Project — Smurf Detection (CS2)

CS2 Smurf Detection (Faceit)

Smurfing is when a strong player jumps on a new account and farms lower lobbies. This project pulls Faceit CS2 stats and recent matches, then outputs a simple Smurf Score (0–100) with the receipts: per-metric bars and plain-English flags. It’s a lightweight API.

Goal

Build a fast system that helps spot suspicious accounts using real match data.

Fetch player + match data from Faceit
Derive a few readable signals
Combine them into a stable 0–100 score
Explain the score with flags

Architecture

API (Cloudflare Worker and FACEIT API)
Nickname → player_id, fetch stats + recent matches, compute features, return JSON.

Frontend
Calls the API and renders score, metrics, flags, and a recent matches table.

Pipeline

GET /api/analyze?player=<nickname>

1) resolve player id
2) fetch lifetime stats (FACEIT CS2)
3) fetch recent matches (last ~20)
4) fetch per-match stats (subset for features)
5) compute signals → score + flags
6) return JSON

Signals

v1 is intentionally boring: a weighted sum of interpretable signals. Boring is good.

• Low match count: fresh account suspicion
• K/D and win rate: dominance signals
• Headshot %: mechanical skill signal
• Win streak: short-term domination
• Recent spike: recent performance above lifetime baseline

Scoring

Each rule adds points. The total is clamped to 0–100. Nothing hidden, nothing mystical.

// sketch
score = 10

if matches_played < 50: score += 25
if kd >= 1.30:          score += 30
if win_rate >= 60:      score += 20
if headshot_pct >= 55:  score += 15

if winstreak >= 5:      score += 15
if recent_kd - lifetime_kd >= 0.30: score += 20

score = clamp(score, 0, 100)

Output

The API returns UI-ready JSON: headline score, raw stats, normalized metric bars, and flags that explain the result.

{
  "smurf_score": 72,
  "win_rate": 61.4,
  "kd_ratio": 1.42,
  "matches_played": 38,
  "metrics": { "headshot": {"raw":"58%","norm":85} },
  "flags": [
    {"level":"high","title":"Low Match Count","description":"Fresh account + strong stats."}
  ]
}

How to Read a High Score

Example
• Matches: 38
• Lifetime K/D: 1.42
• Recent K/D spike
• Win streak: 6

Interpretation: small sample + consistent dominance + spike = suspicious.

Validation

Sanity: legit high-level accounts shouldn’t always score high
Edge cases: low matches but average stats should stay moderate
Stability: score shouldn’t swing wildly between runs
Explainability: flags should match what a human would point at

Visuals

Since v1 is rules, visuals are about insight: score distribution, feature contributions, and recent vs lifetime comparisons.

Smurf Score — Fig 1. Account Banned from FACEIT for Smurfing vs Smurfing Score

Wrap-up

This is a practical baseline: fast, cheap, explainable. If it’s wrong, it’s wrong in a way you can actually fix.

View Source on GitHub