/src/quantlib/ql/pricingengines/mcsimulation.hpp

Source
/* -*- mode: c++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */

/*
 Copyright (C) 2003 Ferdinando Ametrano
 Copyright (C) 2000, 2001, 2002, 2003 RiskMap srl
 Copyright (C) 2007 StatPro Italia srl

 This file is part of QuantLib, a free-software/open-source library
 for financial quantitative analysts and developers - http://quantlib.org/

 QuantLib is free software: you can redistribute it and/or modify it
 under the terms of the QuantLib license.  You should have received a
 copy of the license along with this program; if not, please email
 <quantlib-dev@lists.sf.net>. The license is also available online at
 <https://www.quantlib.org/license.shtml>.

 This program is distributed in the hope that it will be useful, but WITHOUT
 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 FOR A PARTICULAR PURPOSE.  See the license for more details.
*/

/*! \file mcsimulation.hpp
    \brief framework for Monte Carlo engines
*/

#ifndef quantlib_montecarlo_engine_hpp
#define quantlib_montecarlo_engine_hpp

#include <ql/methods/montecarlo/montecarlomodel.hpp>

namespace QuantLib {

    //! base class for Monte Carlo engines
    /*! Eventually this class might offer greeks methods.  Deriving a
        class from McSimulation gives an easy way to write a Monte
        Carlo engine.

        See McVanillaEngine as an example.
    */

    template <template <class> class MC, class RNG, class S = Statistics>
    class McSimulation {
      public:
        typedef typename MonteCarloModel<MC,RNG,S>::path_generator_type
            path_generator_type;
        typedef typename MonteCarloModel<MC,RNG,S>::path_pricer_type
            path_pricer_type;
        typedef typename MonteCarloModel<MC,RNG,S>::stats_type
            stats_type;
        typedef typename MonteCarloModel<MC,RNG,S>::result_type result_type;

        virtual ~McSimulation() = default;
        //! add samples until the required absolute tolerance is reached
        result_type value(Real tolerance,
                          Size maxSamples = QL_MAX_INTEGER,
                          Size minSamples = 1023) const;
        //! simulate a fixed number of samples
        result_type valueWithSamples(Size samples) const;
        //! error estimated using the samples simulated so far
        result_type errorEstimate() const;
        //! access to the sample accumulator for richer statistics
        const stats_type& sampleAccumulator() const;
        //! basic calculate method provided to inherited pricing engines
        void calculate(Real requiredTolerance,
                       Size requiredSamples,
                       Size maxSamples) const;
      protected:
        McSimulation(bool antitheticVariate,
                     bool controlVariate)
        : antitheticVariate_(antitheticVariate),
          controlVariate_(controlVariate) {}
        virtual ext::shared_ptr<path_pricer_type> pathPricer() const = 0;
        virtual ext::shared_ptr<path_generator_type> pathGenerator()
                                                                   const = 0;
        virtual TimeGrid timeGrid() const = 0;
        virtual ext::shared_ptr<path_pricer_type> controlPathPricer() const {
            return ext::shared_ptr<path_pricer_type>();
        }
        virtual ext::shared_ptr<path_generator_type> 
        controlPathGenerator() const {
            return ext::shared_ptr<path_generator_type>();
        }
        virtual ext::shared_ptr<PricingEngine> controlPricingEngine() const {
            return ext::shared_ptr<PricingEngine>();
        }
        virtual result_type controlVariateValue() const {
            return Null<result_type>();
        }
        template <class Sequence>
        static Real maxError(const Sequence& sequence) {
            return *std::max_element(sequence.begin(), sequence.end());
        }
        static Real maxError(Real error) {
            return error;
        }
        
        mutable ext::shared_ptr<MonteCarloModel<MC,RNG,S> > mcModel_;
        bool antitheticVariate_, controlVariate_;
    };


    // inline definitions
    template <template <class> class MC, class RNG, class S>
    inline typename McSimulation<MC,RNG,S>::result_type
        McSimulation<MC,RNG,S>::value(Real tolerance,
                                              Size maxSamples,
                                              Size minSamples) const {
        Size sampleNumber =
            mcModel_->sampleAccumulator().samples();
        if (sampleNumber<minSamples) {
            mcModel_->addSamples(minSamples-sampleNumber);
            sampleNumber = mcModel_->sampleAccumulator().samples();
        }

