The Singular Points Binomial Method for pricing

American path-dependent options

Marcellino Gaudenzi, Antonino Zanette
Dipartimento di Finanza dell’Impresa e dei Mercati Finanziari
Via Tomadini 30/A, Universitá di Udine, Italy
E-Mail: [Link]@[Link] , [Link]@[Link]

Maria Antonietta Lepellere

Dipartimento di Biologia ed Economia Agro-Industriale
Via delle Scienze 208, Universitá di Udine, Italy

Premia 22
Abstract
We introduce a new numerical approach, called "Singular Points Method", for pricing
American path-dependent options. This method, based on a continuous representation
of the price at each node of the binomial tree, allows us to obtain very precise upper
and lower bounds of the discrete binomial price. Moreover, the method provides a-priori
estimates of the difference between upper and lower bounds. The algorithm is convergent
and provides efficient estimates of the continuous price value. We apply the method to
the case of Asian and lookback American options.

Keywords: option pricing, American options, Asian options, lookback options, tree methods.

Introduction
A path-dependent option is an option whose payoff depends not only on the value of the stock
price at maturity but also on the past history of the underlying asset price. In this paper we
are mainly interested in the case of Asian and lookback options.
The pay-off of an Asian option is based on several forms of averaging of the underlying asset
price over the life of the option. The most common cases are those for which the average is
arithmetic or geometric. Lookback options are options whose payoff depend on the maximum
or on the minimum of the underlying asset price reached during the life of the option.
American lookback and American Asian options cannot be valued by closed-form formulae,
even in the Black-Scholes model, and their valuation requires the use of numerical methods.
Here we consider tree methods for pricing these type of options.

1
 The difficulty of applying Cox-Ross-Rubinstein (CRR) method to Asian options with arith-
metic average is well known. This is because the number of possible averages increases expo-
nentially with the number of tree steps. For this reason Hull and White ([7]) and similarly
Barraquand-Pudet ([2]), proposed more feasible approaches. The main idea behind these pro-
cedures is to restrict the range of all the possible arithmetic averages to a set of representative
values. These values are selected in order to span all the possible values of the averages achiev-
able at each node of the tree. The price is then computed by a backward induction procedure
in which the prices associated to the averages not included in the set of representative values,
are obtained by interpolation.
In comparison with the CRR binomial method, these two techniques significantly reduce the
number of computations. In fact the computational complexity of both methods is O(n3 )
(where n is the number of tree steps). However, these techniques have some drawbacks related
both to the precision of the approximations and to the convergence to the continuous value,
7
as observed by Forsyth et al in [11]. Forsyth et al proved that a procedure of order O(n 2 ) is
necessary in order to assure the convergence of these algorithms.
Later Chalasani et al ([3], [4]) proposed a totally different approach which allowed them
to obtain thin upper and lower bounds of the exact CRR binomial price for American Asian
options. Their method requires a forward procedure and a backward induction. This algorithm
significantly increases the precision of the estimates but it requires a very large amount of
memory and has computational complexity O(n4 ).
More recently, very efficient PDE-based methods have been introduced by Vecer[10] and
D’Halluin et al [6]. Vecer proposed a one-dimensional PDE method that runs in O(n2 ). This
approach cannot be applied to American fixed strike Asian options, which, on the other hand,
can be treated using the semi-lagrangian approach of D’Halluin et al.
As regards lookback options, the complexity of the exact CRR binomial algorithm is of
order O(n3 ) and the methods proposed in [7] and [2] do not improve the efficiency. Babbs
([1]) gave an efficient and accurate solution to the problem of American floating strike lookback
options through a procedure of complexity of order O(n2 ). He used a change of “numeraire”
approach, which cannot be applied in the fixed strike case.
In this paper we will introduce a general binomial framework for pricing European/American
path-dependent options in a efficient way. In particular, we apply it for the pricing of both
American Asian and American lookback options.
The method provides very precise upper and lower bounds for the exact binomial discrete
value and it significantly reduces the time of computation with respect to the previous tree
techniques.
The main idea is to give a continuous representation, at each node of the tree, of the
option prices as a piecewise linear convex function of the path-dependent variable (average
or maximum/minimum). These functions are characterized just by a set of points, which
we name ”singular points”. All such functions can be evaluated by backward induction in
a straightforward way. Consequently the method provides an alternative and more efficient
approach to evaluate the exact binomial price associated with the path-dependent options.
Moreover, the convexity property of the piecewise linear function representing the price, allows
us to obtain, simply and naturally upper and lower bounds of the discrete binomial price. A
further appeal of the procedure is that it is possible to fix an a-priori level of precision for the

