QTIP — Quantitative Trading Intelligence Platform

Owner: Richard Patterson © 2026 — De-ASI-INTERFACE Version: 2.0.0 | Status: Production-Hardened & Fully Tested

---

Regime Classifier v2

Hidden Markov Model (HMM)-based market regime classifier that labels price action into four states:

Regime	Description
BULL_TREND	Positive drift, low-moderate vol
BEAR_TREND	Negative drift, low-moderate vol
RANGING	Near-zero drift, low vol
HIGH_VOL_BREAKOUT	Any drift, high vol

---

Quick Start

# 1. Clone
git clone https://github.com/De-ASI-INTERFACE/qtip.git
cd qtip

# 2. Install in editable mode
pip install -e .

# 3. Install dev deps
pip install -e ".[dev]"

# 4. Run full test suite with 90% coverage gate
pytest

---

Usage

import numpy as np
from qtip import RegimeClassifierV2, Regime

# Fit on historical close prices
prices = np.array([...])  # numpy array of close prices
clf = RegimeClassifierV2()
clf.fit(prices)

# Predict regime per bar
regimes = clf.predict(prices)

# Current regime
current = clf.current_regime(prices)
print(current)  # Regime.BULL_TREND

# Posterior probabilities for latest bar
probs = clf.current_proba(prices)
# {'BULL_TREND': 0.821, 'BEAR_TREND': 0.04, 'RANGING': 0.11, 'HIGH_VOL_BREAKOUT': 0.029}

# HMM transition matrix
tm = clf.transition_matrix()

# Log-likelihood score
ll = clf.log_likelihood(prices)

---

Architecture

qtip/
├── __init__.py
└── regime_classifier_v2.py
    ├── Regime (enum)
    ├── RegimeConfig (dataclass)
    ├── build_feature_matrix()
    └── RegimeClassifierV2
        ├── .fit(prices)
        ├── .predict(prices)
        ├── .predict_proba(prices)
        ├── .current_regime(prices)
        ├── .current_proba(prices)
        ├── .transition_matrix()
        └── .log_likelihood(prices)

tests/
└── test_regime_classifier_v2.py   # 35 tests, 90%+ coverage gate

.github/workflows/
└── regime-ci.yml                  # lint → bandit → pytest

pyproject.toml                     # editable install + pytest config

---

Features

• 4-state Gaussian HMM with configurable covariance (full / diag / tied / spherical)
• Automatic regime labelling by mean return + vol signature — no manual state assignment
• Z-scored feature matrix: log returns, realised vol, momentum (ROC), trend strength
• Posterior state probabilities per bar
• Full transition matrix with regime labels
• Log-likelihood scoring for model comparison
• 90%+ branch coverage enforced in CI

---

Dependencies

Package	Min Version	Purpose
hmmlearn	0.3.0	Gaussian HMM engine
numpy	1.26.0	Feature computation
scipy	1.12.0	Z-score normalisation
pytest	8.0.0	Test runner (dev)
pytest-cov	5.0.0	Coverage gate (dev)

---

Roadmap

• Regime Classifier v2 (HMM-based state transitions)
• Full signal ingestion pipeline (Binance, Solana, on-chain)
• Amplitude scorer with backtested weight calibration
• REST API output interface for downstream bot consumption
• Docker Compose deployment with Prometheus integration

---

License

Proprietary — Richard Patterson © 2026. All rights reserved.