arxiv: 2604.03122 · v1 · submitted 2026-04-03 · 🧮 math.NA · cs.NA

Recognition: no theorem link

Nested Multilevel Monte Carlo with Preintegration for Efficient Risk Estimation

Yu Xu , Xiaoqun Wang

Authors on Pith no claims yet

Pith reviewed 2026-05-13 18:50 UTC · model grok-4.3

classification 🧮 math.NA cs.NA

keywords nested Monte Carlomultilevel Monte Carlorisk estimationpreintegrationconvergence ratecomputational complexityquasi-Monte Carlo

0 comments

The pith

Nested MLMC with preintegration reaches a strong convergence rate of -1 for risk estimation.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a nested multilevel Monte Carlo method that adds preintegration to estimate risks more efficiently. Preintegration integrates out one outer random variable, which removes the discontinuity from the indicator function without introducing bias. The multilevel estimator built on this smoothed quantity then converges at a strong rate of -1 instead of the usual -1/2. This change produces nearly optimal overall computational complexity. The same construction also reduces the high kurtosis that indicator functions normally create, and numerical tests show further gains when quasi-Monte Carlo is added in high dimensions.

Core claim

The authors prove that the MLMC estimator combined with preintegration attains a strong convergence rate of -1, compared with -1/2 for standard MLMC, thereby achieving nearly optimal computational complexity for nested risk estimation. Preintegration removes the discontinuity of the indicator function by integrating out one outer random variable while preserving the multilevel hierarchy. The method also mitigates the high-kurtosis phenomenon caused by indicators. Numerical experiments confirm that the smoothed MLMC with preintegration outperforms standard MLMC and reaches the predicted optimal cost, with additional improvement when combined with quasi-Monte Carlo.

What carries the argument

Preintegration applied to integrate out one outer random variable, which smooths the indicator discontinuity while preserving the multilevel hierarchy needed for the MLMC estimator.

If this is right

The combined estimator attains nearly optimal computational complexity for nested risk simulations.
Discontinuities arising from indicator functions in risk measures are handled without bias.
High-kurtosis effects produced by indicator functions are reduced.
Performance improves further when the method is paired with quasi-Monte Carlo sampling in high dimensions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same preintegration step could be applied to other Monte Carlo problems that involve discontinuous payoffs or indicators.
The faster convergence rate may allow practitioners to use smaller sample sizes while maintaining accuracy in risk-management calculations.
Choosing which variable to integrate out may require problem-specific analysis when extending the method to higher-dimensional or path-dependent settings.

Load-bearing premise

Preintegration can integrate out one outer random variable while preserving the multilevel structure and without introducing bias that would invalidate the convergence analysis.

What would settle it

A numerical measurement of the strong error decay for the preintegrated MLMC estimator that shows a rate slower than -1, or total computational cost that exceeds the predicted near-optimal order.

Figures

Figures reproduced from arXiv: 2604.03122 by Xiaoqun Wang, Yu Xu.

**Figure 1.** Figure 1: Expectation for dimensions d = 4, 8, 16 1 2 3 4 5 6 7 level 20.0 17.5 15.0 12.5 10.0 7.5 5.0 2.5 lo g2 v aria nce d=4 MLMC smoothed MLMC smoothed AMLMC 2 2 /2 2 3 /2 1 2 3 4 5 6 7 level 16 14 12 10 8 6 4 2 lo g2 v aria nce d=8 MLMC smoothed MLMC smoothed AMLMC 2 2 /2 2 3 /2 1 2 3 4 5 6 7 level 14 12 10 8 6 4 lo g2 v aria nce d=16 MLMC smoothed MLMC smoothed AMLMC 2 2 /2 2 3 /2 [PITH_FULL_IMAGE:figures/ful… view at source ↗

**Figure 2.** Figure 2: Variance for dimensions d = 4, 8, 16 19 [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗

**Figure 3.** Figure 3: Kurtosis for dimensions d = 4, 8, 16 10 3 10 2 TOL 10 5 10 6 10 7 10 8 10 9 10 10 total cost d=4 MLMC smoothed MLMC smoothed AMLMC TOl 2 TOl 2.5 10 3 10 2 TOL 10 5 10 6 10 7 10 8 10 9 10 10 total cost d=8 MLMC smoothed MLMC smoothed AMLMC TOl 2 TOl 2.5 10 3 10 2 TOL 10 6 10 7 10 8 10 9 10 10 total cost d=16 MLMC smoothed MLMC smoothed AMLMC TOl 2 TOl 2.5 [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗

**Figure 4.** Figure 4: Cost for dimensions d = 4, 8, 16 20 [PITH_FULL_IMAGE:figures/full_fig_p020_4.png] view at source ↗

**Figure 5.** Figure 5: Variance for dimensions d = 32, 50 1 2 3 4 5 6 7 level 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 lo g2 k urtosis d=32 MLMC smoothed MLMC smoothed MLQMC [PITH_FULL_IMAGE:figures/full_fig_p022_5.png] view at source ↗

**Figure 6.** Figure 6: Kurtosis for dimension d = 32 22 [PITH_FULL_IMAGE:figures/full_fig_p022_6.png] view at source ↗

read the original abstract

Nested Monte Carlo is widely used for risk estimation, but its efficiency is limited by the discontinuity of the indicator function and high computational cost. This paper proposes a nested Multilevel Monte Carlo (MLMC) method combined with preintegration for efficient risk estimation. We first use preintegration to integrate out one outer random variable, which effectively handles the discontinuity of the indicator function, then we construct the MLMC estimator with preintegration to reduce the computational cost. Our theoretical analysis proves that the strong convergence rate of the MLMC combined with preintegration reaches -1, compared with -1/2 for the standard MLMC. Consequently, we obtain a nearly optimal computational complexity. Besides, our method can also handle the high-kurtosis phenomenon caused by indicator functions. Numerical experiments verify that the smoothed MLMC with preintegration outperforms the standard MLMC and the optimal computational cost can be attained. Combining our method with quasi-Monte Carlo further improves its performance in high dimensions. Keywords: Nested simulation, Multilevel Monte Carlo, Risk estimation, Preintegration

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Preintegration inside nested MLMC lifts the strong rate to -1 for discontinuous risk functionals, which is the real advance here.

read the letter

The main point is that this paper shows how to preintegrate out one outer variable to remove the indicator jump, then run multilevel Monte Carlo on the smoothed integrand and obtain a strong convergence rate of -1 instead of the usual -1/2. That rate improvement gives nearly optimal complexity for nested risk simulations, which is useful in practice for things like VaR or expected shortfall calculations. They prove the variance of the level differences decays fast enough under the preintegrated function and back it with experiments that show clear cost savings over plain nested MLMC, plus extra gains when they add quasi-Monte Carlo in higher dimensions. The method also reduces the kurtosis blow-up that indicators normally create. The combination itself is new; separate uses of preintegration and MLMC exist, but nesting the smoothing step inside the multilevel hierarchy for this exact setting and proving the rate is not routine. The numerics look clean and use reasonable baselines for risk estimation. The soft spot is that everything rests on the preintegrated payoff inheriting enough regularity so the multilevel coupling still cancels the leading variance term without hidden bias or extra quadrature cost. If the original payoff has kinks that survive or if the preintegration points are chosen poorly, the rate could degrade in some cases, though the paper claims the analysis covers this. The work is aimed at people who already run nested Monte Carlo for finance or insurance and want a concrete efficiency boost without changing the model. It is worth sending to peer review because the claim is specific, the theory is stated clearly, and the experiments are reproducible enough to check.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes combining preintegration with nested Multilevel Monte Carlo (MLMC) for risk estimation. Preintegration is used to integrate out one outer random variable, smoothing the indicator-function discontinuity; an MLMC estimator is then built on the resulting integrand. The central theoretical claim is that this yields a strong convergence rate of -1 (versus -1/2 for standard MLMC), delivering nearly optimal computational complexity while also mitigating high-kurtosis effects. Numerical experiments and a QMC extension are presented to support the claims.

Significance. If the rate -1 is rigorously established without hidden bias or regularity loss, the work would provide a concrete, implementable improvement for nested simulation problems in risk management and finance, where indicator discontinuities and high kurtosis are common. The combination of preintegration with MLMC is a natural extension of existing theory, and the claimed complexity gain would be practically relevant.

major comments (2)

[§3] §3 (Theoretical Analysis, convergence-rate theorem): the proof that preintegration produces an integrand whose level differences satisfy Var[ΔP_l] = O(2^{-2l}) (or the equivalent strong-rate condition) must explicitly verify that the preintegrated payoff inherits the required Lipschitz or C^1 regularity from the original payoff and that the multilevel coupling still cancels the leading variance term. The current sketch does not rule out residual kinks or quadrature bias that would revert the rate to -1/2.
[§4] §4 (Complexity analysis): the nearly-optimal complexity statement assumes the preintegration step is either exact or its error is absorbed into a higher-order term. Please supply the precise quadrature-error bound and show it does not degrade the O(ε^{-2}) (or better) cost when the target RMSE is ε.

