Skip to main content

Methodology

The mathematical foundation behind RankHub rankings

Glicko-2 Rating System

RankHub uses the Glicko-2 rating system, developed by Professor Mark Glickman at Boston University. Unlike simple ELO systems, Glicko-2 provides two key metrics for each item:

Rating (μ)

The estimated skill level of the item. Higher is better. Initial rating: 1500

Rating Deviation (RD)

Uncertainty in the rating. Lower RD = higher confidence. Initial RD: 350

Volatility (σ)

A third metric tracking how much an item's rating tends to fluctuate. Higher volatility indicates unpredictable performance.

Why Not ELO?

Traditional ELO systems have a fundamental flaw: they treat all ratings as equally certain. A new item with 5 votes and rating 1500 is treated the same as an established item with 500 votes at rating 1500.

Glicko-2 solves this by tracking rating deviation. New items start with high RD (uncertainty), which decreases as more votes are collected. This allows the system to make more conservative updates for established items while allowing rapid discovery for new items.

Rating Scale

Rating
Classification
Percentile
Description
> 1800
S-Tier
Top 1%
Exceptional
1600-1800
A-Tier
Top 10%
Excellent
1400-1600
B-Tier
Top 30%
Good
1200-1400
C-Tier
Top 60%
Average
< 1200
D-Tier
Bottom 40%
Below Average

Pairwise Comparison

Voting uses a pairwise comparison model. Voters are presented with exactly two items and must choose which is better. This simplifies the voting task and produces more reliable data than asking voters to rate items on an absolute scale.

The matchup selection algorithm prioritizes:

  • Items with high rating deviation (to reduce uncertainty)
  • Avoiding repeated pairings
  • Similar-rated items for competitive matchups

Bot Protection

To maintain ranking integrity, RankHub employs multiple anti-gaming measures:

Cloudflare Turnstile

Human verification before each vote is accepted.

Rate Limiting

Maximum 30 votes per minute per IP address.

Matchup Expiration

Each matchup expires after 5 minutes.

Idempotent Voting

Duplicate votes are rejected via matchup ID.

Reference

Glickman, M. E. (2013). Example of the Glicko-2 Rating System.

http://www.glicko.net/glicko/glicko2.pdf →

Ready to Vote?

Start contributing to the rankings today.

Browse Directories →