Paper: arXiv 2508.14784
Authors: Yoonsik Hong, Diego Klabjan
Abstract
We propose a two-step graph learning approach for foreign exchange statistical arbitrages (FXSAs), addressing two key gaps in prior studies: the absence of graph-learning methods for foreign exchange rate prediction (FXRP) that leverage multi-currency and currency-interest rate relationships, and the disregard of the time lag between price observation and trade execution. In the first step, to capture complex multi-currency and currency-interest rate relationships, we formulate FXRP as an edge-level regression problem on a discrete-time spatiotemporal graph. This graph consists of currencies as nodes and exchanges as edges, with interest rates and foreign exchange rates serving as node and edge features, respectively. We then introduce a graph-learning method that leverages the spatiotemporal graph to address the FXRP problem. In the second step, we present a stochastic optimization problem to exploit FXSAs while accounting for the observation-execution time lag. To address this problem, we propose a graph-learning method that enforces constraints through projection and ReLU, maximizes risk-adjusted return by leveraging a graph with exchanges as nodes and influence relationships as edges, and utilizes the predictions from the FXRP method for the constraint parameters and node features. Moreover, we prove that our FXSA method satisfies empirical arbitrage constraints. The experimental results demonstrate that our FXRP method yields statistically significant improvements in mean squared error, and that the FXSA method achieves a 61.89% higher information ratio and a 45.51% higher Sortino ratio than a benchmark. Our approach provides a novel perspective on FXRP and FXSA within the context of graph learning.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 6.0/10 (capped from 7.5: transaction costs not modelled; see penalty flags)
- Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 7.5)
Why this score: The paper uses advanced mathematics including stochastic optimization, graph neural networks, projection/ReLU constraints, and proofs of constraint satisfaction, indicating high complexity. It reports specific backtest results (information ratio, Sortino ratio, MSE improvements) on real market data, showing substantial empirical testing and implementation detail.
Research Flowchart
flowchart TD
A["Research Goal:<br>FX Prediction & Statistical Arbitrage"] --> B["Data Inputs:<br>FX Rates, Interest Rates"]
B --> C["Step 1: FX Rate Prediction<br>Spatiotemporal Graph Learning"]
C --> D["Step 2: Statistical Arbitrage<br>Graph-based Optimization"]
D --> E{"Compute<br>Risk-Adjusted Return"}
E --> F["Key Findings"]
subgraph F ["Outcomes"]
F1["MSE ↓ (Significant)"]
F2["Info Ratio ↑ 61.89%"]
F3["Sortino Ratio ↑ 45.51%"]
end
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 two-step graph learning approach for foreign exchange rate prediction and statistical arbitrage.
Source sentence
We propose a two-step graph learning approach for foreign exchange statistical arbitrages (FXSAs), addressing two key gaps in prior studies: the absence of graph-learning methods for foreign exchange rate prediction (FXRP) that leverage multi-currency and currency–interest rate relationships, and the disregard of the t
Assumes 2
- The benchmark expected return is zero. execution
Source sentence
we assume that the expected return of the benchmark that invests
- Trading occurs only on weekdays; weekends are excluded. data
Source sentence
Trading is assumed to occur only on weekdays, with weekends excluded from trading dates.
Data
- Foreign exchange rates and interest rate data market-prices public
Source sentence
These values are identified by comparison with historical charts from Yahoo Finance [33].
Costs
Not modeled
Finds 4
- The FXRP graph neural network method yields statistically significant improvements in mean squared error over the MLP benchmark. predictive-accuracy
Source sentence
Table 2 shows our FXRP method outperforms the benchmark MLP in mean squared error (MSE).
- The FXSA graph learning method achieves a 61.89% higher information ratio than the benchmark LP. performance
Source sentence
our FXSA method (GNN) achieves, on average, 61.89% higher information ratio
- The FXSA method achieves a 45.51% higher Sortino ratio than the benchmark LP. performance
Source sentence
and a 45.51% higher Sortino ratio than the benchmark
- The FXSA method reduces exposure to FX rate risk, showing lower average absolute holdings and an 83.12% lower Herfindahl‑Hirschman index. performance
Source sentence
Our method consistently exhibits lower values than LP throughout most of the test period
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 55 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (8)
- Scilab — link
- Scilab — link
- The Computational Geometry Algorithms Library — link
- The Computational Geometry Algorithms Library — link
- The Parmap library — link
- The Parmap library — link
- BabyJ : from Object Based to Class Based Programming via Types (2003)
- tag-sound
0 of 8 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.