Rating methodology

An ordinary Grimoire score is available only for a completed fantasy novel or novella, tied to an identified edition. The reviewer must attest to the completed read and rubric scope, record confidence, explain who the book may and may not suit, provide evidence for every component, and clear all conflict and disclosure checks.

• Did not finish (DNF): record the stopping point and reason, publish only as an explicitly unscored discussion, and exclude it from score aggregates.
• Advance reader copy (ARC): disclose the copy. A score based on an ARC is provisional until the final edition is checked; provisional ARC scores are excluded from aggregates.
• Everlore-owned work: do not assign an ordinary in-house score.
• Hard conflict: editing, representing, designing, financially participating in, or being unable to assess the work impartially disqualifies the reviewer.
• Manageable relationship: friendship or a prior publisher or professional relationship must be resolved by the publisher and disclosed publicly before a score can proceed.

Each component is scored from 1.00 to 10.00 in quarter-point steps. The component score is an evidence-backed editorial judgment about the book’s execution and aims, never a judgment of the author as a person.

• Prose and voice — 20%: how effectively and deliberately the language creates the intended reading experience.
• Character and relationships — 20%: whether characters are dimensional, motivated, and consequential to one another.
• World and imagination — 20%: whether the fantasy world is coherent, specific, immersive, and meaningfully integrated.
• Narrative design and pacing — 20%: how well the narrative builds, moves, reveals, and delivers payoff for its aims.
• Thematic and emotional impact — 15%: what the book examines, makes the reader feel, and leaves behind.
• Distinctiveness and ambition — 5%: what the book attempts, risks, or contributes beyond familiar execution.

The CMS calculates the raw score as the sum of each component score multiplied by its weight. It then rounds that single weighted result to two decimal places using round-half-up. An editor cannot type or override the overall number.

Example: component scores of 8.25, 8.25, 9.00, 8.00, 8.50, and 8.25 produce (8.25 × .20) + (8.25 × .20) + (9.00 × .20) + (8.00 × .20) + (8.50 × .15) + (8.25 × .05) = 8.3875, displayed as 8.39.

The hundredths place comes from the weighted formula; it is not a claim that 8.39 is meaningfully better than 8.38. Differences under 0.25 should usually be treated as effectively tied and are never used as a ranking tiebreaker.

Bands describe the displayed score. They are interpretation aids, not quotas or promises that every reviewer will produce the same distribution.

• 9.50–10.00 — Masterwork: rare, field-defining execution with exceptional control and lasting impact.
• 9.00–9.49 — Exceptional: outstanding and memorable; among the strongest books in its area.
• 8.00–8.99 — Excellent: highly accomplished, distinctive, and easy to recommend to its intended reader.
• 7.00–7.99 — Strong: clearly successful with meaningful strengths and manageable limitations.
• 6.00–6.99 — Good: worthwhile for the right reader, though notable weaknesses affect the whole.
• 5.00–5.99 — Mixed: strengths and weaknesses are roughly balanced; fit matters substantially.
• 4.00–4.99 — Limited: interesting elements exist, but the execution is more unsuccessful than successful.
• 3.00–3.99 — Poor: major craft or coherence problems prevent the book from fulfilling its aims.
• 2.00–2.99 — Very poor: severe, pervasive failures with little successful execution.
• 1.00–1.99 — Fundamentally unsuccessful: fails most of its core promises or is not responsibly publishable as presented.

Before any scored review can publish, every active scored-book reviewer must independently score at least three real anchor books spanning different fantasy subgenres and likely score bands. Each reviewer must first accept the current review standards and conflict policy.

The publisher compares the evidence, component choices, overall spread, component spread, and reviewer distributions. A round passes only when every required submission exists, the overall score spread for each anchor is no more than 0.75, and every component spread is no more than 1.00. The database blocks score publication while calibration is pending, failed, suspended, or tied to another rubric version.

Calibration repeats for a new scored-book reviewer, a changed reviewer roster, a new rubric version, and at least quarterly while the review desk is active. A disagreement outside the thresholds leads to discussion and a new independent round; the publisher does not edit reviewers into artificial agreement.

A public aggregate requires at least three current, public, high-confidence reviews of the same work under the same rubric version. Only non-ARC or final-edition-verified scores qualify. The CMS averages the stored raw weighted scores and rounds once to two decimals.

The aggregate displays “Critics split” when the highest and lowest qualifying raw scores differ by 1.50 or more. Aggregate numbers never mix reader ratings, different rubric versions, DNF discussions, provisional ARC scores, low-confidence reviews, or disqualified conflicts.

Every score change creates a new immutable revision linked to the prior one. If a published total changes, the public review receives a visible correction note; the old revision remains in history. Only Travis DiPerna, as publisher and final methodology approver, may authorize the revision.

Scores are structured editorial judgments, not universal truths. Read the evidence, fit notes, edition facts, confidence, and disclosures alongside the number. Review copies, affiliate availability, advertising, friendship, access, or sales potential never raise a score.