My Optimized N4 Anki Setup (with stats)
I've been iterating on this for about four months and I'm happy enough with the numbers to share. Current retention rate: 91.3% on mature cards. Here's the full config.
Deck structure
I run three separate decks rather than one combined deck:
- N4-Vocab — 1,700 cards (core vocabulary + context sentences)
- N4-Kanji — 320 cards (kanji recognition + reading)
- N4-Grammar — 180 cards (grammar pattern → example sentence, cloze format)
Keeping them separate lets me weight reviews differently and track where my retention is actually dropping.
Interval settings
| Setting | Value | Notes |
| New cards/day | 15 | Split 8 vocab / 5 kanji / 2 grammar |
| Starting ease | 250% | Slightly above default |
| Interval modifier | 105% | Conservative — I'd rather review more |
| Maximum interval | 180 days | Caps at 6 months |
| New interval (lapse) | 20% | Don't reset fully on fail |
| Lapse threshold | 8 | Card becomes leech after 8 fails |
Review schedule
Morning session: 20–30 minutes before anything else (willpower is highest, interference from new input is lowest).
Average daily reviews: 87 cards. Sounds like a lot but it plateaus once you've built the deck up — mature cards come back slowly.
Card format (vocab cards)
Front: Japanese word (kanji + kana)
Back:
- Reading
- English meaning
- Example sentence (Japanese)
- Example sentence (English)
- JLPT level tag
I don't put audio on the front. Some people do — I find it creates a listening-only pathway that doesn't transfer to reading recognition.
Stats after 4 months
- Total cards: 2,147 active
- Mature cards (interval >21 days): 1,489
- Retention on mature cards: 91.3%
- Average time per card: 6.2 seconds
- Total review time this month: 41 hours
One thing I'd change
I started too many new cards in the first month and built up a review backlog that took six weeks to clear. If you're starting from zero, cap yourself at 10 new cards/day for the first month no matter how tempted you are to go faster.
Happy to answer questions about the setup.