        Size nextBatch;
        Real order;
        result_type error(mcModel_->sampleAccumulator().errorEstimate());
        while (maxError(error) > tolerance) {
            QL_REQUIRE(sampleNumber<maxSamples,
                       "max number of samples (" << maxSamples
                       << ") reached, while error (" << error
                       << ") is still above tolerance (" << tolerance << ")");

            // conservative estimate of how many samples are needed
            order = maxError(Real(error*error))/tolerance/tolerance;
            nextBatch =
                Size(std::max<Real>(static_cast<Real>(sampleNumber)*order*0.8 - static_cast<Real>(sampleNumber),
                                    static_cast<Real>(minSamples)));

            // do not exceed maxSamples
            nextBatch = std::min(nextBatch, maxSamples-sampleNumber);
            sampleNumber += nextBatch;
            mcModel_->addSamples(nextBatch);
            error = result_type(mcModel_->sampleAccumulator().errorEstimate());
        }

        return result_type(mcModel_->sampleAccumulator().mean());
    }


    template <template <class> class MC, class RNG, class S>
    inline typename McSimulation<MC,RNG,S>::result_type
        McSimulation<MC,RNG,S>::valueWithSamples(Size samples) const {

        Size sampleNumber = mcModel_->sampleAccumulator().samples();

        QL_REQUIRE(samples>=sampleNumber,
                   "number of already simulated samples (" << sampleNumber
                   << ") greater than requested samples (" << samples << ")");

        mcModel_->addSamples(samples-sampleNumber);

        return result_type(mcModel_->sampleAccumulator().mean());
    }


    template <template <class> class MC, class RNG, class S>
    inline void McSimulation<MC,RNG,S>::calculate(Real requiredTolerance,
                                                  Size requiredSamples,
                                                  Size maxSamples) const {

        QL_REQUIRE(requiredTolerance != Null<Real>() ||
                   requiredSamples != Null<Size>(),
                   "neither tolerance nor number of samples set");

        //! Initialize the one-factor Monte Carlo
        if (this->controlVariate_) {

            result_type controlVariateValue = this->controlVariateValue();
            QL_REQUIRE(controlVariateValue != Null<result_type>(),
                       "engine does not provide "
                       "control-variation price");

            ext::shared_ptr<path_pricer_type> controlPP =
                this->controlPathPricer();
            QL_REQUIRE(controlPP,
                       "engine does not provide "
                       "control-variation path pricer");

            ext::shared_ptr<path_generator_type> controlPG = 
                this->controlPathGenerator();

            this->mcModel_ =
                ext::shared_ptr<MonteCarloModel<MC,RNG,S> >(
                    new MonteCarloModel<MC,RNG,S>(
                           pathGenerator(), this->pathPricer(), stats_type(),
                           this->antitheticVariate_, controlPP,
                           controlVariateValue, controlPG));
        } else {
            this->mcModel_ =
                ext::shared_ptr<MonteCarloModel<MC,RNG,S> >(
                    new MonteCarloModel<MC,RNG,S>(
                           pathGenerator(), this->pathPricer(), S(),
                           this->antitheticVariate_));
        }

        if (requiredTolerance != Null<Real>()) {
            if (maxSamples != Null<Size>())
                this->value(requiredTolerance, maxSamples);
            else
                this->value(requiredTolerance);
        } else {
            this->valueWithSamples(requiredSamples);
        }

    }

    template <template <class> class MC, class RNG, class S>
    inline typename McSimulation<MC,RNG,S>::result_type
        McSimulation<MC,RNG,S>::errorEstimate() const {
        return mcModel_->sampleAccumulator().errorEstimate();
    }

    template <template <class> class MC, class RNG, class S>
    inline const typename McSimulation<MC,RNG,S>::stats_type&
    McSimulation<MC,RNG,S>::sampleAccumulator() const {
        return mcModel_->sampleAccumulator();
    }