minor comments (2)

[Numerical experiments] Figure 2 and Table 1: the plotted empirical rates should include 95% confidence intervals obtained from repeated independent runs; single-run slopes can be misleading for MLMC variance estimates.
[§2] Notation: the symbol for the preintegrated payoff function is introduced without a clear definition distinguishing it from the original payoff; a short notational table or explicit definition in §2 would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which help strengthen the theoretical foundations of our work. We address each major comment below and will incorporate the requested clarifications into the revised manuscript.

read point-by-point responses

Referee: §3 (Theoretical Analysis, convergence-rate theorem): the proof that preintegration produces an integrand whose level differences satisfy Var[ΔP_l] = O(2^{-2l}) (or the equivalent strong-rate condition) must explicitly verify that the preintegrated payoff inherits the required Lipschitz or C^1 regularity from the original payoff and that the multilevel coupling still cancels the leading variance term. The current sketch does not rule out residual kinks or quadrature bias that would revert the rate to -1/2.

Authors: We agree that the proof sketch in §3 requires additional detail for full rigor. In the revision we will insert a new lemma establishing that, under the paper's standing assumptions on the payoff (Lipschitz continuity in the outer variable and suitable smoothness in the inner variable), the preintegrated integrand is C^1. We will then show that the standard multilevel coupling still cancels the leading O(2^{-l}) variance term, yielding Var[ΔP_l] = O(2^{-2l}) with no residual kinks or quadrature bias that would reduce the rate. The expanded argument will be placed immediately after the current sketch in §3. revision: yes
Referee: §4 (Complexity analysis): the nearly-optimal complexity statement assumes the preintegration step is either exact or its error is absorbed into a higher-order term. Please supply the precise quadrature-error bound and show it does not degrade the O(ε^{-2}) (or better) cost when the target RMSE is ε.

Authors: We thank the referee for highlighting this point. We will add an explicit quadrature-error analysis to §4. Using a composite trapezoidal rule (or higher-order quadrature) with mesh size proportional to 2^{-l}, the preintegration error is bounded by O(2^{-2l}). This term is absorbed into the existing MLMC bias and variance bounds without altering the leading constants. Consequently, the total cost remains O(ε^{-2} (log ε^{-1})^2) for RMSE ε, which is still nearly optimal. The revised §4 will contain the corresponding theorem and the choice of quadrature parameters that guarantee the bound. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on independent theoretical analysis of preintegrated MLMC

full rationale

The paper claims a strong convergence rate of -1 for the combined MLMC+preintegration estimator via theoretical analysis, contrasting with the standard -1/2 rate. No quoted equations or steps reduce the claimed rate, variance decay, or complexity result to a fitted parameter, self-definition, or self-citation chain by construction. The preintegration step is described as integrating out one variable to handle discontinuity while preserving multilevel structure, with the rate proved separately. This is a standard extension of existing MLMC theory; the central result does not collapse to renaming inputs or load-bearing self-citations. Score 0 is appropriate as the derivation remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper builds on standard Monte Carlo and multilevel theory plus the preintegration technique; no new free parameters, invented entities, or ad-hoc axioms are introduced in the abstract description.

axioms (1)

domain assumption The integrand after preintegration satisfies the regularity conditions required for the multilevel variance decay analysis.
Needed to obtain the claimed strong convergence rate of -1.

pith-pipeline@v0.9.0 · 5477 in / 1209 out tokens · 40366 ms · 2026-05-13T18:50:24.765914+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

[1]

Conditional sampling for barrier option pricing under the Heston model

Nico Achtsis, Ronald Cools, and Dirk Nuyens. Conditional sampling for barrier option pricing under the Heston model. InMonte Carlo and Quasi-Monte Carlo Methods 2012, pages 253–269. Springer, 2013

work page 2012
[2]

Conditional sampling for barrier option pricing under the LT method.SIAM Journal on Financial Mathematics, 4(1):327–352, 2013

Nico Achtsis, Ronald Cools, and Dirk Nuyens. Conditional sampling for barrier option pricing under the LT method.SIAM Journal on Financial Mathematics, 4(1):327–352, 2013

work page 2013
[3]

