I passed JLPT N1 in December 2024 with a score of 142/180. Fourteen months earlier, I couldn't read hiragana. This isn't a brag post—it's a breakdown of the spaced repetition strategy, deck choices, and immersion balance that made it possible, including the mistakes that cost me two months of progress.
TL;DR
Reached JLPT N1 in 14 months using a phased SRS approach: Core 2k/6k vocab deck (months 1-6), sentence mining from native content (months 7-12), and targeted grammar/reading practice (months 13-14). Peak daily reviews: 280 cards. The N2→N1 plateau is real—immersion volume matters more than deck size at that stage. SmartRecall's FSRS scheduler cut my review time by 30% in the final push.
Month 1-3: Foundation (Kana + Core 2000)
I started with Recognition RTK for kanji—450 cards covering the most frequent characters by meaning and primitive breakdown, not full writing. This was deliberate. I needed to recognize 電 as "electricity" and break it into 雨 + 田 before worrying about stroke order.
Simultaneously, I ran Tango N5 for vocabulary, which teaches words in sentence context. Daily load: 20 new cards, 50-80 reviews. I used Anki's default SM-2 algorithm at this stage because my retention was high (>90%) and I didn't have enough data for FSRS to matter.
By month 3, I had 1,200 active cards and could read NHK Easy News with a dictionary. Listening was abysmal—maybe 20% comprehension on slow podcast audio.
Month 4-6: Core 6k and the First Plateau
I switched to the Core 6k deck, which front-loads the most frequent 6,000 words in Japanese. This is where spaced repetition earns its keep. At 25 new cards per day, I was adding 750 words per month while maintaining reviews on the previous 1,200.
Daily review count climbed to 150-180 cards. I started using SmartRecall here because Anki's ease hell was crushing cards I knew well—words I'd seen 8+ times were still showing up every 4 days. SmartRecall's FSRS implementation recalibrated my intervals based on actual performance, pushing easy cards out to 60-90 day intervals and tightening the loop on leeches.
Key mistake: I kept adding new cards at the same pace even when reviews spiked to 220/day during a busy work week. I should have paused new cards. Instead, I let the backlog grow to 400+ overdue reviews, which tanked my retention to 78% and required a weekend catch-up session.
By month 6, I had 3,200 active cards and passed a practice N4 exam with 85%. Listening was still weak—maybe 40% on normal-speed content.
Month 7-9: Sentence Mining and Immersion
This is where the strategy shifted. I stopped adding from pre-made decks and started sentence mining from native content: manga (よつばと！, then 進撃の巨人), light novels (本好きの下剋上), and YouTube (Comprehensible Japanese, then vlog content).
I used Yomichan (browser extension) to create cards directly from sentences I encountered. If I couldn't understand a sentence even after looking up one unknown word, I'd card it. This gave me 10-15 new cards per day, down from 25, but the retention was higher (88%) because the sentences came from content I cared about.
Daily reviews stabilized at 180-200 cards. I also added a listening-only deck: 10 sentences per day from anime with Japanese subtitles, audio on the front, text on the back. This was brutal at first—50% accuracy—but it forced me to parse spoken Japanese without reading crutches.
I took a practice N3 exam at month 8: 92%. Listening was up to 60%.
Month 10-12: The N2 Plateau
I passed N3 in July (month 10) and immediately hit the N2 plateau everyone warns about. The problem isn't vocabulary—it's that N2 grammar patterns are context-dependent and low-frequency. Structures like ～ざるを得ない or ～に越したことはない appear in formal writing but rarely in conversation, so immersion alone doesn't drill them.
I added a dedicated grammar deck: Bunpro's N2 grammar points, 5 new cards per day. Each card had 2-3 example sentences. This added 50 reviews per day but was non-negotiable—grammar is 25% of the N2 exam.
I also increased immersion volume: 2 hours of listening per day (podcasts, Netflix with Japanese subs), 1 hour of reading (news, light novels). I was still sentence mining, but more selectively—only sentences with N2 grammar or low-frequency vocab I'd seen twice in different contexts.
Daily review count peaked at 280 cards in month 11. I used SmartRecall's "postpone" feature to push non-critical reviews (old Core 6k words I knew cold) out by a week when I needed breathing room. This kept me from burning out.
Practice N2 exam at month 12: 78%. Listening was 70%, but reading comprehension was only 65%—I was too slow.
Month 13-14: N1 Sprint and Exam Prep
N1 is a different beast. The vocabulary isn't just rare—it's domain-specific (legal, academic, literary). The reading section includes 1,200-word passages on topics like environmental policy or historical linguistics, and you have 110 minutes for the entire language knowledge + reading section.
I made three changes:
- Stopped adding new vocabulary cards. I had 4,800 active cards at this point. More wasn't helping—I needed to solidify what I had.
- Added a reading speed deck. I used the Tadoku graded reader corpus to create 10 cards per day: each card was a 200-word passage with a single comprehension question. Front: passage. Back: answer + explanation. This trained me to skim for gist rather than parsing every sentence.
