Add `BlackVolatilitySurfaceDelta` class for FX options by paolodelia99 · Pull Request #2368 · lballabio/QuantLib

paolodelia99 · 2025-11-06T16:36:26Z

Added BlackVolatilitySurfaceDelta class from ORE, a Black volatility surface parameterized by market deltas, mainly used in Building FX/equity volatility surfaces from market delta quotes.

QL changes @lballabio:

BlackVolatilitySurfaceDelta class
time extrapolation logic imported from the ORE's QL fork
added the possibility to a have a flat extrapolation in the strike space in the InterpolatedSmileSection class as an additional constructor parameter, required from the BlackVolatilitySurfaceDelta class and from ORE's implementation of InterpolatedSmileSection class
strikes sanity check added in the InterepolatedSmileSection class (logic implemented in checkStrikes() private method), thought you guys might need it since it was underlined in the comments

ORE Changes @pcaspers:

used the QL's InterpolatedSmileSection class to retrive the SmileSection from the blackVolSmile method in the BlackVolatilitySurfaceDelta class instead of the QLE InterpolatedSmileSection since they provide the same functionalities, (QL's InterpolatedSmileSection is actually more complete than the QLE ones, but the latter is derived form FxSmileSection not present in QL) thus there no need to introduce another redundant class.

…ility

…e> for the vols

…mileSection

… iterpolateSmileSection

coveralls · 2025-11-06T17:19:04Z

coverage: 74.273% (+0.007%) from 74.266%
when pulling d9af61e on paolodelia99:feature/fx-options-utils
into 0a02cce on lballabio:master.

lballabio

Thanks! A few comments, some of them for @pcaspers.

ql/termstructures/volatility/equityfx/blackvariancetimeextrapolation.hpp

lballabio · 2025-11-11T10:22:55Z

ql/termstructures/volatility/equityfx/blackvariancetimeextrapolation.hpp

+
+
+    //! Extrapolate black variance using flat vol extrapolation in time direction
+    Real timeExtrapolatationBlackVarianceFlat(const Time t, const std::vector<double>& times,


"extrapolatation" (here and everywhere)

lballabio · 2025-11-11T10:27:18Z

ql/termstructures/volatility/equityfx/blackvolsurfacedelta.cpp

+ Copyright (C) 2019 Quaternion Risk Management Ltd
+ Copyright (C) 2022 Skandinaviska Enskilda Banken AB (publ)
+ Copyright (C) 2025 Paolo D'Elia
+ All rights reserved.


it seems this is outdated, see here - feel free to remove this line in this PR. We will update the ORE files accordingly.

lballabio · 2025-11-11T10:29:28Z

ql/termstructures/volatility/equityfx/blackvolsurfacedelta.hpp

+                                    DeltaVolQuote::DeltaType dt = DeltaVolQuote::DeltaType::Spot,
+                                    DeltaVolQuote::AtmType at = DeltaVolQuote::AtmType::AtmDeltaNeutral,
+                                    ext::optional<DeltaVolQuote::DeltaType> atmDeltaType = ext::nullopt,
+                                    const Period& switchTenor = 0 * Days,
+                                    DeltaVolQuote::DeltaType ltdt = DeltaVolQuote::DeltaType::Fwd,
+                                    DeltaVolQuote::AtmType ltat = DeltaVolQuote::AtmType::AtmDeltaNeutral,
+                                    ext::optional<DeltaVolQuote::DeltaType> longTermAtmDeltaType = ext::nullopt,
+                                    SmileInterpolationMethod interpolationMethod =
+                                        SmileInterpolationMethod::Linear,
+                                    bool flatStrikeExtrapolation = false,
+                                    BlackVolTimeExtrapolation timeExtrapolation =
+                                        BlackVolTimeExtrapolation::FlatVolatility);


Lots of default parameters here—@pcaspers, is their order based on your usage patterns? (i.e., are the last ones also the least likely to be passed?)

No I don't think so to be honest, they were added over time at the back to keep backwards compatibility. Feel free to reorder them if that seems more reasonable?