Numerical smoothing with hierarchical adaptive sparse grids and quasi-Monte Carlo methods for efficient option pricing

Christian Bayer, Chiheb Ben Hammouda, and Ra´ ul Tempone. Numerical smoothing with hierarchical adaptive sparse grids and quasi-Monte Carlo methods for efficient option pricing. Quantitative Finance, 23(2):209–227, 2023

work page 2023
[4]

Multilevel Monte Carlo with numerical smoothing for robust and efficient computation of probabilities and densities.SIAM Journal on Scientific Computing, 46(3):A1514–A1548, 2024

Christian Bayer, Chiheb Ben Hammouda, and Ra´ ul Tempone. Multilevel Monte Carlo with numerical smoothing for robust and efficient computation of probabilities and densities.SIAM Journal on Scientific Computing, 46(3):A1514–A1548, 2024

work page 2024
[5]

Smoothing the payoff for efficient computation of basket option prices.Quantitative Finance, 18(3):491–505, 2018

Christian Bayer, Markus Siebenmorgen, and Raul Tempone. Smoothing the payoff for efficient computation of basket option prices.Quantitative Finance, 18(3):491–505, 2018

work page 2018
[6]

Importance sampling for a robust and efficient multilevel Monte Carlo estimator for stochastic reaction networks.Statistics and Computing, 30(6):1665–1689, 2020

Chiheb Ben Hammouda, Nadhir Ben Rached, and Ra´ ul Tempone. Importance sampling for a robust and efficient multilevel Monte Carlo estimator for stochastic reaction networks.Statistics and Computing, 30(6):1665–1689, 2020

work page 2020
[7]

Efficient risk estimation via nested se- quential simulation.Management Science, 57(6):1172–1194, 2011

Mark Broadie, Yiping Du, and Ciamac C Moallemi. Efficient risk estimation via nested se- quential simulation.Management Science, 57(6):1172–1194, 2011

work page 2011
[8]

Karolina Bujok, Ben M Hambly, and Christoph Reisinger. Multilevel simulation of functionals of bernoulli random variables with application to basket credit derivatives.Methodology and Computing in Applied Probability, 17(3):579–604, 2015

work page 2015
[9]

Monte Carlo and quasi-Monte Carlo methods.Acta numerica, 7:1–49, 1998

Russel E Caflisch. Monte Carlo and quasi-Monte Carlo methods.Acta numerica, 7:1–49, 1998

work page 1998
[10]

Valuation of mortgage-backed securities using Brownian bridges to reduce effective dimension.Journal of Computational Finance, 1997

Russel E Caflisch, William Morokoff, and Art Owen. Valuation of mortgage-backed securities using Brownian bridges to reduce effective dimension.Journal of Computational Finance, 1997

work page 1997
[11]

Multilevel higher- order quasi-Monte Carlo bayesian estimation.Mathematical Models and Methods in Applied Sciences, 27(05):953–995, 2017

Josef Dick, Robert N Gantner, Quoc T Le Gia, and Christoph Schwab. Multilevel higher- order quasi-Monte Carlo bayesian estimation.Mathematical Models and Methods in Applied Sciences, 27(05):953–995, 2017

work page 2017
[12]

Cambridge University Press, 2010

Josef Dick and Friedrich Pillichshammer.Digital nets and sequences: discrepancy theory and quasi-Monte Carlo integration. Cambridge University Press, 2010

work page 2010
[13]

Multilevel Monte Carlo path simulation.Operations research, 56(3):607–617, 2008

Michael B Giles. Multilevel Monte Carlo path simulation.Operations research, 56(3):607–617, 2008

work page 2008
[14]

Multilevel Monte Carlo methods.Acta numerica, 24:259–328, 2015

Michael B Giles. Multilevel Monte Carlo methods.Acta numerica, 24:259–328, 2015

work page 2015
[15]

MLMC for nested expectations

Michael B Giles. MLMC for nested expectations. InContemporary computational mathematics- A celebration of the 80th birthday of ian sloan, pages 425–442. Springer, 2018. 24

work page 2018
[16]

Decision-making under uncertainty: using MLMC for efficient estimation of EVPPI.Statistics and computing, 29(4):739–751, 2019

Michael B Giles and Takashi Goda. Decision-making under uncertainty: using MLMC for efficient estimation of EVPPI.Statistics and computing, 29(4):739–751, 2019

work page 2019
[17]

Multilevel nested simulation for efficient risk estimation.SIAM/ASA Journal on Uncertainty Quantification, 7(2):497–525, 2019

