Cosine-Sim-Matchmaking

A proof-of-concept matchmaking system that uses cosine similarity to pair players with compatible playstyles. Players are represented as vectors across three dimensions — Damage, Skill, and Synergy — and matched based on directional similarity combined with a skill gap penalty.

This was built as a companion project to the paper "Implementasi Cosine Similarity dalam Sistem Matchmaking pada Game Online" (IF2123 Linear Algebra and Geometry, STEI ITB).

---

How It Works

Each player is encoded as a 3D vector:

[DamageScore, SkillScore, SynergyScore]

Match quality between two players is scored as:

score = cosine_similarity(a, b) / max(skill_gap, 1)

This rewards similar playstyle direction (cosine similarity) while penalizing large rating disparities.

The top 9 most compatible players are then split into ally and enemy teams in alternating order.

---

Files

File	Description
matchmaking.py	Core matchmaking logic
generatePlayers.py	Generates a random player pool
generateNames.py	Generates random gaming usernames
players.json	Player dataset (pre-generated)
gaming_usernames.txt	Username pool for generation

---

Requirements

• Python 3
• numpy

pip install numpy

---

Usage

# Generate a fresh player pool (optional)
python generatePlayers.py

# Run matchmaking for a random player
python matchmaking.py

Sample output:

Match for xXShadowBlade99Xx with rating 7.43:
(Min: 1.02, Max: 9.87, Avg: 5.31)

Ally Team:
xXShadowBlade99Xx
NightOwl42 (0.98)
...

Enemy Team:
PixelProwler (0.97)
...

---

License

MIT