Paper: arXiv 2609.06137
Authors: Evgeny Lakshtanov
Abstract
Pathwise differentiation of Monte Carlo estimators fails at payoff discontinuities, producing zero or biased sensitivities for barriers, autocallables, and digital options. The industry workaround — smoothing the indicator functions — introduces bias and requires per-product calibration. We derive a correction formula that restores unbiased Greeks without smoothing. For a payoff \(F(Z,θ)\) that is piecewise smooth with discontinuities on surfaces \(\{g_i = 0\}\), we show that the sensitivity decomposes into a pathwise term (computed by standard AAD) plus a sum of boundary corrections, each involving the payoff jump, the Gaussian density at the boundary, and the sensitivity of the boundary to the parameter. The correction is computed by Newton root-finding in the normal-random space, with the jump evaluated by two forward replays of the pricing kernel. The implementation uses AADC (\texttt{pip install aadc}), whose tape replay and automatic discontinuity tracking make the method fully automatic — the quant writes standard pricing code, and the correction driver identifies and handles all discontinuities. We prove the formula for arbitrary compositions of smooth functions and indicator functions (not just outer products), covering real autocallable payoff structures with recursive alive/dead logic. Benchmarks on QuantLib models (GBM, Heston, Hull-White) show all Greeks within 0.1–4% of analytic or bump-and-revalue references.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 8.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: This paper presents a highly novel and mathematically sophisticated method for computing unbiased Monte Carlo Greeks for discontinuous payoffs. The empirical validation against QuantLib models and comparison with industry workarounds demonstrates strong rigor. The clear exposition and explicit mention of an open-source library for implementation further enhance its value.
Research Flowchart
flowchart TD
A["Research Goal: Unbiased Monte Carlo Greeks for Discontinuous Payoffs"] --> B{"Problem: Pathwise Differentiation Fails at Discontinuities"};
B --> C["Methodology: Derive Correction Formula for Unbiased Greeks"];
C --> D{"Computational Process: AADC for Automatic Discontinuity Tracking & Correction"};
D --> E["Correction Formula Inputs: Payoff Jump, Gaussian Density, Boundary Sensitivity"];
E --> F["Key Outcome: Unbiased Greeks for Discontinuous Payoffs (within 0.1-4% of references)"];
D --> G["Key Outcome: Automatic Implementation via AADC (Standard Pricing Code)"];
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 correction formula that restores unbiased Monte‑Carlo Greeks for discontinuous payoffs without smoothing.
Source sentence
We derive a correction formula that restores unbiased Greeks without smoothing.
Assumes 5
- Each discontinuity surface g_i is smooth with non‑zero gradient ∇_Z g_i on the surface. model
Source sentence
Let g1, . . . , gK : Rd × R →R be smooth with ∇Zgi̸ = 0 on Γi = {gi = 0}
- The discontinuity surfaces Γ_i are pairwise transversal (normals not parallel). model
Source sentence
Surfaces Γi, Γj are transversal if their normals vi, vj are not parallel at intersection points.
- The payoff F(Z,θ) is piecewise smooth with discontinuities only on the defined surfaces. model
Source sentence
Let F(Z, θ) be piecewise smooth with discontinuities only on Γ1, . . . , ΓK.
- The method works for any underlying model (e.g., GBM, Heston, Hull‑White, hybrid). model
Source sentence
The correction is unbiased, requires no parameter tuning, and works for any model (GBM, Heston, Hull-White, hybrid).
Show 1 more
- Newton root‑finding in the normal‑random space converges to the boundary point u* for each indicator. execution
Source sentence
Newton: find u∗ i s.t. gi(Z∗) = 0 along vi.
Data
- GBM model (Black‑Scholes process) synthetic public
Source sentence
All benchmarks use QuantLib models via the AADC library (pip install aadc).
Costs
Not applicable
Finds 4
- Greeks computed with the correction are within 0.1–4 % of analytic or bump‑and‑revalue references across all tested models. theoretical
Source sentence
Benchmarks on QuantLib models (GBM, Heston, Hull-White) show all Greeks within 0.1--4% of analytic or bump-and-revalue references.
- For a down‑and‑out call under GBM, the correction matches the analytic delta within –0.1 % and vega within –0.1 %, while smoothing shows larger errors. pricing-calibration
Source sentence
For 2% accuracy: correction ∼8K paths, bump-and-revalue ∼12M paths.
- In the GBM + Hull‑White case, the correction is essential for vega: the pathwise estimator overestimates vega by 69 %, while the corrected vega matches the total. performance
Source sentence
Correction is essential for vega: pathwise alone overestimates by 69%.
- Using a screening threshold σskip = 20 skips about 70–80 % of indicators per path, reducing computational cost without loss of accuracy. statistical
Source sentence
With σskip = 20: ∼70–80% of indicators are skipped per path.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 22 related papers). Coverage is partial, so this is neither replication nor refutation.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.