I have reorder the optional parameters in the following way:

Short-term conventions (dt_, at_, atmDeltaType_)

Interpolation & extrapolation (interpolationMethod_, flatStrikeExtrapolation_, timeExtrapolation_)

Long-term conventions (switchTenor_, ltdt_, ltat_, longTermAtmDeltaType_)

It made more sense to me. Let me know what do you think.

ql/termstructures/volatility/interpolatedsmilesection.hpp

paolodelia99 · 2025-12-17T07:54:37Z

Whenever you guys have time @lballabio @pcaspers

Made mistake by using vols instead of stdDevs when building SmileSection in the test suite

github-actions · 2026-02-17T02:17:50Z

This PR was automatically marked as stale because it has been open 60 days with no activity. Remove stale label or comment, or this will be closed in two weeks.

lballabio · 2026-02-17T08:11:59Z

@pcaspers — there were a couple of questions for you here if you have some time. Thanks!

pcaspers · 2026-02-17T11:35:07Z

this is fine, we can move to ql's interpolated smilesection and remove fxsmilesection in ore:

ORE Changes @pcaspers:

used the QL's InterpolatedSmileSection class to retrive the SmileSection from the blackVolSmile method in the BlackVolatilitySurfaceDelta class instead of the QLE InterpolatedSmileSection since they provide the same functionalities, (QL's InterpolatedSmileSection is actually more complete than the QLE ones, but the latter is derived form FxSmileSection not present in QL) thus there no need to introduce another redundant class.

paolodelia99 · 2026-02-18T19:02:12Z

@pcaspers actually I think you still need the FxSmileSection for the VannaVolgaSmileSection since it needs the spot, rd, and rf to calculate the vols. But in this case the InterpolatedSmileSection in ORE was redudant to the one already present in QuantLib.

lballabio · 2026-03-02T15:58:09Z

ql/termstructures/volatility/equityfx/blackvariancetimeextrapolation.hpp

+
+
+    //! Extrapolate black variance using flat vol extrapolation in time direction
+    Real timeExtrapolationBlackVarianceFlat(const Time t, const std::vector<double>& times,


The functions in this file seem to belong together with BlackVolTimeExtrapolation. Instead of having an enum class and a separate set of functions, How about something like

class BlackVolTimeExtrapolation { public: enum Type { FlatVolatility, UseInterpolatorVariance, UseInterpolatorVolatility }; static Real extrapolate(Type type, const Time t, const std::vector<double>& times, etc); };

where extrapolate calls one of these functions (that can be declared in namespace detail) based on type?

Yeah, actually it's more elegant to have a class to do that, then just those plain functions.

lballabio · 2026-03-02T15:59:12Z

ql/termstructures/volatility/equityfx/blackvariancecurve.cpp

-            return varianceCurve_(times_.back(), true)*t/times_.back();
+            return timeExtrapolationBlackVarianceFlat(t, times_, varianceCurve_);
+        } else if (timeExtrapolation_ == BlackVolTimeExtrapolation::UseInterpolatorVolatility) {
+            return timeExtrapolationBlackVarianceInVolatility(t, times_, varianceCurve_);


I would have expected this case to use the existing interpolator, but timeExtrapolationBlackVarianceInVolatility always uses linear interpolation. Maybe the enum name is misleading?

I guess, it's because if you interpolate/extrapolate variances linearly you are sure that you are not going to have calendar spreads.

Yes, I think so, but then the enum should be something like "LinearVariance", not "UseInterpolatorVolatility".

…ad of plain functions

lballabio · 2026-03-12T15:14:58Z

Thanks, Paolo. I'll probably make some more refactoring to wrap my head around this and push it.

lballabio · 2026-03-12T15:15:04Z

@pcaspers : a question outside the scope of this PR (otherwise we'll never merge it). Given the recent work in #2443 and #2478, is there any difference between this class and what we would get from instantiating PiecewiseBlackVarianceSurface with smiles based on the same data?

lballabio · 2026-03-12T17:15:08Z

ql/termstructures/volatility/equityfx/blackvoltermstructure.cpp

+            vols[0] = close_enough(times[0], 0.0) ? 0.0 : std::sqrt(variances[0] / times[0]);
+            vols[1] = close_enough(times[1], 0.0) ? 0.0 : std::sqrt(variances[1] / times[1]);
+            LinearInterpolation interpolation(times.begin(), times.end(), vols.begin());
+            return std::max(interpolation(t, true), 0.0);


This interpolates linearly between volatilities. It should interpolate linearly between variances, shouldn't it?

yes, an oversight of mine. I thought it was interpolating variances, for the non arb reasons.

No problem, I'll modify it while I'm at it.

Also, use std::function to avoid templates (allowing us to keep the implementation in the cpp file).

paolodelia99 · 2026-03-12T17:41:57Z

@pcaspers : a question outside the scope of this PR (otherwise we'll never merge it). Given the recent work in #2443 and #2478, is there any difference between this class and what we would get from instantiating PiecewiseBlackVarianceSurface with smiles based on the same data?

I guess it's fine @lballabio. We just need to update the code that is initializing the interpolators_ in the ctor. If we also want to keep the current ctor then we can use the makeFromGrid static method in the PiecewiseBlackVarianceSurface class.

lballabio · 2026-03-12T17:53:08Z

I guess it's fine @lballabio. We just need to update the code that is initializing the interpolators_ in the ctor. If we also want to keep the current ctor then we can use the makeFromGrid static method in the PiecewiseBlackVarianceSurface class.