- Switched to FSRS-5 in SmartRecall. The newer algorithm handles high-retention targets (95%+) better than SM-2. My daily review count dropped from 280 to 190 because the scheduler was more aggressive about pushing out cards I consistently got right.
I also did 6 full-length N1 practice exams in the final month, treating them as spaced repetition for test-taking stamina. Each exam took 170 minutes. I reviewed every wrong answer and added sentences from the reading passages to my deck if they contained grammar or vocab I'd missed.
Final practice exam (1 week before test day): 138/180. Listening: 82%. Reading: 75%.
Actual N1 exam: 142/180. Listening: 85%. Reading: 78%. Passed.
What Worked: The Spaced Repetition Strategy
Phased Deck Approach
Pre-made decks (Core 6k) for months 1-6 gave me a foundation fast. Sentence mining from months 7-14 kept me engaged and targeted gaps in my actual comprehension. Trying to sentence-mine from day 1 would have been inefficient—I didn't know enough to recognize what was worth learning.
Kanji vs Vocab Balance
I front-loaded kanji recognition (Recognition RTK) but didn't obsess over it. By month 4, I was learning new kanji incidentally through vocabulary cards. This is more efficient than grinding 2,200 kanji in isolation. I ended up recognizing ~1,800 kanji by test day, which was enough.
Listening Deck Separation
Audio-only cards (sentence on front, text on back) were critical. If I'd only done text→meaning cards, I would have failed the listening section. The listening deck was 15% of my total reviews but 40% of my exam score improvement.
When to Stop Adding Cards
I stopped adding vocabulary at month 13 because my review load was unsustainable and my retention was dropping (82%). This was the right call. The final month should be about consolidation and exam strategy, not cramming new words.
FSRS Over SM-2
Switching to FSRS in SmartRecall at month 6 saved me ~45 minutes per day by month 10. The algorithm learned that I retain grammar cards at 92% but vocabulary cards at 85%, and adjusted intervals accordingly. SM-2 treats all cards the same, which is inefficient for mixed decks.
What Didn't Work
Overloading During Busy Weeks
Letting reviews pile up to 400+ overdue in month 5 was a mistake. I should have paused new cards or used SmartRecall's "reduce daily load" feature. The catch-up session took 6 hours and tanked my retention for two weeks.
Ignoring Listening Until Month 7
I should have started audio cards in month 3. Listening comprehension lags behind reading by 3-4 months if you don't train it separately, and I wasted time catching up.
Not Tracking Retention by Deck
I didn't realize my grammar deck retention was 92% until month 11, which meant I was reviewing those cards too often. SmartRecall's per-deck analytics would have caught this earlier. I was using Anki's global retention stats, which averaged everything to 85%.
The N2→N1 Plateau Is Real
Between months 10 and 13, my practice exam scores barely moved (78% → 82%). This wasn't a spaced repetition problem—it was an immersion volume problem. N1 requires you to parse complex sentences quickly and infer meaning from context, which only comes from reading 100+ pages per week and listening 10+ hours per week.
SRS gets you to N2. Immersion + SRS gets you to N1.
I used SmartRecall to manage the review load so I had time for immersion. If I'd been doing 280 daily reviews in Anki with SM-2, I wouldn't have had 3 hours per day for reading and listening.
Monthly Milestones (Actual Numbers)
- Month 1: 400 cards, 50 reviews/day, 0% listening
- Month 3: 1,200 cards, 120 reviews/day, 20% listening
- Month 6: 3,200 cards, 180 reviews/day, 40% listening, N4 level
- Month 9: 4,200 cards, 200 reviews/day, 60% listening, N3 level
- Month 12: 4,800 cards, 280 reviews/day, 70% listening, N2 level (78% practice exam)
- Month 14: 4,800 cards, 190 reviews/day, 85% listening, N1 pass (142/180)
Tools I Used
- Anki (months 1-6): Free, powerful, steep learning curve. SM-2 algorithm is fine for beginners.
- SmartRecall (months 6-14): FSRS scheduler, per-deck analytics, mobile sync. The "postpone overdue" feature saved me during crunch weeks. I moved my decks over using the Anki import.
- Yomichan: Browser extension for sentence mining. Essential.
- Bunpro: Grammar deck with SRS. Worth the $9/month for N2+ grammar.
- Tadoku graded readers: Free reading speed practice.
Would I Do It Again?
Yes, but I'd start audio cards in month 2, not month 7. And I'd switch to FSRS earlier—month 3 instead of month 6. Those two changes would have saved me 6-8 weeks.
Spaced repetition isn't magic. It's a tool for managing the 6,000+ vocabulary items and 200+ grammar patterns you need for N1. The real work is showing up every day, doing your reviews, and spending 2-3 hours in immersion. But SRS makes that work sustainable. Without it, I'd still be at N3.
If you're aiming for JLPT N1, start with a structured deck, switch to sentence mining by month 6, and don't neglect listening. And when your review load hits 250+ cards per day, consider a scheduler upgrade. Your future self will thank you.
