How ranking works

HuntScore in one line

HuntScore = rating(0–60) + Σ weightᵢ × ln(1 + votesᵢ) × 8 + freshness(0–4)

1. Tree rating → 0–60 points

People rate tools from 1 to 5 trees. To stop one enthusiastic rating from beating a thousand steady ones, we use a Bayesian average with a prior of 3.5 trees and a prior weight of 5 ratings:

smoothed = (5 × 3.5 + Σ ratings) / (5 + n)
rating_points = (smoothed − 1) / 4 × 60

A tool with no ratings sits at 37.5 points; a tool with many 5-tree ratings approaches 60.

2. Signals → logarithmic points

Each one-tap signal has an admin-set weight (default 1). Its contribution is weight × ln(1 + votes) × 8, so the first ten votes matter much more than the ten-thousandth. Default signals:

• Must Have — you would miss it if it vanished.
• Pro Use — you rely on it for paid work.
• Personal — you use it for yourself, outside work.
• Favorite — you love it and recommend it.
• Wishlist — you want it next.
• Category-specific — e.g. Built to Last for physical tools. Admins can add, edit, weight and scope these at any time.

3. Freshness → 0–4 points

New listings receive up to 4 points that decay linearly to zero over 30 days. It surfaces newcomers briefly without letting them camp at the top.

Trend arrow

Scores are snapshotted daily. The arrow compares today with seven days ago: ▲ gained, ▼ lost, — steady (within ±0.5).

Rank

Rank is a tool’s position by HuntScore on the whole board, and it belongs to the tool: filtering by category, kind or pricing, searching, or sorting the table by another column never changes the number next to it. Ties are broken by number of hunters, then name.

What can never move a rank

• Money — no paid placement, ever.
• The submitter — you cannot vote on or review your own tool.
• Admins — we can categorise, hide spam and fix typos, not edit scores.
• Bots — every write needs a Google account, one vote per person per signal, rate limits on everything, and an audit log we review.

Same metrics for everyone

A bullock cart and a browser extension are ranked by the same six core signals, so you never need to learn a new scale. Category-specific metrics only compare like with like. See the worked example for a full walkthrough.