Paper: arXiv 2305.06805

Authors: Ludovic Goudenège, Andrea Molent, Antonino Zanette

Abstract

In this article, we introduce an algorithm called Backward Hedging, designed for hedging European and American options while considering transaction costs. The optimal strategy is determined by minimizing an appropriate loss function, which is based on either a risk measure or the mean squared error of the hedging strategy at maturity. The proposed algorithm moves backward in time, determining, for each time-step and different market states, the optimal hedging strategy that minimizes the loss function at the time the option is exercised, by assuming that the strategy used in the future for hedging the liability is the one determined at the previous steps of the algorithm. The approach avoids machine learning and instead relies on classic optimization techniques, Monte Carlo simulations, and interpolations on a grid. Comparisons with the Deep Hedging algorithm in various numerical experiments showcase the efficiency and accuracy of the proposed method.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced stochastic calculus, convex risk measures, and optimization theory, indicating high mathematical complexity, while the method is validated through extensive numerical experiments, Monte Carlo simulations, and comparisons with Deep Hedging, demonstrating strong empirical implementation.

Research Flowchart

  flowchart TD
  A["Research Goal: <br>Hedge American Options w/ Transaction Costs"] --> B["Methodology: <br>Backward Hedging Algorithm"]
  B --> C["Input Data: <br>Monte Carlo Simulated Asset Paths"]
  C --> D["Computational Process: <br>Backward Time Iteration"]
  D --> E{"At Exercise Date: <br>Minimize Loss Function?"}
  E -- Yes --> F["Calculate Optimal Hedge <br>Classic Optimization"]
  E -- No --> G["Interpolate Hedge from Grid"]
  F --> H["Key Findings: <br>High Accuracy & Efficiency"]
  G --> H
  H --> I["Comparison: <br>Superior to Deep Hedging"]

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 Backward Hedging algorithm for hedging European and American options with transaction costs, using classic optimization, Monte Carlo simulations, and grid interpolation.

Source sentence
we introduce an algorithm called Backward Hedging, designed for hedging European and American options while considering transaction costs.

Assumes 5

  • The interest rate is assumed to be zero in the illustrative discrete‑time example. model
    Source sentence
    Also, for simplicity, we assume the interest rate r to be zero.
  • The market state at each time can be described by the triple (S1_t, V_t, I_t) where I_t = ∫_0^t V_s ds. market
    Source sentence
    V_s ds in place of all filtration F_tn, so the market state can be described by this triple.
  • Transaction costs are proportional and constant (e.g., 1% for S1, 10% for S2). execution
    Source sentence
    Proportional cost for trading S1 is 1%, while the proportional cost for trading S2 has been raised to 10%.
  • The underlying follows Heston’s stochastic volatility model with possibly non‑zero drift and interest rate. model
    Source sentence
    we present it with reference to Heston’s (one-dimensional) stochastic model, which is also studied in the original manuscript.
Show 1 more
  • Monte Carlo simulations are used to estimate risk measures and to build the grid of market states. data
    Source sentence
    First of all, we propose to estimate the risk measure ρ through (forward) M Monte Carlo simulations, from t_n to T, of the underlying processes S1 and S2 conditioned by F_tn.

Data

  • Synthetic Monte Carlo paths for S1, S2 and variance process synthetic unknown
    Source sentence
    ... and we apply the algorithm on this new out‑of‑sample data.
  • CRR binomial model (T=2, N=8) for illustration synthetic unknown
    Source sentence
    We stress out that these values are the same as those generated at time‑steps 0, 4 and 8 by the CRR model with T = 2 and N = 8 discretization steps

Costs

Modeled Costs are part of the reported headline results.

Proportional costs of 1% for asset S1 and 10% for asset S2 are incorporated in the hedging simulations.

Source sentence
Proportional cost for trading S1 is 1%, while the proportional cost for trading S2 has been raised to 10%.

Finds 2

  • The Backward Hedging algorithm achieves comparable or better hedging performance than Deep Hedging while being computationally efficient. performance
    Source sentence
    Comparisons with the Deep Hedging algorithm in various numerical experiments showcase the efficiency and accuracy of the proposed method.
  • For an American put option with r = µ = 0.1, minimizing CVaR loss yields hedging costs reported in Table 13 (e.g., BH cost 8.3280, DH cost 8.6478). pricing-calibration
    Source sentence
    Table 13: Hedging cost for an American put option with r = µ = 0.1, while minimizing loss based on CVaR.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 18 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 2 papers in the archive (1 by independent authors): Hedging with Sparse Reward Reinforcement Learning (uses its method or data)

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.