2
distance between the estimates and the exact binomial value. This can be done very efficiently
keeping the amount of time and memory space at low level. Moreover, the error control process
permits to automatically obtain the convergence of the approximations to the continuous value.
The choice of providing an a-priori control of the option price error in the discrete model
gives rise to problems in determining the theoretical complexity of the procedure. Nevertheless,
the numerical experiments show that the method is very competitive in practice. Moreover,
the observed complexity is O(n3 ).
The paper is organized as follows: in Section 1 we will describe the standard binomial
techniques for American Asian and lookback options. Section 2 is devoted to the singular points
method in the Asian case, including a description of the implementation of the algorithm. In
Section 3 we will propose an algorithm for American lookback options. Finally, in Section 4,
we will compare our technique with the best lattice based methods known in the literature.
Furthermore, we will study the convergence of our method to the continuous price value by
comparing it with the PDE-based methods.

1 The exact binomial algorithm

In this paper, we consider a market model where the evolution of a risky asset is governed by
the Black-Scholes stochastic differential equation
dSt
= (r − q)dt + σdBt , S0 = s 0 , (1)
St
where (Bt )0≤t≤T is a standard Brownian motion, under the risk neutral measure Q. The
nonnegative constant r is the force of interest rate, q is the continuous dividend yields and
σ is the volatility of the risky asset. Then ST is the value of the underlying asset at maturity
T:
σ2
ST = s0 e(r−q− 2 )T +σBT .
We will consider two examples of path-dependent options written on the underlying St : arith-
metic Asian options and lookback options.

1.1 American Asian options

The price of an American Asian option of initial time 0 and maturity T is:
h i
P (0, S0 , A0 ) = sup E e−rτ ψ(Sτ , Aτ )|S0 = s0 , A0 = s0 ,
τ ∈T0,T

where: T0,T is the set of all stopping times with values in [0, T ], ψ denotes the payoff function
and Aτ Ris the arithmetic average of the price of the underlying asset over the period [0, τ ], i.e.
Aτ = τ1 0τ St dt.
Let K be the strike price. Some examples of payoff functions useful for Asian options pricing
are:

• Fixed Asian Call: the payoff is (AT − K)+

3
 • Fixed Asian Put: the payoff is (K − AT )+

• Floating Asian Call: the payoff is (ST − AT )+

• Floating Asian Put: the payoff is (AT − ST )+ .

Consider now the binomial approach. Let n be the number of steps of the binomial tree
and ∆T = Tn the corresponding time-step. The lognormal diffusion process (Si∆T )0≤i≤n is
approximated by the Cox-Ross-Rubinstein binomial process
i
Y
Si = (s0 Yj )0≤i≤n
j=1

where the random variables Y1 , . . . , Yn are independent and identically distributed with values
in {d, u}. Let us denote
√
by π = P(Yn = u). The Cox-Ross-Rubinstein tree corresponds to the
1 σ ∆T
choice u = d = e and √
er∆T − e−σ ∆T
π= √ √
eσ ∆T − e−σ ∆T
In a discrete-time setting, the payoff function at maturity n of an Asian option is given by
f (Sn , An ) where
n
1 X
An = Si
n + 1 i=0
and the average process (Ai )0≤i≤n is recursively computed by

(i + 1)Ai + Si+1
Ai+1 = , A0 = s 0 .
i+2
In the Cox-Ross-Rubinstein model, the price at time 0 of the American (resp. European)
Asian option with payoff function ψ is given by v(0, s0 , s0 ) where the functions v(i, x, y) can be
computed by the following backward dynamic programming equations


v(n, x, y) = ψ(x, y)
 h 
 ′ −r∆T (i + 1)y + xu (i + 1)y + xd i
v(i, x, y) = max ψ (x, y), e πv(i + 1, xu, ) + (1 − π)v(i + 1, xd, ) ,
i+2 i+2
(2)
where ψ ′ = ψ in the American case and ψ ′ ≡ 0 in the European case.
The obtained tree is not recombining so that, from a practical point of view, the valuation
of v(0, s0 , s0 ) is unfeasible just for very small number of steps.

