CIPHER v1 DEPLOYED
Player Performance · Gradient Boosting (regression)

Predicts individual player output for upcoming matches.

0.307
held-out test set

About this Model

CIPHER predicts individual player performance metrics for upcoming matches: Rating, ADR, KAST, and Impact Score. Trained on 120,000+ player-map performances from V4 parsed demos, CIPHER uses each player's rolling stats (last 15 maps), opponent quality, map specialization, and team context to generate per-player projections. R²=0.307 on held-out data means CIPHER explains ~31% of individual performance variance — strong for a fundamentally noisy domain.

Top Feature Drivers

Player rolling Rating (15-map) 31%

Recent HLTV-style rating trend

ADR rolling average 22%

Average damage per round history

Map specialization score 17%

Player's historical performance on this specific map

Opponent quality (Elo) 14%

Facing stronger opponents downgrades projected stats

Team context (lineup CQR mean) 9%

Better teammates = slightly inflated individual numbers

LAN/Online indicator 7%

Some players significantly better/worse on LAN

Feature importances reflect normalized split gain across the training set. Values are approximate — the model uses all features simultaneously.

How to Interpret

CIPHER projected Rating is a regression estimate, not a hard prediction. R²=0.307 means it captures about 31% of real performance variance — useful for identifying systematic over/underperformers but not precise enough for single-match confidence. Players projected >0.15 Rating above their opponent are meaningful edges. CIPHER is most reliable for stable rosters with 10+ map history.

No settled predictions recorded for this model yet.

Live Performance

No live predictions recorded yet.

Selector Role

Player intelligence — feeds player projections into team comparison and TITAN.

Version History

v1 2026-Q2

Launch; fixed opening_kill_rate source (demo_kill_events); R² 0.307

All Models