Paper: arXiv 2306.16422

Authors: Ariel Neufeld, Julian Sester

Abstract

In this paper we demonstrate both theoretically as well as numerically that neural networks can detect model-free static arbitrage opportunities whenever the market admits some. Due to the use of neural networks, our method can be applied to financial markets with a high number of traded securities and ensures almost immediate execution of the corresponding trading strategies. To demonstrate its tractability, effectiveness, and robustness we provide examples using real financial data. From a technical point of view, we prove that a single neural network can approximately solve a class of convex semi-infinite programs, which is the key result in order to derive our theoretical results that neural networks can detect model-free static arbitrage strategies whenever the financial market admits such opportunities.

Complexity vs Empirical Score

  • Math Complexity: 9.0/10
  • Empirical Rigor: 6.0/10 (capped from 6.5: transaction costs not modelled; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 6.5)

Why this score: The paper relies heavily on advanced mathematics, including convex semi-infinite programming and neural network approximation theory, to prove theoretical guarantees. It complements this with numerical experiments on real financial data, demonstrating practical implementation and backtest-readiness.

Research Flowchart

  flowchart TD
  A["Research Goal: Can NN detect<br>model-free static arbitrage<br>in high-dimensional markets?"] --> B["Key Methodology"]
  
  B --> C["Theoretical Proof<br>Convex Semi-Infinite Programming<br>approximation via NN"]
  B --> D["Computational Process<br>Single Neural Network<br>optimizes 1st-order conditions"]
  
  C --> E["Data Input<br>Real Financial Data<br>Multi-Asset Markets"]
  D --> E
  
  E --> F["Key Findings:<br>1. NN detects arbitrage if market admits it<br>2. Model-free approach<br>3. Handles high dimensions<br>4. Fast execution"]

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 a neural network based method that can detect model‑free static arbitrage opportunities in markets with many securities.

Source sentence
we demonstrate both theoretically as well as numerically that neural networks can detect model-free static arbitrage opportunities whenever the market admits some.

Assumes 3

  • The market allows static trading in options. market
    Source sentence
    In this section, we study a financial market in which a financial agent can trade statically in various types of options
  • Underlying assets are scaled to 1 and their values at maturity lie in [0,2] per dimension (0%–200% of spot). model
    Source sentence
    The spot values of the underlying assets are scaled to 1, therefore the strikes and corresponding prices are included as percentage values w.r.t. the spot value of the underlying asset. We assume S = [0, 2]5, i
  • Option data are taken from the 10 most liquidly traded call options per S&P 500 constituent with a fixed maturity. data
    Source sentence
    To create a training set we consider for each of the 500 constituents of the S&P 500 the 10 most liquidly traded

Data

  • S&P 500 call options dataset · Yahoo Finance options public
    maturity T = 19 May 2023 (data downloade · 500 S&P 500 constituents, 10 most liquid call options each
    Source sentence
    To create a training set we consider for each of the 500 constituents of the S&P 500 the 10 most liquidly traded

Recognised sources: Yahoo Finance

Costs

Not modeled A trading or execution result is reported without costs.

Finds 3

  • The neural network correctly identifies arbitrage on 92.81% of training samples (41 768 / 45 000). performance
    Source sentence
    After training on the 45 000 samples, the neural network assigns to 41 768 out of 45 000 the correct sign of the price of the strategy learned by the neural network, i.e., in 92.81% of cases on the training set, the neural network can correctly decide whether the market admits arbitrage or not.
  • On the test set the network achieves 89.20% correct arbitrage identification (4 460 / 5 000). performance
    Source sentence
    On the test set we have 4460 out of 5000 correct identifications which corresponds to 89.20%.
  • In the vast majority of 1 000 000 evaluated test cases the net profit of the suggested strategy is positive. theoretical
    Source sentence
    The results verify that on the test set the net profit is in the vast majority of the 1 000 000 evaluated cases positive

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 8 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive

Code

https://github.com/juliansester/Deep-Arbitrage — 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.