1.2 Lookback options

The price of an American lookback option is:
h i
P (0, S0 , S0∗ ) = sup E e−rτ ψ(Sτ , Sτ∗ )|S0 = s, S0∗ = s .
τ ∈T0,T

4
where ψ denotes the payoff function of the option and

Sτ∗ = Mτ = max Su or Sτ∗ = mτ = min Su

u∈[0,τ ] u∈[0,τ ]

Let K be the strike. Some examples of payoff function useful in lookback option pricing are:

• Fixed lookback Call: the payoff is (MT − K)+ .

• Fixed lookback Put: the payoff is (K − mT )+ .

• Floating lookback Call: the payoff is (ST − mT )+ .

• Floating lookback Put: the payoff is (MT − ST )+ .

In a discrete-time setting, the payoff at maturity n of an European lookback option, written

on the maximum, is given by ψ(Sn , Mn ) where

Mn = max(S0 , ..., Sn ).

The maximum process (Mi )0≤i≤n can be computed recursively by

Mi+1 = max(Mi , Si+1 ), M0 = s0 .

In the Cox-Ross-Rubinstein model, the price at time 0 of the corresponding American lookback
option is given by v(0, s0 , s0 ) where the functions v(i, x, y) can be computed by the following
backward dynamic programming equations


v(n, x, y) = ψ(x, y)
  
 −r∆T (3)
v(i, x, y) = max ψ(x, y), e πv(i + 1, xu, max(xu, y)) + (1 − π)v(i + 1, xd, y) ,

where ψ(x, y) is the payoff function. The valuation of v(0, s0 , s0 ) requires a number of compu-
tations of order O(n3 ).

2 The Singular Points Method

In this section we will introduce a new backward procedure. The main idea is to give a
continuous representation of the option price as a piecewise linear function at each node of
the tree, which describes the path-dependent nature of the option. Such a representation only
depends on a finite number of points (i.e. the points were the slope of the function changes)
called ”singular points”.

2.1 Piecewise linear convex functions and singular points

Henceforth we will use the following notations:

5
Definition 1. Given a set of points: (x1 , y1 ), ..., (xn , yn ), such that a = x1 < x2 < ... < xn = b
and
yi − yi−1 yi+1 − yi
< , i = 2, ..., n − 1, (4)
xi − xi−1 xi+1 − xi
let us consider the function f (x), x ∈ [a, b], obtained by interpolating the given points linearly.
The points (x1 , y1 ), ..., (xn , yn ) (which characterize the piecewise linear function f ), will be called
the singular points of f , while x1 , ..., xn will be called the singular values of f .

Remark 1. In the previous definition we considered only piecewise linear functions with strictly
increasing slopes, this implies that the resulting function f is convex.
From here on we shall consider only piecewise linear functions that are continuous and
convex on the interval [a, b]. For each of these functions we can find a set of singular points
characterizing them and satisfying equation (4).

The following results, which have a very simple geometrical interpretation (see Fig.1 and
Fig.2), allow us to construct the upper and the lower bounds of the discrete option price.

Lemma 1. Let f be a piecewise linear and convex function defined on [a, b], and let C =
{(x1 , y1 ), ..., (xn , yn )} be the set of its singular points.
Removing a point (xi , yi ), 2 ≤ i ≤ n − 1, from the set C, the resulting piecewise linear
function fe, whose set of singular points is C \ {(xi , yi )}, is again convex in [a, b] and we have:

f (x) ≤ fe(x), ∀x ∈ [a, b].

Proof. The previous inequality and the convexity of fe follow from the fact that fe is the max-
imum between f and the function given by the straight line joining the points (xi−1 , yi−1 ),
(xi+1 , yi+1 ). ♦
Remark 2. From the previous Lemma it follows that every piecewise linear function fe whose
singular points are a subset of C (containing the first and the last singular point) is still convex
and satisfies fe ≥ f .
Lemma 2. Let f be a piecewise linear and convex function defined on [a, b], and let C =
{(x1 , y1 ), ..., (xn , yn )} be the set of its singular points. Let us denote by (x, y) the intersection
between the straight line joining (xi−1 , yi−1 ), (xi , yi ) and the one joining (xi+1 , yi+1 ), (xi+2 , yi+2 ),
2 ≤ i ≤ n − 2.
If we consider the new set of n − 1 singular points

{(x1 , y1 ), ..., (xi−1 , yi−1 ), (x, y), (xi+2 , yi+2 ), ..., (xn , yn )},

the associated piecewise linear function fe is again convex on [a, b] and we have:

f (x) ≥ fe(x), ∀x ∈ [a, b].

