IntrinsicTime 0.1.3

Utilities for intrinsic time decomposition and fractal event analysis

IntrinsicTime

Utilities for decomposing intrinsic time events and analyzing fractal scaling behavior in price or signal data.

This package provides:

• dcos_core: A Directional Change and Overshoot (DcOS) event detector.
• dcos_fractal: Tools for fractal scaling and multi-threshold analysis.
• dcos_plot: Plotly visualization for interactive fractal plots.

---

Installation

From PyPi

pip install IntrinsicTime

From GitHub

pip install git+https://github.com/THouwe/IntrinsicTime.git

Local Install

git clone https://github.com/THouwe/IntrinsicTime.git
cd IntrinsicTime
pip install -e .

Dependencies

See requirements.txt.

---

Overview

DcOS events are first‑passage moves of size δ in log space

Robust power‑law behaviour can be observed for many phenomena. For instance, Intrinsic Time event density of BTCUSDT price ticks scale linearly with DcOS δ threshold in log space (cit), consistently with first‑passage theory (cit) plus market microstructure. However, this is the case only within a given range of δ thresholds, as the power law may brake at 'extremely low' or 'extremely high' δs.

For small δs, issues relate to microstructure noise (tick size, latency, and irregular sampling inject high‑frequency mean reversion. This raises event frequency toward a ceiling and flattens the log–log curve) and discretization limits (time and sample - e.g., price - granularity cap how many distinct first‑passage events you can observe).

For large δs, issues relate to data scarcity: too few events reduce fit quality and increase variance.

Fractal brakepoint formalization

Compute local slopes (b(\delta)) with a sliding window (w) on ((\log \delta,\log f)). Mark the smallest δ where either (R^2 < R^2_{\min}) or (|\Delta b|) exceeds one standard error from adjacent windows. Your windowed method will return something close to δ ≈ 6e‑4 for this dataset if your visual read is correct.

Example Usage

import numpy as np
import pandas as pd
from IntrinsicTime import DcOS_fractal

# Example input DataFrame
df = pd.DataFrame({
    "Timestamp": range(1000),
    "Price": 100 + np.cumsum(np.random.randn(1000))
})

# Initialize and run
analyzer = DcOS_fractal(debugMode=True)
results, ranges = analyzer.run(df)

# Display results
print(results.head())
print(ranges)