SYNERGY v2 DEPLOYED
Team Chemistry & Roster · Gradient Boosting

Some lineups are greater than the sum of their parts.

AUC
0.639
held-out test set

About this Model

SYNERGY models team chemistry and roster composition effects on match outcomes. AUC 0.639 after a critical v1→v2 fix: team Elo was completely missing in v1 and an O(N²) roster-build bug was fixed. SYNERGY predicts how well a specific lineup of 5 players performs together, factoring in pairwise co-occurrence win rates, role balance, and experience playing together. The pairwise chemistry matrix identifies which player combinations are statistically strongest.

Top Feature Drivers

Team Elo differential 32%

Foundation quality signal — was missing in v1

Pairwise co-win rate (top pair) 24%

Best player pair's historical win rate together

Roster novelty score 19%

How new is this exact lineup? New = less data = regression

Role balance index 15%

Entry/AWPer/Support/Lurk balance score

Mean games played together 10%

Lineup experience — more games = stronger signal

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

How to Interpret

SYNERGY chemistry score is a 0–100 percentile versus all CS2 pro teams. High SYNERGY (>70) with high player individual CQR means a well-functioning team where the parts amplify each other. Low SYNERGY despite high individual CQR scores signals a roster that hasn't gelled — perhaps a recent signing or role conflict. Check the pairwise matrix for the specific weak links.

No settled predictions recorded for this model yet.

Live Performance

No live predictions recorded yet.

Selector Role

Sub-model — roster chemistry feeds into TITAN ensemble and the team comparison tool.

Version History

v1 2026-Q1

AUC 0.53 — missing team Elo entirely; O(N²) roster-build bug

v2 2026-Q2

Fixed Elo + roster build; AUC 0.639 (+0.11)

All Models

PISTOL
SYNERGY (current)