Proof. The singular points of f satisfy the property of increasing slopes (4). The set of slopes
associated to the singular points of fe are obtained removing the slope of the line joining (xi , yi ),
(xi+1 , yi+1 ), hence (4) is again satisfied and fe is convex. The inequality f ≥ fe is trivial. ♦

6
 Figure 1: Upper estimate: x4 has been removed.

Figure 2: Lower estimate: x3 and x4 have been removed, x has been inserted.

2.2 Fixed strike European Asian options

First at all we will describe the proposed algorithm in the framework of a fixed strike European
Asian call option.
In this case the method consists in valuating the price of the option, at each node of the tree,
for each possible choice of the average at that point. So we consider not only averages which
are effectively achievable, but all the possible averages between the minimum and maximum
realized at that point. In this way, we will show that it is possible to give a continuous
representation of the price function as a piecewise linear convex function of the average. This
function is characterized just by its singular points.
We will now introduce some further notations.
Let us denote by Ni,j the node of the tree whose underlying asset is Si,j = s0 u2j−i , i = 0, ..., n,
j = 0, ..., i.
To each node Ni,j we will associate a set of singular points, whose number is Li,j . The singular

7
points will be denoted by
(Ali,j , Pi,j
l
), l = 1, ..., Li,j .
As regards Asian options, the singular values Ali,j are called singular averages and Pi,j
l
are called
singular prices.
Let us consider first the nodes Nn,j , j = 0, ..., n, of the tree at maturity. At each node the
average values vary between a minimum average Amin n,j (corresponding to the path with n − j
down movements followed by j up movements) and a maximum average Amax n,j (corresponding
to the path with j up movements followed by n − j down movements). These minimum and
maximum are easily valuable:

s0 1 − dn−j+1 1 − uj+1
Amin
n,j = ( + dn−j ( − 1)),
n+1 1−d 1−u
s0 1 − uj+1 1 − dn−j+1
Amax
n,j = ( + uj ( − 1)).
n+1 1−u 1−d
For each A ∈ [Amin max
n,j , An,j ] the price of the option can be continuously defined by vn,j (A) =
(A − K)+
(remark that vn,j (A) ≡ v(n, Sn,j , A) where v(n, x, y), is the price function introduced in Section
1.1).
Note that the function vn,j (A) is a piecewise linear function satisfying Definition 1, whose
singular points are valuable in a straightforward way. In fact:

• if K ∈ (Amin max
n,j , An,j ) then the price value function vn,j (A) is characterized by the 3 singular
l l
points (An,j , Pn,j ), l = 1, 2, 3 (hence Ln,j = 3), where

A1n,j = Amin 1
n,j , Pn,j = 0;
A2n,j = K, 2
Pn,j = 0; (5)
3 3
An,j = An,j , Pn,j = Amax
max
n,j − K.
.

• if K 6∈ (Amin max
n,j , An,j ) then the price value function vn,j (A) is characterized by the 2 singular
points (Aln,j , Pn,j
l
), l = 1, 2, (Ln,j = 2), where

A1n,j = Amin 1 min

n,j , Pn,j = (An,j − K)+ ;
2 2 (6)
An,j = An,j , Pn,j = (Amax
max
n,j − K)+ .

• In the case j = 0 and j = n the minimum and maximum of the averages coincide and
Ln,j = 1.

Therefore we can conclude

Lemma 3. At each node at maturity the function vn,j (A) that provides the price of the option,
is a piecewise linear function on the interval [Amin max
n,j , An,j ]. Moreover, such a function is convex
on its domain.

8
 Now consider the step i, 0 ≤ i ≤ n − 1. At the node Ni,j we can evaluate recursively the
minimum and the maximum of the averages, respectively

(i + 2)Amin
i+1,j+1 − Si+1,j+1 (i + 2)Amax
i+1,j − Si+1,j
Amin
i,j = , Amax
i,j = .
i+1 i+1
Lemma 4. At each node Ni,j , i = 0, ..., n, j = 0, ..., i, the function vi,j (A), which provides the
price of the option as function of the average A, is piecewise linear and convex in the interval
[Amin max
i,j , Ai,j ].

