CounterQuant header

Counter-Strike 2 Intelligence

Real-time CS2 match data, event highlights and AI predictions.

UK-based analytics built for serious CS2 teams, content creators and quant traders. Live scores, historic match features, and data-driven insight from an independent Counter-Strike research platform.

Live feed

Updates every 30 seconds

Match scores, map state and predictions refresh instantly.

Trusted data

HLTV, Liquipedia, GRID

Sourced, normalised, and exposed via API.

AI Predictions

CS2-APEX v1.5

CQ-GESUS-CS2-APEX v1.5 — calibrated win probabilities + SHAP explainability on 118,000+ match history. AUC 0.6928, isotonic calibration.

View Plans →
Live Artifacts

Model Intelligence Suite

● 2 models
T1 matches → CQ-GESUS-APEX-T1 v1 | T2/T3 matches → CQ-GESUS-APEX v1.5
CQ-GESUS
CQ-GESUS-APEX-T1 v1
Tier 1 only (dedicated)
LIVE
Test AUC
0.69
Val AUC
0.7353
Brier Score
0.2221
Features
450
Train N
4862
Test N
2022
Trained 2026-05-27
CQ-GESUS
CQ-GESUS-APEX-T1T2 v2
T1 + T2 (Optuna-tuned)
LIVE
Test AUC
0.6329
Val AUC
0.6597
Brier Score
0.2363
Features
462
Train N
7497
Test N
5713
Trained 2026-05-27
T1 AUC 0.6853 | ECE 0.0191 | 80 Optuna trials
Step 2 of 3

Model Training — CQ-GESUS-CS2-APEX v1.5

auto-refresh 15s
APEX v2 Retrain
DS (176.9.51.147)
COMPLETE
started 2026-05-27T10:47 UTC · 290 features · 31934 train rows
Complete
Benchmark (no tuning): Val 0.6856 Test 0.7062
Optuna trials — 160/160 100%
r1 80/80
best loss 0.63367
r2 50/50
best loss 0.63364
r3 30/30
best loss 0.63364
Val AUC (cal)0.69212
Test AUC (cal)0.7022
Calibrationisotonic
TierBELOW_MIN
Step 1 of 3

Data Pipeline

auto-refresh 30s
118872
Total matches
Jan 2012 – present
25999
Matches 2024–now
Primary training target
36963
/ 29693
Demos parsed
100% complete
48435
Feature vectors
49894 demo aggregates
Scraping: 118,020 matches (Jan 2012 – present) — complete & live.
Demos: 36963 parsed, running continuously. 2024+ first, then backfill.
Feature extraction: 509-feature pipeline scaling. Training at ~80% 2024 coverage.
CQ-GESUS-CS2-APEX v1.5: Live (AUC 0.6928 isotonic)

Upcoming Predictions

Elo Baseline v0 — win probabilities for next 72 hours

Elo Baseline v0
📅
No upcoming matches in next 72 hours

Predictions will appear here as matches are scheduled.

CounterQuant Intelligence

CQ-GESUS Model Suite

6 models 509 max features

GESUS — Game Environment Statistical Understanding System — six purpose-built models from general match outcome to full-signal market arbitrage.

CQ-GESUS
APEX
v1 · General Match Outcome
▶ Running

"The pinnacle of match prediction"

The flagship all-purpose match outcome predictor. Built on 50+ features including Elo, form, roster changes, and event tier. Designed as your reliable daily driver for general betting and predictions.

50+ features Hobbyist+
Feature extraction in progress · expected Q3 2026
CQ-GESUS
PHANTOM
v1 · Economy Specialist
Planned

"Invisible efficiency. Unseen edges."

Economy specialist that reveals invisible edges in team financial play. Analyzes 120+ demo-derived features like eco win rates, force-buy discipline, trade efficiency, and clutch economy management.

120+ features Pro+
Awaiting demo feature pipeline scale
CQ-GESUS
ORACLE
v1 · Deep Stats Ensemble
Planned

"Sees the unseen through 509 dimensions"

The deep-stats powerhouse. A 509-feature LightGBM ensemble that processes everything from KAST, utility impact, multi-kills, and advanced player synergies to deliver highly accurate predictions.

509 features Analyst+
Q3 2026
CQ-GESUS
NEXUS
v1 · Full Signal + Market
Planned

"Connects all data with live market reality"

The ultimate market-aware model. Combines all 509 demo/stats features with live Polymarket odds for real-time calibration, uncovering mispricings and delivering the strongest calibrated signals.

509 + market Enterprise
Q4 2026
CQ-GESUS
MIRAGE
v1 · Map-Specific Predictor
Planned

"Master of every map. Controls perception."

Map-specific prediction expert. Delivers precise per-map win probabilities by analyzing map pool depth, CT/T-side tendencies, economy patterns, and historical map performance.

Map-specific Analyst+
Q3 2026
CQ-GESUS
TITAN
v1 · Ultimate All-Rounder
Planned

"The complete battlefield dominator"

The complete battlefield dominator. An elite hybrid that blends the strongest signals from all GESUS models into one ultra-powerful all-round predictor for maximum accuracy.

All signals Enterprise
2027

Who Gets What

Access tiers unlock progressively more powerful models

Model Free Hobbyist Pro Analyst Enterprise
APEXGeneral Match Outcome
PHANTOMEconomy Specialist
ORACLEDeep Stats Ensemble
MIRAGEMap-Specific Predictor
NEXUSFull Signal + Market
TITANUltimate All-Rounder
Confidence score basic full full full
Market edge vs Polymarket
API access

Simulated Track Record

Awaiting First Model

Paper bets placed before every match. No real money — a public log to prove accuracy before you commit.

🎲
No simulated bets yet

Every prediction from CS2-APEX v1.5 is logged here as a paper bet. Win rate, ROI, P&L — fully public, fully auditable.

Expected Q3 2026

Latest Model Progress

live
CQ-GESUS-APEX apex_v2_retrain
✓ Trained
Val AUC 0.66709
Test AUC
Brier (val) 0.22849 ✗
Tier BELOW_MIN
5 training runs ▾
20260524_200123 below_min AUC=0.66709
20260523_184801 below_min AUC=0.66132
20260523_181557 below_min AUC=0.66183
20260523_102128 below_min AUC=0.65009
20260523_051006 below_min AUC=0.63391
CQ-GESUS- PHANTOM
planned
CQ-GESUS- ORACLE
planned
CQ-GESUS- NEXUS
planned
CQ-GESUS- MIRAGE
planned
CQ-GESUS- TITAN
planned

Launch Thresholds

MinimumAUC≥0.73 · LL≤0.62
TargetAUC≥0.77 · LL≤0.58
StrongAUC≥0.79 · LL≤0.56
CQ-GESUS- APEX
Feature Intelligence

Global Feature Importance

Elo advantage 0.0260
Elo implied probability 0.0220
Elo win probability 0.0190
Schedule difficulty edge 0.0120
Avg opponent quality diff 0.0100
Consensus Win Prob Team1 0.0080
Days since last match (T1) 0.0070
Glicko2 rating (T2) 0.0070
Elo (T1) 0.0060
Glicko2 rating (T1) 0.0060
Elo (T2) 0.0060
Days since last match (T2) 0.0060
Elo freshness (T1) 0.0060
Rest days (T1) 0.0060
Team1 Schedule Difficulty 90D 0.0060

Mean |SHAP| on validation set — higher = stronger predictor

Training Data

Total matches118872
From 2024+25999
Demos parsed36963
Feature vectors48435
View Pipeline Dashboard →