Anki for Poker¶

Anki is spaced retrieval practice, automated — your daily ~10-minute layer that keeps ranges and heuristics from decaying. It does not replace drilling; it maintains what drilling builds.

What Anki is (and isn't) for¶

It is for: the why-backed heuristics you extract, boundary range decisions, and rare-but-important spots the random Trainer draw won't surface often.
It is not for: replicating the Trainer. The Trainer drills the full spot distribution in context; Anki holds the principles and edge cases. Don't card your whole game.

Keep the deck lean

A huge deck of easy recognition cards is the fluency-illusion trap — it feels productive and isn't. A small deck of atomic, production-forcing cards beats a big passive one.

What to card¶

Heuristic cards: front = a concrete spot; back = action + size + one-line why. Always include the why (it's what transfers). > "BTN vs BB SRP, turn is an overcard that favours my range — barrel freq + why?" → "High freq, small/med — my range stays uncapped and the card improves me more than the caller."
Boundary range cards: drill the marginal hands where the decision is close (not the obvious nuts/trash). Or use Image Occlusion on a range chart (mask a region, recall it).
A formation isn't a card — it's a tag. Tag cards formation::srp-btn-vs-bb, node::turn-barrel, so a formation's cards accumulate as you study its nodes over time.

Four rules for good cards¶

Atomic — one fact per card.
Force production — the front must make you generate, not just recognize.
Decision, not definition — front is a spot, back is action + why.
Be specific — exact position, action, size, depth, or cards interfere.

Settings that match the routine¶

Turn on FSRS, desired retention ~0.90 (it implements expanding intervals automatically).
Learning steps → 1m (so one "Good" graduates a card to days, no in-session bouncing).
New cards/day ~5–10, max reviews/day ~150–200.
Maximum interval ~6–12 months, so even well-known material resurfaces at least once or twice a year (nothing decays silently).

How to grade a card¶

You're scoring recall of the heuristic, not solver-perfection (that's the Trainer's job). Use a tolerance band:

Again — blanked or wrong primary action.
Hard — right action but missed the size/why, or struggled.
Good — recalled action + size + why within tolerance, promptly. (Your default.)
Easy — instant and trivial (rare).

Grade honestly — lenient grading recreates the illusion of competence and tells the scheduler to space cards too far out.

When to make cards¶

Make the card when you extract the heuristic (during study), while it's fresh — not months later when it retires. FSRS then ages it automatically: seen often when new, rarely once mastered. The "long tail" is where old cards end up, not where you start adding them.

Key takeaway¶

Small, atomic, why-backed cards, made the moment you learn something, graded honestly. Anki maintains the map daily; the Trainer builds the skill. Don't mix up their jobs.

Sources: players using Anki for ranges · FSRS / spaced repetition