Paper: arXiv 2611.25860

Authors: Siddharta Laloux, Bhargavi Srinivasan

Abstract

We study the intrinsic geometry of the Nikkei-225 through the discrete Ollivier-Ricci curvature (ORC) of its empirical correlation graph, using the graph Ricci flow and surgery methodology of arXiv:2510.15942. The 226-node correlation graph, built from daily returns between January 2020 and June 2026, is ORC-spherical: every edge has strictly positive curvature before any flow, close to the equal-weight value \(κ\approx0.502\). Under a fixed 20-iteration Ricci flow the curvature distribution polarises into a high-curvature bulk and a low-curvature minority, and cutting the low-curvature edges splits the graph into components of 184 and 42 nodes. The recursive ARCUS procedure (Adaptive Ricci Curvature Unsupervised Surgery) stops the flow at each node once a curvature gap ratio \(G\) exceeds \(G_{\min}\) and the surgery set is stable. With \(G_{\min}=10\) and a minimum cluster size of 15, ARCUS performs 28 surgeries over 13 levels and ends in 29 terminal clusters of 2-14 nodes. The stopping rule fires at every node, with \(G\) between 10.0 and 92.4, and the top-level split reproduces the fixed-schedule 184/42 partition exactly. Eight small subgraphs stop with \(G<15\), so the deeper levels are conditional on \(G_{\min}\). Many terminal clusters are single industries by inspection, among them nine banks, five insurers, five general trading companies, three steelmakers and three shipping lines. Two delisted stocks without price data enter as zero-correlation placeholders; with them excluded, every terminal cluster is more strongly correlated internally than with the rest of the index.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: This paper applies advanced geometric concepts (Ricci flow, Ollivier-Ricci curvature) to financial market analysis, demonstrating high mathematical sophistication. It rigorously tests the methodology on real Nikkei-225 data, providing strong empirical evidence for its claims. The approach of using intrinsic graph geometry for market segmentation is highly novel.

Research Flowchart

  flowchart TD
    A["Research Goal: Intrinsic Geometry of Nikkei-225"] --> B{"Data: Nikkei-225 Daily Returns<br>Jan 2020 - Jun 2026"};
    B --> C{"Methodology: Ollivier-Ricci Curvature (ORC) of Correlation Graph"};
    C --> D["Process 1: Fixed 20-Iteration Ricci Flow<br>+ Surgery"];
    C --> E["Process 2: ARCUS Procedure<br>(Adaptive Ricci Curvature Unsupervised Surgery)"];
    D --> F{"Finding 1: ORC-spherical graph (κ≈0.502)<br>Polarization, 184/42 node split"};
    E --> G{"Finding 2: 28 surgeries, 29 terminal clusters (2-14 nodes)<br>Replication of 184/42 split, industry-specific clusters"};
    F & G --> H["Outcome: Strong internal correlation within terminal clusters"];

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 applying discrete Ollivier‑Ricci curvature and the adaptive ARCUS (Adaptive Ricci Curvature Unsupervised Surgery) procedure to the Nikkei‑225 correlation graph.

Source sentence
We study the intrinsic geometry of the Nikkei-225 through the discrete Ollivier-Ricci curvature (ORC) of its empirical correlation graph, using the graph Ricci flow and surgery methodology of arXiv:2510.15942. The recursive ARCUS procedure (Adaptive Ricci Curvature Unsupervised Surgery) stops the flow at each node once a curvature gap ratio $G$ exceeds $G_{\min}$ and the surgery set is stable.

