Paper: arXiv 2511.14408

Authors: Thomas Houweling

Abstract

Directional-change Intrinsic Time analysis has long revealed scaling laws in market microstructure, but the origin of their stability remains elusive. This article presents evidence that Intrinsic Time can be modeled as a memoryless exponential hazard process. Empirically, the proportion of directional changes to total events stabilizes near \(1 - 1/e = 0.632\), matching the probability that a Poisson process completes one mean interval. This constant provides a natural heuristic to identify scaling regimes across thresholds and supports an interpretation of market activity as a renewal process in intrinsic time.

Complexity vs Empirical Score

  • Math Complexity: 7.0/10
  • Empirical Rigor: 5.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper presents a sophisticated mathematical model using renewal theory and Poisson processes to derive a scaling constant, but it also implements a concrete empirical analysis on real cryptocurrency data with specific statistical tests and reproducibility references.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Identify origin of stable<br>scaling laws in Intrinsic Time"] --> B["Key Methodology<br>Directional-change Intrinsic Time<br>Analysis & Poisson Process Modeling"]
  B --> C["Data Inputs<br>Empirical Intrinsic Time Series<br>Directional Changes & Total Events"]
  C --> D["Computational Process<br>Measure Ratio of Directional<br>Changes to Total Events"]
  D --> E{"Analysis"}
  E --> F["Evidence: Ratio stabilizes near<br><b>1 - 1/e = 0.632</b>"]
  E --> G["Interpretation<br>Market activity modeled as<br>Memoryless Exponential Hazard Process"]
  F --> H["Key Findings/Outcomes<br>1. Identified Hidden Constant<br>2. Natural heuristic for scaling regimes<br>3. Renewal Process interpretation"]
  G --> H

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes modeling Intrinsic Time as a renewal process with a memoryless exponential hazard, yielding a constant directional‑change share of approximately 1‑e⁻¹.

Source sentence
This article presents evidence that Intrinsic Time can be modeled as a memoryless exponential hazard process. This motivates modelling intrinsic time as a renewal process.

Assumes 5

  • Inter‑renewal durations are independent and identically distributed (i.i.d.) non‑negative random variables. model
    Source sentence
    (1) A renewal process assumes that the random variables (Tn)n≥1 are i.i.d. nonnegative with cumulative distribution function F(t) = Pr(T ≤t), (2) and density f(t) = F ′(t) when it exists.
  • The inter‑renewal distribution follows an exponential law, implying a constant hazard rate. model
    Source sentence
    A constant hazard rate implies that the proportion of Directional-change events among all DcOS events is constant across thresholds within the scaling regime. Let X be the (normalized) overshoot length, and assume X ∼Exp(λ).
  • Overshoot lengths are exponential in units of the threshold δ. model
    Source sentence
    If overshoot lengths are exponential in units of the threshold δ, then the number of Overshoots per cycle follows a geometric distribution. Let X be the (normalized) overshoot length, and assume X ∼Exp(λ).
  • Directional‑change and overshoot events are correctly identified by the dcos core module. data
    Source sentence
    Directional-change and Overshoot events are extracted using the dcos core module, while event counts across thresholds are generated through the dcos fractal mod-ule.
Show 1 more
  • Event counts across thresholds are normalized by the number of ticks in the sample. statistical
    Source sentence
    5 Unless stated otherwise, event frequencies refer to counts normalized by the number of ticks in the sample.

Data

  • Dataset A (BTCUSDT) crypto-onchain public
    2022-02-08 to 2023-03-24 · BTCUSDT
    Source sentence
    Dataset Instrument Date range (YYYY–MM–DD) Days A BTCUSDT 2022-02-08 to 2023-03-24 410 B BTCUSDT 2023-03-24 to 2024-07-01 466 C ETHUSDT 2023-12-27 to 2024-08-09 227 For each dataset, we compute the Intrinsic Time representation of the midprice series using a grid of fifty logarithmically spaced Directional-change thresholds δ ∈[10−5, 1].
  • Dataset B (BTCUSDT) crypto-onchain public
    2023-03-24 to 2024-07-01 · BTCUSDT
    Source sentence
    Dataset Instrument Date range (YYYY–MM–DD) Days A BTCUSDT 2022-02-08 to 2023-03-24 410 B BTCUSDT 2023-03-24 to 2024-07-01 466 C ETHUSDT 2023-12-27 to 2024-08-09 227 For each dataset, we compute the Intrinsic Time representation of the midprice series using a grid of fifty logarithmically spaced Directional-change thresholds δ ∈[10−5, 1].
  • Dataset C (ETHUSDT) crypto-onchain public
    2023-12-27 to 2024-08-09 · ETHUSDT
    Source sentence
    Dataset Instrument Date range (YYYY–MM–DD) Days A BTCUSDT 2022-02-08 to 2023-03-24 410 B BTCUSDT 2023-03-24 to 2024-07-01 466 C ETHUSDT 2023-12-27 to 2024-08-09 227 For each dataset, we compute the Intrinsic Time representation of the midprice series using a grid of fifty logarithmically spaced Directional-change thresholds δ ∈[10−5, 1].

