Back to eSports Models
CS2 Moneyline Model

CS2 Moneyline Documentation

Complete technical reference for the Counter-Strike 2 Moneyline prediction model—methodology, edge detection, and operational details.

Frequently Asked Questions

How are team ratings calculated?

Elo-style ratings that update based on match results, opponent strength, and margin of victory. Ratings decay over time to account for roster changes.

Why tournament tier matters?

S-Tier events (Majors, top LANs) carry more weight than online qualifiers. Teams perform differently under high-stakes conditions.

How handle roster changes?

Significant roster changes trigger rating adjustments. New rosters start with blended ratings based on individual player histories.

What is Bo1 vs Bo3 adjustment?

Best-of-1 matches have higher variance than Bo3/Bo5. We adjust win probabilities to account for format-specific upset potential.

Why LAN vs online distinction?

Some teams overperform on LAN due to crowd energy and bootcamp preparation. Others struggle with travel and pressure. We model this explicitly.

What events are covered?

Major tournaments, top-tier LANs, and significant online leagues. Lower-tier matches are excluded due to limited data reliability.

How is data quality scored?

Each pick includes a data quality score based on team history depth, recent form stability, and roster continuity.

What does edge percentage mean?

The gap between our model probability and the implied market probability. Higher edge = larger detected mispricing.

How validate model accuracy?

Long-term ROI tracking, CLV analysis, and performance by tournament tier. All metrics are publicly displayed.

Methodology

Team Rating System

Each team maintains a dynamic Elo-style rating that updates based on match results. The system accounts for:

Opponent Quality: Defeating a top-10 team yields larger rating gains than beating an unranked squad
Map Score Margin: 2-0 sweeps carry more weight than close 2-1 series
Recency Weighting: Recent matches matter more, accounting for form changes and meta shifts
Roster Stability: Teams with recent roster changes have ratings adjusted for uncertainty

Tournament Tier System

Not all matches carry equal significance. We weight results by tournament importance:

S-Tier
Majors, top-tier LANs, premier league finals. Maximum weight for rating updates.
A-Tier
Regional LANs, major online leagues. High weight for competitive matches.
B-Tier & Below
Qualifiers, minor online cups. Reduced weight due to motivation variance.

Match Format Adjustments

Different formats have different variance profiles:

  • Best-of-1: High variance, map pick/ban crucial, upset potential elevated
  • Best-of-3: Standard format, balanced skill expression, most reliable predictions
  • Best-of-5: Low variance, favors deeper map pools, skill differential magnified

LAN vs. Online

Venue type significantly affects performance:

  • LAN advantages: Some teams thrive under crowd pressure and bootcamp preparation
  • Online advantages: Home comfort, no travel fatigue, familiar setup
  • Ping considerations: Cross-regional online matches introduce latency factors

Edge Detection & Filtering

Pick Selection Criteria

Strict filtering ensures only high-quality edges are released:

Edge Threshold
Model probability must exceed implied market probability by a meaningful margin
Data Quality Score
Sufficient match history and roster stability required for confident projections
Tournament Tier Filter
Low-tier matches with unreliable data are excluded from pick generation
Odds Floor
Extreme favorites beyond certain thresholds are excluded due to unfavorable risk/reward

Data Quality Framework

Quality Scoring System

Each pick includes a data quality score (0-100) indicating projection confidence:

90+ (High Quality)
Established teams, stable rosters, extensive match history, reliable predictions
70-89 (Medium Quality)
Good data available but some uncertainty factors present (recent roster change, limited H2H)
Below 70 (Lower Quality)
New team, major roster changes, or limited match history. Higher uncertainty.

Pick Lock Timing

8-Hour Lock Window

CS2 picks are locked and published 8 hours before match start. This timing ensures:

Time to Act
Users have sufficient time to review picks and place bets before lines move significantly
Odds Stability
8 hours out, odds are generally stable but still offer value before sharp money arrives
CLV Tracking
Locked odds are captured at publish time, enabling accurate closing line value measurement
Global Coverage
Works for all timezones—EU, NA, and Asia matches all follow the same 8-hour rule

Performance Tracking

Validation Metrics

All picks are tracked with full transparency:

Win Rate
Overall and by confidence tier, tracked against expected hit rates
ROI
Return on investment using flat unit sizing by tier
Closing Line Value
How odds move after we lock—consistent CLV validates edge detection
View Live Picks