Assumes 6

  • Stock returns within the index are on average positively correlated. market
    Source sentence
    1 Introduction Stock returns within a market index are, on average, pos-itively correlated, and the resulting correlation matrix has long been used both as a risk-management tool and as the basis for constructing a network representation of the mar-ket [2, 3, 4].
  • Daily closing prices from Yahoo! Finance are used without any survivorship correction or filtering of the correlation matrix. data
    Source sentence
    We use daily closing prices from Yahoo! With N = 226, T/N ≈7, so the empirical correlation matrix car-ries substantial estimation noise [22, 23, 24]; we do not filter it, so that the flow acts on the raw matrix, as in Ref. [1].
  • The analysis uses a fixed list of 226 Nikkei‑225 tickers, ignoring index changes during the sample window. data
    Source sentence
    The node set is a fixed list of Nikkei-225 tickers. The list is not updated for changes in the index during the window, so the analysis concerns a fixed cross-section rather than the index as composed on each date, and no survivorship correction is made.
  • Two delisted stocks are kept as zero‑correlation placeholder nodes, which do not affect correlations among other stocks. other
    Source sentence
    Two tickers returned no data because they have been delisted, although both traded for most of the win-dow: NTT Data Group (delisted 26 September 2025 after a tender offer by NTT; 89% of the window’s sessions) and Hino Motors (delisted 30 March 2026 on the formation of AR-CHION with Mitsubishi Fuso; 97%). They are kept as place-holder nodes with Cij = 0 for all j̸ = i, which makes each equidistant from every other
Show 2 more
  • Ricci flow is run for a fixed 20‑iteration schedule (or until the ARCUS stopping rule fires) with a curvature gap ratio threshold Gmin. model
    Source sentence
    Under a fixed 20-iteration Ricci flow the curvature distribution polarises into a high-curvature bulk and a low-curvature minority, and cutting the low-curvature edges splits the graph into components of 184 and 42 nodes. The recursive ARCUS procedure (Adaptive Ricci Curvature Unsupervised Surgery) stops the flow at each node once a curvature gap ratio $G$ exceeds $G_{\min}$ and the surgery set is stable.
  • A minimum cluster size of 15 nodes is required for terminal clusters. model
    Source sentence
    With $G_{\min}=10$ and a minimum cluster size of 15, ARCUS performs 28 surgeries over 13 levels and ends in 29 terminal clusters of 2-14 nodes.

Data

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • The Nikkei‑225 correlation graph is ORC‑spherical; every edge has strictly positive curvature (≈0.502) before any flow. theoretical
    Source sentence
    The 226-node correlation graph, built from daily returns between January 2020 and June 2026, is ORC-spherical: every edge has strictly positive curvature before any flow, close to the equal-weight value $κ\approx0.502$. The Nikkei-225 correlation graph is ORC-spherical.
  • A fixed 20‑iteration Ricci flow polarises curvature into a high‑curvature bulk and a low‑curvature minority, and cutting the low‑curvature edges splits the graph into components of 184 and 42 nodes. statistical
    Source sentence
    Under a fixed 20-iteration Ricci flow the curvature distribution polarises into a high-curvature bulk and a low-curvature minority, and cutting the low-curvature edges splits the graph into components of 184 and 42 nodes. 7 Conclusion The Nikkei-225 correlation graph is ORC-spherical, and dis-crete Ricci flow develops a bimodal curvature distribution whose low-curvature population can be cut to separate the graph
  • ARCUS with Gmin = 10 and Nmin = 15 performs 28 surgeries over 13 levels, producing 29 terminal clusters containing 2‑14 nodes each. statistical
    Source sentence
    With $G_{\min}=10$ and a minimum cluster size of 15, ARCUS performs 28 surgeries over 13 levels and ends in 29 terminal clusters of 2-14 nodes. Applied recur-sively with Gmin = 10 and Nmin = 15, ARCUS stops on its gap-and-stability criterion at all 28 nodes it visits and produces a 13-level hierarchy of 29 terminal clusters, though the lower levels sit close to the Gmin threshold and the hierarchy mostly 7 sheds
  • Most terminal clusters correspond to single industries (e.g., nine banks, five insurers, five trading companies, three steelmakers, three shipping lines) and have higher internal mean correlation than external correlation. statistical
    Source sentence
    Many terminal clusters are single industries by inspection, among them nine banks, five insurers, five general trading companies, three steelmakers and three shipping lines. For every terminal cluster, the mean internal cor-relation ρin exceeds the mean correlation with the rest of the index ρout (Table 4, placeholders excluded).
  • Every subgraph examined is ORC‑spherical at the start of its own Ricci‑flow run. theoretical
    Source sentence
    Every subgraph is ORC-spherical at the start of its flow. All 28 subgraphs were ORC-spherical at the start of their own flow: the implementation’s per-node flag is true at every node.

Checked by others

Supported by

  • Utilizing Technical Data to Discover Similar Companies in Dhaka Stock Exchange (2023)
    On the question: Do companies in a stock market form industry‑based clusters with higher internal correlation as revealed by a correlation‑matrix network?
    Compare the findings

    This paper: Most terminal clusters correspond to single industries (e.g., nine banks, five insurers, five trading companies, three steelmakers, three shipping lines) and have higher internal mean correlation than external correlation.

    That paper: A correlation matrix based on closing‑price returns reveals that companies cluster together, forming visible cliques in the network.

    Both findings report industry clusters; B quantifies them (9 banks, 5 insurers, 5 trading firms, 3 steelmakers, 3 shippers) confirming A’s visible cliques and higher internal vs external correlation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-20 · same card as JSON · A language model chose these items: check the source sentence before relying on one.