later. Clearing three days of missed reviews on top of a full schedule is not.

Measure minutes as well as cards

The weak spot in all of this is the seconds-per-review number. People underestimate it, and it drifts as you add harder cards. Logging your flashcard sessions as real minutes, separately from reading or listening, keeps it honest. In LangTrack that's a timer you start and stop around the session, or two taps to repeat yesterday's entry, and because each session is logged by activity, the per-language stats show whether SRS time is crowding out the rest of your study.

If flashcards start eating the time you meant for input, that's your signal to lower the new-card number. Cards are a support for reading and listening, not a replacement. For the background on why spaced reviews work at all, see the science of spaced repetition, and for choosing between SRS apps, see Anki vs other SRS tools. If you're weighing card counts against hours as a progress measure, tracking vocabulary vs tracking time covers the trade-offs.