Michael B Giles and Abdul-Lateef Haji-Ali. Multilevel nested simulation for efficient risk estimation.SIAM/ASA Journal on Uncertainty Quantification, 7(2):497–525, 2019

work page 2019
[18]

Conditioning on one-step survival for barrier option simulations.Operations Research, 49(6):923–937, 2001

Paul Glasserman and Jeremy Staum. Conditioning on one-step survival for barrier option simulations.Operations Research, 49(6):923–937, 2001

work page 2001
[19]

Multilevel Monte Carlo estimation of expected information gains.Stochastic Analysis and Applications, 38(4):581–600, 2020

Takashi Goda, Tomohiko Hironaka, and Takeru Iwamoto. Multilevel Monte Carlo estimation of expected information gains.Stochastic Analysis and Applications, 38(4):581–600, 2020

work page 2020
[20]

Decision-theoretic sensitiv- ity analysis for reservoir development under uncertainty using multilevel quasi-Monte Carlo methods.Computational Geosciences, 22(4):1009–1020, 2018

Takashi Goda, Daisuke Murakami, Kei Tanaka, and Kozo Sato. Decision-theoretic sensitiv- ity analysis for reservoir development under uncertainty using multilevel quasi-Monte Carlo methods.Computational Geosciences, 22(4):1009–1020, 2018

work page 2018
[21]

Nested simulation in portfolio risk measurement.Man- agement Science, 56(10):1833–1848, 2010

Michael B Gordy and Sandeep Juneja. Nested simulation in portfolio risk measurement.Man- agement Science, 56(10):1833–1848, 2010

work page 2010
[22]

High dimensional in- tegration of kinks and jumps—smoothing by preintegration.Journal of Computational and Applied Mathematics, 344:259–274, 2018

Andreas Griewank, Frances Y Kuo, Hernan Le¨ ovey, and Ian H Sloan. High dimensional in- tegration of kinks and jumps—smoothing by preintegration.Journal of Computational and Applied Mathematics, 344:259–274, 2018

work page 2018
[23]

Monte Carlo complexity of global solution of integral equations.Journal of complexity, 14(2):151–175, 1998

Stefan Heinrich. Monte Carlo complexity of global solution of integral equations.Journal of complexity, 14(2):151–175, 1998

work page 1998
[24]

Kernel smoothing for nested estimation with application to portfolio risk measurement.Operations Research, 65(3):657–673, 2017

L Jeff Hong, Sandeep Juneja, and Guangwu Liu. Kernel smoothing for nested estimation with application to portfolio risk measurement.Operations Research, 65(3):657–673, 2017

work page 2017
[25]

Pearson Education India, 2016

John C Hull and Sankarshan Basu.Options, futures, and other derivatives. Pearson Education India, 2016

work page 2016
[26]

Recent advances in randomized quasi-Monte Carlo methods.Modeling uncertainty: An examination of stochastic theory, methods, and applica- tions, pages 419–474, 2002

Pierre L’Ecuyer and Christiane Lemieux. Recent advances in randomized quasi-Monte Carlo methods.Modeling uncertainty: An examination of stochastic theory, methods, and applica- tions, pages 419–474, 2002

work page 2002
[27]

SIAM, 1992

Harald Niederreiter.Random number generation and quasi-Monte Carlo methods. SIAM, 1992

work page 1992
[28]

Randomly permuted (t, m, s)-nets and (t, s)-sequences

Art B Owen. Randomly permuted (t, m, s)-nets and (t, s)-sequences. InMonte Carlo and Quasi-Monte Carlo Methods in Scientific Computing: Proceedings of a conference at the Uni- versity of Nevada, Las Vegas, Nevada, USA, June 23–25, 1994, pages 299–317. Springer, 1995

work page 1994
[29]

Practical Quasi-Monte Carlo integration.URL https://artowen

Art B Owen. Practical Quasi-Monte Carlo integration.URL https://artowen. su. do- mains/mc/practicalqme. pdf, 2023

work page 2023
[30]

Efficient risk estimation via nested mul- tilevel quasi-Monte Carlo simulation.Journal of Computational and Applied Mathematics, 443:115745, 2024

Zhenghang Xu, Zhijian He, and Xiaoqun Wang. Efficient risk estimation via nested mul- tilevel quasi-Monte Carlo simulation.Journal of Computational and Applied Mathematics, 443:115745, 2024. 25

work page 2024