System Architecture v5.1.0
Live API Feed Active

QUANTITATIVE
SUPERIORITY.

The SOAR Method transitions from raw data ingestion via cbbd and cfbd SDKs to a specialized ensemble execution pipeline. We isolate mathematical edges by processing 13+ efficiency variables per matchup.

Follow @TheRealSOAR
01

Multi-Threaded Ingestion

College Basketball

Library: cbbd==1.22.0

Our system utilizes concurrent.futures to perform parallel sweeps of 365+ D1 programs. We ingest real-time game-by-team results to calculate recursive SRS and Adjusted Net Ratings.

eFG%TOV%ORB%FTA Rate

College Football

Library: cfbd==5.13.2

The football engine flattens complex API responses via pd.json_normalize. We synthesize Team Combined metrics, merging recruiting talent grades with PPA (Predicted Points Added) efficiency data.

PPA EfficiencyTalent RankSOS WeightReturning Prod
02

The Edge Algorithm

RANDOM FOREST
REGRESSION ENGINE

By processing 800+ decision trees per game, our model identifies the "True Margin." We look beyond the box score to analyze high-dimensional interactions between team efficiency deltas.

input_features = [# 13 Total
"edge_AdjNRtg", "edge_eFG", "edge_TOV_pct",
"edge_ORB_pct", "edge_FTA_rate", "edge_SRS" ...
]

Ensemble Methodology

We use Scikit-Learn ensembles to prevent overfitting on single-game statistical noise.

Vegas Delta Calculation

The system identifies the 'spread-gap'—the delta between market lines and our projected margin.

Train/Test Validation

Models are validated against a 20% test split of historical 2024-2025 season data.

03

Platinum Verification

Before a prediction is authorized as a Platinum Play, it must survive a secondary Logistic Regression confidence check.

Back-Tested Threshold

65%+

Confidence Interval

95% CI

Computational Superiority

UNMATCHED
TRANSPARENCY.

Connect on X

Daily Recalibration | Python 3.12

SOAR Analytics uses proprietary datasets and scikit-learn ensemble methods. Models are retrained every 24 hours via automated CRON jobs to ensure zero-day data relevance.