Not sure if I'm getting it — the interpolators are constant delta, the piecewise surface works for a set of smiles which are constant time. It would be more like replacing the ones with the others, I guess?

paolodelia99 added 9 commits October 12, 2025 22:07

Added first fx vol utils classes from ORE

490046d

Add unit test for BlackVolatilitySurfaceDelta with non-constant volat…

be091ed

…ility

make InterpolatedFxSmileSection lazy, possibility to pass handle<quot…

5d346f8

…e> for the vols

Fix compilation errors

93c57cb

Replace InterpolatedFxSmileSection with already present InterpolatedS…

dddeafd

…mileSection

renaming vars, added docs string for the blackvolsurfacedelta

94f4762

set flatStrikeExtrapolation to false, add checkStrike sanity check in…

3e61f8d

… iterpolateSmileSection

fix win compilation error, in generate doc error

9a58b8a

Updated initial comment in the tests

813c1bf

lballabio reviewed Nov 11, 2025

View reviewed changes

paolodelia99 added 2 commits November 27, 2025 13:13

Fix various oversights and add interpolatedsmilesection first tests

ff66504

no anonymous namespace anymore, removed #pragma once

8e6afe6

paolodelia99 and others added 3 commits December 18, 2025 12:57

Correct logical error when creating InterpolatedSmileSection in tests

856b096

Made mistake by using vols instead of stdDevs when building SmileSection in the test suite

Update old license links

57372ce

Update copyright list in license

12c44d1

github-actions bot added the stale label Feb 17, 2026

lballabio removed the stale label Feb 17, 2026

paolodelia99 mentioned this pull request Feb 17, 2026

Add PiecewiseBlackVarianceSurface for ragged vol grids #2443

Merged

Reorder some params of the blackvolsurfacedelta ctor

2494b90

lballabio added 3 commits February 23, 2026 10:32

Remove obsolete rights attribution

239504d

Merge branch 'master' into feature/fx-options-utils

ce07bb4

Fix typo in function names

ddfc9be

lballabio reviewed Mar 2, 2026

View reviewed changes

Refactoring according to suggestions: BlackVolTimeExtrapolation inste…

a068265

…ad of plain functions

Fix docstring in blackvolsurfacedelta class

d74ebac

lballabio reviewed Mar 12, 2026

View reviewed changes

Move BlackVolTimeExtrapolation to own file.

859f97b

Also, use std::function to avoid templates (allowing us to keep the implementation in the cpp file).

lballabio added 11 commits March 12, 2026 18:55

Don't make the common case look like an early exit

7ac8677

Add missing header

19a597a

Use linear extrapolation in variance

38da695

Merge branch 'master' into feature/fx-options-utils

becd335

Clean up inclusions

cd132fc

Let the inner interpolator extrapolate when needed

3881c9c

Standard syntax for enum

a84e248

Time extrapolation tests

86a07d8

Prevent negative volatility when extrapolating

569052a

Add smile interpolation/extrapolation test

e344bfc

Clean up test messages

d9af61e

lballabio changed the title ~~Fx options utils - BlackVolatilitySurfaceDelta class~~ Add BlackVolatilitySurfaceDelta class for FX options Mar 13, 2026

lballabio added this to the Release 1.42 milestone Mar 13, 2026

lballabio merged commit 7f06be1 into lballabio:master Mar 13, 2026
43 checks passed



		//! Extrapolate black variance using flat vol extrapolation in time direction
		Real timeExtrapolatationBlackVarianceFlat(const Time t, const std::vector<double>& times,



		//! Extrapolate black variance using flat vol extrapolation in time direction
		Real timeExtrapolationBlackVarianceFlat(const Time t, const std::vector<double>& times,

Conversation

paolodelia99 commented Nov 6, 2025

Uh oh!

coveralls commented Nov 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lballabio left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

paolodelia99 commented Dec 17, 2025

Uh oh!

github-actions bot commented Feb 17, 2026

Uh oh!

lballabio commented Feb 17, 2026

Uh oh!

pcaspers commented Feb 17, 2026

Uh oh!

paolodelia99 commented Feb 18, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lballabio commented Mar 12, 2026

Uh oh!

lballabio commented Mar 12, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

paolodelia99 Mar 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

paolodelia99 commented Mar 12, 2026

Uh oh!

lballabio commented Mar 12, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

coveralls commented Nov 6, 2025 •

edited

Loading

paolodelia99 Mar 12, 2026 •

edited

Loading