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • The empirical directional‑change share ¯p stabilizes within the scaling zone at about 0.61 ± 0.02, close to the theoretical constant 1‑e⁻¹ ≈ 0.632. statistical figure from elsewhere in the paper
    Source sentence
    The empirical Directional-change share ¯p demonstrates remarkable stability within each dataset’s scaling zone. For Dataset A (BTCUSDT, 2022–2023), ¯p varies only between 0.601 and 0.635 across thirteen thresholds spanning nearly two orders of magnitude, with a central tendency of 0.623 ± 0.009 (stan-dard error).
  • Event frequencies follow a power‑law scaling f(δ) ∝ δ⁻ᵝ with β ≈ ‑1.91 ± 0.02 across all datasets and event types within the scaling zone. theoretical
    Source sentence
    3.4 Summary of Findings • Power-law scaling: Within the scaling zone (10−3 ≲δ ≲10−1 in the datasets analyzed), event frequencies follow f(δ) ∝δ−β with β ≈−1.91 ± 0.02 across all datasets and event types (R2 > 0.99, p < 0.001). • Stable Directional-change share: Within the scaling zone, the empirical proba-bility ¯p stabilizes at 0.61±0.02, close to the theoretical 1−e−1 ≈0.632 benchmark. • Renewal validation: Three
  • Estimated hazard rates from normalized overshoot lengths cluster around unity (λ̂ ≈ 0.94 ± 0.03) in the scaling zone, supporting the exponential‑hazard model. statistical
    Source sentence
    Normalized overshoot lengths pro-duce estimated hazard rates ˆλ that cluster tightly around unity within the scaling zone. Dataset A yields ˆλ = 0.94 ± 0.03; Dataset B gives 0.93 ± 0.02; and Dataset C produces 0.92 ± 0.03.
  • Geometric overshoot‑count fits yield a parameter p̂_geom that mirrors the empirical directional‑change share, confirming internal consistency of the renewal framework. statistical
    Source sentence
    The geometric fit parameter ˆpgeom mir-rors ¯p almost exactly, differing by less than 10−4 in most cases. This algebraic consistency validates the internal logic of the renewal framework: the mean overshoot count per cy-cle ¯K yields a geometric parameter that reproduces the empirical event ratio.
  • Goodness‑of‑fit tests reject the geometric overshoot‑count model in the microstructure‑noise regime but show markedly higher p‑values within the scaling zone, indicating approximate renewal behavior. statistical
    Source sentence
    In the microstructure-noise regime (δ < 10−3), both the chi-squared and Kolmogorov–Smirnov p-values are effec-tively zero, correctly rejecting the geometric null where market microstructure dominates. Within the scaling zone, p-values increase dramatically—for Dataset B, pKS reaches 0.69 at δ = 0.0115, and for Dataset C, pKS = 0.98 at δ = 0.0596—yet remain below conven-tional significance thresholds for most

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 21 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

https://github.com/THouwe/IntrinsicTime — the authors' own repository

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (17)

  1. A Directional-Change Events Approach for Studying Financial Time Series (2012) — link
  2. A Modern Paradigm for Algorithmic Trading — → 2501.06032
  3. Volatility Measurement with Directional Change in Chinese Stock Market: Statistical Property and Investment Strategy (2017) — link
  4. Instantaneous Volatility Seasonality of High-Frequency Markets in Directional-Change Intrinsic Time (2019) — link
  5. Intrinsic Time Directional-Change Methodology in Higher Dimensions (2019) — link
  6. Agent-based modelling in directional-change intrinsic time (2020) — link
  7. The Theory of Intrinsic Time: Market Microstructure 2.0 (2024) — link
  8. From the bird’s eye to the microscope: A survey of new stylized facts of the intra-daily foreign exchange markets (1997) — link
and 9 more
  1. Patterns in high-frequency FX data: Discovery of 12 empirical scaling laws (2011) — link
  2. Robust inference for nondestructive one-shot device testing under step-stress model with exponential lifetimes (2023) — link
  3. Queueing Systems, Volume I: Theory (1975)
  4. Non-exponential and oscillatory decays in quantum mechanics (2014) — link
  5. The Transient Response of Damped Linear Networks with Particular Regard to Wideband Amplifiers (1948) — link
  6. Stopping at the Last Renewal up to a Terminal Time (2007)
  7. Power-law distributions in empirical data (2009) — link
  8. Autoregressive Conditional Duration: A New Model for Irregularly Spaced Transaction Data (1998) — link
  9. Econometric analysis of realized volatility and its use in estimating stochastic volatility models (2002) — link

1 of 17 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.