}


#endif

Coverage Report

Created: 2026-06-23 06:40

Line	Count	Source
1		/* -- mode: c++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -- */
2
3		/*
4		Copyright (C) 2003 Ferdinando Ametrano
5		Copyright (C) 2000, 2001, 2002, 2003 RiskMap srl
6		Copyright (C) 2007 StatPro Italia srl
7
8		This file is part of QuantLib, a free-software/open-source library
9		for financial quantitative analysts and developers - http://quantlib.org/
10
11		QuantLib is free software: you can redistribute it and/or modify it
12		under the terms of the QuantLib license. You should have received a
13		copy of the license along with this program; if not, please email
14		<quantlib-dev@lists.sf.net>. The license is also available online at
15		<https://www.quantlib.org/license.shtml>.
16
17		This program is distributed in the hope that it will be useful, but WITHOUT
18		ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
19		FOR A PARTICULAR PURPOSE. See the license for more details.
20		*/
21
22		/*! \file mcsimulation.hpp
23		\brief framework for Monte Carlo engines
24		*/
25
26		#ifndef quantlib_montecarlo_engine_hpp
27		#define quantlib_montecarlo_engine_hpp
28
29		#include <ql/methods/montecarlo/montecarlomodel.hpp>
30
31		namespace QuantLib {
32
33		//! base class for Monte Carlo engines
34		/*! Eventually this class might offer greeks methods. Deriving a
35		class from McSimulation gives an easy way to write a Monte
36		Carlo engine.
37
38		See McVanillaEngine as an example.
39		*/
40
41		template <template <class> class MC, class RNG, class S = Statistics>
42		class McSimulation {
43		public:
44		typedef typename MonteCarloModel<MC,RNG,S>::path_generator_type
45		path_generator_type;
46		typedef typename MonteCarloModel<MC,RNG,S>::path_pricer_type
47		path_pricer_type;
48		typedef typename MonteCarloModel<MC,RNG,S>::stats_type
49		stats_type;
50		typedef typename MonteCarloModel<MC,RNG,S>::result_type result_type;
51
52	709	virtual ~McSimulation() = default;
53		//! add samples until the required absolute tolerance is reached
54		result_type value(Real tolerance,
55		Size maxSamples = QL_MAX_INTEGER,
56		Size minSamples = 1023) const;
57		//! simulate a fixed number of samples
58		result_type valueWithSamples(Size samples) const;
59		//! error estimated using the samples simulated so far
60		result_type errorEstimate() const;
61		//! access to the sample accumulator for richer statistics
62		const stats_type& sampleAccumulator() const;
63		//! basic calculate method provided to inherited pricing engines
64		void calculate(Real requiredTolerance,
65		Size requiredSamples,
66		Size maxSamples) const;
67		protected:
68		McSimulation(bool antitheticVariate,
69		bool controlVariate)
70	709	: antitheticVariate_(antitheticVariate),
71	709	controlVariate_(controlVariate) {}
72		virtual ext::shared_ptr<path_pricer_type> pathPricer() const = 0;
73		virtual ext::shared_ptr<path_generator_type> pathGenerator()
74		const = 0;
75		virtual TimeGrid timeGrid() const = 0;
76	0	virtual ext::shared_ptr<path_pricer_type> controlPathPricer() const {
77	0	return ext::shared_ptr<path_pricer_type>();
78	0	}
79		virtual ext::shared_ptr<path_generator_type>
80	0	controlPathGenerator() const {
81	0	return ext::shared_ptr<path_generator_type>();
82	0	}
83	0	virtual ext::shared_ptr<PricingEngine> controlPricingEngine() const {
84	0	return ext::shared_ptr<PricingEngine>();
85	0	}
86	0	virtual result_type controlVariateValue() const {
87	0	return Null<result_type>();
88	0	}
89		template <class Sequence>
90		static Real maxError(const Sequence& sequence) {
91		return *std::max_element(sequence.begin(), sequence.end());
92		}
93	0	static Real maxError(Real error) {
94	0	return error;
95	0	}
96
97		mutable ext::shared_ptr<MonteCarloModel<MC,RNG,S> > mcModel_;
98		bool antitheticVariate_, controlVariate_;
99		};
100
101
102		// inline definitions
103		template <template <class> class MC, class RNG, class S>
104		inline typename McSimulation<MC,RNG,S>::result_type
105		McSimulation<MC,RNG,S>::value(Real tolerance,
