Skip to content
EH

Command Palette

Search for a command to run...

One-page scoring brief

A clear, presentation-ready view of how ErrataHub scores books, credibility, and contributors

This page is designed for demos and internal presentations. It shows the exact formulas, the factors behind them, and worked examples so the scoring logic is easy to explain without jumping between screens.

Book score

0 to 10 weighted rating

Claim status, metadata, cover, followers, issue health, and resolution speed.

Credibility

0 to 100 closure rate

Publisher and author credibility use total issues versus open issues.

Contributor trust

Leaderboard signal

Trust score combines reputation points and confirmed issue outcomes.

Methodology Overview

How the scoring model is structured

Book ratings are a weighted 0 to 10 score. Publisher and author credibility are 0 to 100 closure-rate scores. Contributor trust uses reputation points plus confirmed issues.

Everything is recomputed from auditable claims, issue lifecycle states, and reputation events so the numbers stay explainable.

Books

Weighted rating out of 10

Requires at least one approved claim and blends metadata, cover quality, followers, issue health, resolution rate, and speed.

Publisher / Author

Closure rate out of 100

Calculated from total issues vs open issues after a profile is verified. Unverified profiles show N/A instead of a fake 100.

Contributors

Trust score plus reputation

Leaderboard is sorted by trust score. Profile pages also show raw reputation points and a points-based rank.

Book Rating Formula

Exact book scoring logic

Eligibility

At least 1 approved claim

Final score

round(clamp(10 * sum(factor * weight), 0, 10), 1)

Label thresholds

9+ Exceptional8+ Strong7+ Solid6+ Fair5+ Fragile
FactorExact calculation
Claim1 if an approved claim exists, otherwise 0
Metadatapresent_fields / 6 using author, publisher, year, language, description, and ISBN
Covermin(1, front_bonus + back_bonus), where front is 0.7 or 0.55, and back adds 0.3
Followersmin(1, log(1 + followers) / log(251))
Issue healthmax(0, 1 - min(1, issueLoad / (totalIssues * 1.05)))
Resolution ratemin(1, (resolvedClosed + 0.5 * resolvedOpen) / totalIssues)
Recognition speedspeedFactor(avgRecognitionDays, 14, 0.55)
Resolution speedspeedFactor(avgResolutionDays, 30, 0.55)
ModeClaimMetadataCoverFollowersIssue healthResolutionRecognitionResolution
strict0.050.100.050.050.300.200.150.10
balanced0.050.150.100.100.250.150.100.10
generous0.050.200.150.150.180.120.080.07
Issue statusWeightIssue statusWeight
suggested1.00resolved_open0.58
unconfirmed0.92resolved_rejected0.35
confirmed0.85resolved_closed0.25
rejected0.10default0.55

Presentation Example

Worked book example

Example inputs: approved claim = 1, metadata = 5/6, cover = 1.0, followers = 250, issue health = 0.60, resolution rate = 0.50, recognition speed = 0.55, resolution speed = 0.55.

FactorValueWeightContribution
Claim1.0000.050.050
Metadata0.8330.150.125
Cover1.0000.100.100
Followers1.0000.100.100
Issue health0.6000.250.150
Resolution rate0.5000.150.075
Recognition speed0.5500.100.055
Resolution speed0.5500.100.055
TotalWeighted sum = 0.710Score = 7.1 / 10

Presentation line: this book would land at 7.1/10, which maps to Solid.

Credibility

Publisher and author credibility

Exact formula

verified ? (totalIssues == 0 ? 100 : round(((totalIssues - openIssues) / totalIssues) * 100)) : N/A

Meaning

Closure rate only, after verification. Unverified profiles stay N/A.
ExampleTotal issuesOpen issuesResolved issuesCredibility
Author or publisher123975
Verified profile, no issues yet000100

Contributor Leaderboard

Trust and reputation

Leaderboard order

trust_score DESC, id DESC

Trust formula

min(1000, points * 2 + approvedIssues * 10)

Trust levels

250+ trusted100+ established30+ active
ExamplePointsApproved issuesTrust scoreLevel
Strong contributor1208320trusted
Profile fallback example1147 confirmed...n/aDisplay points only

Important distinction: the contributor leaderboard uses trust score, while the reputation leaderboard sorts by raw points.

Ranking Rule

Why authors and publishers stay separate

We do not merge author and publisher credibility into one leaderboard. Publishers usually have many more books than authors, so a shared rank would bias the results and make the comparison unfair.

The safe setup is two separate verified leaderboards: authors compete with authors, and publishers compete with publishers.