A novel Adaptive Dynamic Rating (ADR) algorithm for skill assessment in Free-For-All (FFA) arenas.
ADR-Algorithm-EN.md— English paper (authoritative)ADR-Algorithm-ZH.md— Chinese translationADR-Algorithm-EN.pdf— Compiled English PDFADR-Algorithm-ZH.pdf— Compiled Chinese PDFLICENSE— CC BY 4.0 Licensefigures/adr_comparison.png— Cumulative standardized rating comparison figurefigures/adr_comparison.csv— Simulation numeric datafigures/adr_per_step_change.png— Per-event standardized rating change figurefigures/adr_per_step_change.py— Script to regenerate the per-event figurefigures/prediction_quality.png— Illustrative prediction quality (log loss / AUC) figurefigures/prediction_quality.csv— Prediction quality numeric datafigures/prediction_quality.py— Script to regenerate the prediction quality figure
ADR is an original online rating algorithm designed for FFA arenas. It combines:
- self-adaptive learning rate
- uncertainty propagation
- streak / volatility / trend factors
- time decay based on active-player mean
- population-based tier calibration
- returning-player safeguard (temporary score pool, per-event change limit)
The algorithm is not a direct copy of Elo, Glicko-2, or TrueSkill.
This paper uses Minecraft FFA arenas as the target environment. Mechanics such as fall damage are treated as environmental non-kill deaths, which are specific to Minecraft-like sandbox games.
- Ruixi Deng (Neamyoo-dev)
- GitHub: https://github.com/Neamyoo-dev
- ORCID: https://orcid.org/0009-0004-3643-6049
- Email: superailva@gmail.com, superailva@qq.com
AI-assisted tools (DeepSeek V4 Flash 0731 and GPT 5.6 Sol) were used for mathematical formulation, LaTeX/Markdown typesetting, numerical simulation implementation, and iterative revision. The author defined the research problem, evaluated and selected candidate approaches, identified errors, and reviewed all final formulas and simulation results.
This work is licensed under the Creative Commons Attribution 4.0 International License.
Copyright (c) 2026 Ruixi Deng (Neamyoo-dev)