106		Size maxSamples,
107	0	Size minSamples) const {
108	0	Size sampleNumber =
109	0	mcModel_->sampleAccumulator().samples();
110	0	if (sampleNumber<minSamples) {
111	0	mcModel_->addSamples(minSamples-sampleNumber);
112	0	sampleNumber = mcModel_->sampleAccumulator().samples();
113	0	}
114
115	0	Size nextBatch;
116	0	Real order;
117	0	result_type error(mcModel_->sampleAccumulator().errorEstimate());
118	0	while (maxError(error) > tolerance) {
119	0	QL_REQUIRE(sampleNumber<maxSamples,
120	0	"max number of samples (" << maxSamples
121	0	<< ") reached, while error (" << error
122	0	<< ") is still above tolerance (" << tolerance << ")");
123
124		// conservative estimate of how many samples are needed
125	0	order = maxError(Real(error*error))/tolerance/tolerance;
126	0	nextBatch =
127	0	Size(std::max<Real>(static_cast<Real>(sampleNumber)order0.8 - static_cast<Real>(sampleNumber),
128	0	static_cast<Real>(minSamples)));
129
130		// do not exceed maxSamples
131	0	nextBatch = std::min(nextBatch, maxSamples-sampleNumber);
132	0	sampleNumber += nextBatch;
133	0	mcModel_->addSamples(nextBatch);
134	0	error = result_type(mcModel_->sampleAccumulator().errorEstimate());
135	0	}
136
137	0	return result_type(mcModel_->sampleAccumulator().mean());
138	0	}
139
140
141		template <template <class> class MC, class RNG, class S>
142		inline typename McSimulation<MC,RNG,S>::result_type
143	709	McSimulation<MC,RNG,S>::valueWithSamples(Size samples) const {
144
145	709	Size sampleNumber = mcModel_->sampleAccumulator().samples();
146
147	709	QL_REQUIRE(samples>=sampleNumber,
148	709	"number of already simulated samples (" << sampleNumber
149	709	<< ") greater than requested samples (" << samples << ")");
150
151	709	mcModel_->addSamples(samples-sampleNumber);
152
153	709	return result_type(mcModel_->sampleAccumulator().mean());
154	709	}
155
156
157		template <template <class> class MC, class RNG, class S>
158		inline void McSimulation<MC,RNG,S>::calculate(Real requiredTolerance,
159		Size requiredSamples,
160	709	Size maxSamples) const {
161
162	709	QL_REQUIRE(requiredTolerance != Null<Real>() \|\|
163	709	requiredSamples != Null<Size>(),
164	709	"neither tolerance nor number of samples set");
165
166		//! Initialize the one-factor Monte Carlo
167	709	if (this->controlVariate_) {
168
169	0	result_type controlVariateValue = this->controlVariateValue();
170	0	QL_REQUIRE(controlVariateValue != Null<result_type>(),
171	0	"engine does not provide "
172	0	"control-variation price");
173
174	0	ext::shared_ptr<path_pricer_type> controlPP =
175	0	this->controlPathPricer();
176	0	QL_REQUIRE(controlPP,
177	0	"engine does not provide "
178	0	"control-variation path pricer");
179
180	0	ext::shared_ptr<path_generator_type> controlPG =
181	0	this->controlPathGenerator();
182
183	0	this->mcModel_ =
184	0	ext::shared_ptr<MonteCarloModel<MC,RNG,S> >(
185	0	new MonteCarloModel<MC,RNG,S>(
186	0	pathGenerator(), this->pathPricer(), stats_type(),
187	0	this->antitheticVariate_, controlPP,
188	0	controlVariateValue, controlPG));
189	709	} else {
190	709	this->mcModel_ =
191	709	ext::shared_ptr<MonteCarloModel<MC,RNG,S> >(
192	709	new MonteCarloModel<MC,RNG,S>(
193	709	pathGenerator(), this->pathPricer(), S(),
194	709	this->antitheticVariate_));
195	709	}
196
197	709	if (requiredTolerance != Null<Real>()) {
198	0	if (maxSamples != Null<Size>())
199	0	this->value(requiredTolerance, maxSamples);
200	0	else
201	0	this->value(requiredTolerance);
202	709	} else {
203	709	this->valueWithSamples(requiredSamples);
204	709	}
205
206	709	}
207
208		template <template <class> class MC, class RNG, class S>
209		inline typename McSimulation<MC,RNG,S>::result_type
210		McSimulation<MC,RNG,S>::errorEstimate() const {
211		return mcModel_->sampleAccumulator().errorEstimate();
212		}
213
214		template <template <class> class MC, class RNG, class S>
215		inline const typename McSimulation<MC,RNG,S>::stats_type&
216		McSimulation<MC,RNG,S>::sampleAccumulator() const {
217		return mcModel_->sampleAccumulator();
218		}
219
220		}
221
222
223		#endif