Paper: arXiv 2305.12264
Authors: Masanori Hirano, Kentaro Imajo, Kentaro Minami, Takuya Shimada
Abstract
Deep hedging is a framework for hedging derivatives in the presence of market frictions. In this study, we focus on the problem of hedging a given target option by using multiple options. To extend the deep hedging framework to this setting, the options used as hedging instruments also have to be priced during training. While one might use classical pricing model such as the Black-Scholes formula, ignoring frictions can offer arbitrage opportunities which are undesirable for deep hedging learning. The goal of this study is to develop a nested deep hedging method. That is, we develop a fully-deep approach of deep hedging in which the hedging instruments are also priced by deep neural networks that are aware of frictions. However, since the prices of hedging instruments have to be calculated under many different conditions, the entire learning process can be computationally intractable. To overcome this problem, we propose an efficient learning method for nested deep hedging. Our method consists of three techniques to circumvent computational intractability, each of which reduces redundant computations during training. We show through experiments that the Black-Scholes pricing of hedge instruments can admit significant arbitrage opportunities, which are not observed when the pricing is performed by deep hedging. We also demonstrate that our proposed method successfully reduces the hedging risks compared to a baseline method that does not use options as hedging instruments.
Complexity vs Empirical Score
- Math Complexity: 7.5/10
- Empirical Rigor: 6.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper presents a complex nested optimization problem involving deep neural networks for hedging and pricing, requiring advanced stochastic calculus and machine learning theory, while also providing empirical evidence through experiments on simulated data comparing arbitrage opportunities and risk reduction.
Research Flowchart
flowchart TD
A["Research Goal: Develop Efficient Nested Deep Hedging"]
B["Methodology: Fully-Deep Approach<br>Pricing hedging instruments via NNs"]
C["Input Data: Market Frictions &<br>Target & Hedging Options"]
D{"Computational Challenge<br>Intractability of Pricing Under Frictions"}
E["Proposed Solution: Efficient Learning Method<br>- Sub-network Reuse<br>- Shared Initial Layers<br>- Reduced Sampling Steps"]
F["Computational Process:<br>Arbitrage-Free Training"]
G["Key Findings"]
A --> B
B --> C
C --> D
D --> E
E --> F
F --> G
subgraph G_Outcomes ["Outcomes"]
G1["Arbitrage Present in Black-Scholes Pricing"]
G2["Risk Reduced vs Baseline"]
end
G --> G1
G --> G2
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 an efficient learning method for nested deep hedging where an option is hedged using multiple options and stocks.
Source sentence
We propose an efficient learning method of deep hedging for the situation where an option is hedged by options and stocks.
Assumes 3
- Market frictions are present and modeled as proportional transaction costs for both the underlying stock and tradable options. execution
Source sentence
Proportional cost coefficient for the underlying stock: c0 = 0.01, 0.005, 0.001, 0.0005, 0.0001
- The underlying asset follows the Heston stochastic volatility model with specified parameters. model
Source sentence
Parameters for the Heston model3: S(0) t0 = 1.0, Vt0 = 0.04, κ = 1.0, θ = 0.04, ρ = −0.7, and σ = 0.3
- The primary trader’s pricing of hedge instruments must be recomputed at every time step under many conditions. other
Source sentence
Since the secondary trader is allowed to trade hedging instruments at every time step until maturity, the prices of the hedging instruments have to be updated many times.
Data
- Synthetic Heston model data synthetic unknown
Source sentence
Parameters for the Heston model3: S(0) t0 = 1.0, Vt0 = 0.04, κ = 1.0, θ = 0.04, ρ = −0.7, and σ = 0.3
Costs
Modeled
Proportional cost coefficients are set for the underlying stock (c0) and tradable options (ci) at values 0.01, 0.005, 0.001, 0.0005, 0.0001.
Source sentence
Proportional cost coefficient for the underlying stock: c0 = 0.01, 0.005, 0.001, 0.0005, 0.0001
Finds 4
- Black‑Scholes pricing of hedge instruments creates significant arbitrage opportunities, yielding a much smaller hedging cost for the secondary trader. performance
Source sentence
In the case of Black–Scholes, the hedging cost of the secondary trader is significantly small. It means that the pricing of the Black–Scholes-based primary trader is inappropriate and gives an arbitrage chance to the secondary trader.
- The proposed nested deep hedging method reduces hedging costs compared to a stock‑only baseline. performance
Source sentence
According to the results, the proposed secondary trader successfully reduces the hedging costs.
- The proposed method lowers tail risk as measured by CVaR(0.1), improving the distribution of P&L. performance
Source sentence
According to this plot, the proposed secondary trader successfully reduces only the tail risk, which effectively reduces CVaR(0.1).
- The proposed method reduces most Greeks (gamma, theta, vega) relative to the stock‑only trader, though delta performance is slightly lower. performance
Source sentence
secondary trader successfully reduces the greeks except for delta more than the stock-only secondary traders.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 28 related papers). Coverage is partial, so this is neither replication nor refutation.
2 papers in the archive
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.