How we audit Anki decks
Most coverage tools stop at the system level: "this deck covers Cardiovascular." That's not useful — every Step 1 deck "covers" Cardiovascular. The interesting question is which parts. This page explains the work behind our numbers.
Granular by design — 915 subtopics
We broke Step 1 down into 915 testable subtopics across 16 areas, and MCAT into 408 subtopics across 32 areas. The granularity is the point: a deck can hit 100% of "Cardiovascular" and still miss preload/afterload entirely. The outline catches that.
Authored from the resources students actually study from — First Aid, Pathoma, Boards & Beyond, and UWorld for Step 1; AAMC, Khan Academy, and Princeton Review for MCAT.
Within the outline, each subtopic carries a high-yield designation derived from emphasis across primary sources. About 64% of Step 1 and a third of MCAT subtopics are flagged high-yield.
How we determine high-yield
High-yield designations are calibrated against the canonical Step 1 sources — First Aid bold/italic items, UWorld topic frequency, Pathoma chapter emphasis, Boards & Beyond video time, AMBOSS HY badges, and community deck tags (AnKing #HighYield, Zanki/Lightyear card density).
Our latest pass also incorporated multi-agent editorial review across every one of the 915 subtopics, with web-source verification on the 170+ cases where signals disagreed. We aim to capture both the granular exam-pattern HY topics (drug classes, microbe specifics) that community tagging catches well, and the foundational concepts (TCA cycle, glycolysis, biostats fundamentals) that community tagging structurally under-emphasizes.
We periodically re-review the designations. No source — including ours — is authoritative; this is editorial judgment grounded in source consensus. Reasonable people may disagree on edge cases.
How we map to the official NBME outline
Our areas follow how students actually study (First Aid, by discipline). On top of that, every subtopic is mapped to its single best-fit official NBME system and discipline, drawn from the official USMLE/NBME content outline. That lets us report coverage against the blueprint the exam is built from — see the by-system comparison.
The mapping is done by multi-agent review: each subtopic is assigned one system and one discipline from the fixed NBME lists, grounded against the official outline text, then validated system-by-system against the conditions the outline names. Foundational, cross-system topics (metabolism, molecular biology, general pharmacokinetics) are grouped under Multisystem.
One honest caveat: this is a best-fit placement, not a canonical NBME classification. The outline files some conditions in a different system than their clinical home (it lists vasculitides under Multisystem; we place them in Cardiovascular), and each subtopic gets a single system even when it spans several. So the by-system percentages reflect our placement choices, and the discipline split reflects how we categorize a subtopic — not the exam's weighting. As everywhere here: one data point, not gospel.
What we measure
Per deck:
- Subtopic coverage — subtopics with at least one card.
- High-yield coverage — same metric, restricted to high-yield subtopics.
- Gaps — subtopics with zero cards.
- Thin spots — subtopics with 1–3 cards.
Denominators are fixed: every deck audit compares against the same 915 (or 408) subtopics, so percentages map directly across decks.
Semantic matching, not tag counting
Community Anki decks tag their cards, but tags are organizing tools — not exam-content labels. A card teaching how preload affects stroke volume might be filed under "Frank-Starling" because that's the parent concept in the deck's hierarchy. A tag-only audit would miss the preload teaching entirely.
Our pipeline reads every card. Front and back. It identifies the specific subtopics each card actually teaches and credits each one. Cards that teach multiple concepts are credited to all of them.
That's why our coverage numbers don't match what a simple tag scan would produce. The cards in a deck are doing more work than the tag tree implies — our job is to see what they teach, not what they're filed under.
What we don't claim
These audits are useful, not authoritative. They're a way to see the shape of a deck's coverage — not a verdict on whether you'll pass.
- Not an exam blueprint. The NBME and AAMC don't publish subtopic-level lists. Our outline is the closest working approximation of what students study — the exam itself may emphasize differently.
- Coverage ≠ quality. One card on a topic counts as covered. We don't grade clarity or freshness.
- Gaps are relative to our outline. A gap means our outline lists a subtopic the deck doesn't cover. It doesn't mean the exam will test it.
- High-yield is our designation, weighted from sources students use. Reasonable people may disagree on edge cases.
- Card-level mistakes happen. Our matching reads each card and makes judgment calls. It gets most of them right, but individual cards can be miscredited or missed. Treat the headline numbers as directional, not exact.
Use the audit to spot where you might want to supplement; don't use it as the last word on a deck.
Updated as we improve
We re-audit every deck when the outline or matching improves. Numbers on this site reflect the current pipeline; older snapshots are not promised.
You can verify any audit on your own deck — upload it for a free report and you'll see the same pipeline applied to your cards, with per-subtopic attributions.