ML SCORING METHOD

A probabilistic decision layer that scores every incoming trade before execution. It estimates trade quality from live market features, compares the setup against model thresholds, and helps route capital toward higher-conviction opportunities.

Block 02
A PROBABILITY ENGINE FOR TRADE QUALITY IN REAL TIME

The ML Scoring Method sits inside the live trading path and evaluates each new signal using a trained model family built from historical trade behavior.

01
Probability-Based SelectionEvery trade is transformed into a feature vector and scored by the model. The output is not a vague label but a measurable estimate of expected trade quality, including win probability and confidence relative to the active threshold.
02
Policy-Aware DecisioningThe score is interpreted through live policy states such as COLD, WARM, and MATURE. That allows the system to be more cautious with low-sample models, stricter with mature models, and explicit about when a trade deserves approval or rejection.
03
Backoff and Coverage LogicWhen a highly specific model does not have enough support, the system can fall back to pooled or global coverage. That preserves live usability while still keeping the scoring path grounded in the best available evidence.
ML Scoring
Active — v3.1
TOTAL BUY-BACKS

Real-time cumulative performance of the ML Scoring Method. Every data point reflects the model's live impact on trade selection quality and downstream outcomes.

Total Buy-Backs
ML Scoring Method — Cumulative P&L
$0.00
$0.00
Created
Budget
Buy-backs
Upcoming buy
CONTINUOUSLY TRAINED, FULLY MEASURABLE

The ML Scoring Method improves through accumulated trade history, model monitoring, and live policy control. It is built to evolve without becoming opaque, so every score can still be traced back to model outputs, thresholds, and decision state.

Feature-Rich Inputs

The model can learn from momentum, structure, volatility, macro signals, projection edges, and other live context features that are difficult to express cleanly in static rule form.

Threshold Discipline

Scoring alone is not enough. Each output is judged against active thresholds, precision floors, margin rules, and policy states so the system remains operationally disciplined.

Adaptive Maturity Model

As more validated trade data becomes available, the method can transition from broad fallback coverage toward more specialized model decisions with stronger confidence.

Audit-Ready Outputs

Every live decision can be logged with pwin, threshold, margin, final score, policy state, and backoff level, making it easier to monitor real behavior and tighten the system over time.