Proof. The claim is true at step i = n (at maturity) by Lemma 3. At step i = n − 1, the price
function vi,j (A), with A ∈ [Amin max
i,j , Ai,j ], is obtained by considering the discounted expectation
value:
vi,j (A) = e−r∆T [πvi+1,j+1 (A′ ) + (1 − π)vi+1,j (A′′ )], (7)
where
(i + 1)A + s0 u2j−i+1 (i + 1)A + s0 u2j−i−1
A′ = , A′′ = . (8)
i+2 i+2
2j−i+1
As vn,j (A) is piecewise linear and convex on its domain and h1 (A) = vi+1,j+1 ( (i+1)A+s 0u
i+2
)
is a function composed by a linear function of A and a piecewise linear convex function, then
h1 (A) is piecewise linear and convex as a function of A. The same holds true for h2 (A) =
2j−i−1
vi+1,j ( (i+1)A+s
i+2
0u
). We can conclude that vi,j (A) is piecewise linear and convex on its
domain.
The claim of the Lemma now follows by backward induction. ♦

Figure 3: The price function vn−1,j (A) is obtained from vn,j (A) and vn,j+1 (A). It is piecewise
linear and convex and its internal singular points arise from the singular points of vn,j (A) and
vn,j+1 (A).

Consider again the step i = n − 1 and the node Ni,j . By Lemma 4, vi,j (A) is piecewise linear
and convex, hence it is characterized by its singular points (see Fig.3).
The valuation of the singular points can be carried out recursively by a backward algorithm,
which will be described in the sequel.
Each average Ali+1,j , l = 1, ..., Li+1,j , associated to a singular point of the node Ni+1,j is
projected in a new average value B l at the node Ni,j by the relation

l (i + 2)Ali+1,j − s0 u2j−i−1
B = . (9)
i+1
Note that B l is the average evaluated at the node Ni,j which becomes Ali+1,j after a down
movement of the underlying.
Observe that B l is increasing with respect to l, B Li+1,j = Amax
i,j for all j, and B 1 6∈
min max l min max
[Ai,j , Ai,j ] if 0 < j < i. Each B belonging to the interval [Ai,j , Ai,j ] becomes the first
coordinate of a singular point associated to the node Ni,j .

9
 In order to evaluate the price vi,j (B l ) associated to the singular average B l ∈ [Amin max
i,j , Ai,j ],
we remark that after a down movement of the underlying, B l transforms into Ali+1,j and the
l
corresponding price is Pi+1,j . Consider now an up movement of the underlying. In this case B l
2j−i+1
transforms into the average: Bup l
= (i+1)Bl +s
i+2
0u
. This average clearly could not belong to
the set of singular averages associated to the node Ni+1,j+1 . Therefore we need to evaluate the
index s such that Bup l
∈ [Asi+1,j+1 , As+1
i+1,j+1 ]. Since in this interval the price function is linear,
we have
s+1 s
l Pi+1,j+1 − Pi+1,j+1 l
vi+1,j+1 (Bup ) = s+1 s
(Bup − Asi+1,j+1 ) + Pi+1,j+1
s
.
Ai+1,j+1 − Ai+1,j+1
We can evaluate the price associated to the singular average B l evaluating the discounted
expectation value:

vi,j (B l ) = e−r∆T [πvi+1,j+1 (Bup

l
) + (1 − π)vi+1,j (Ali+1,j )]. (10)

In a similar way each singular average Ali+1,j+1 , l = 1, ..., Li+1,j+1 associated to the node
Ni+1,j+1 is projected in a new average C l at the node Ni,j by the relation

(i + 2)Ali+1,j+1 − s0 u2j−i+1
Cl = . (11)
i+1
Now C 1 = Amin i,j for all j, and C
Li+1,j+1
6∈ [Amin max l min max
i,j , Ai,j ] if 0 < j < i. For each C ∈ [Ai,j , Ai,j ]
l
we can evaluate the corresponding price vi,j (C ) similarly as before.
Finally we proceed by sorting the averages B l and C l belonging to [Amin max
i,j , Ai,j ], obtaining
L L
an ordered set {(A1i,j , Pi,j
1
), ..., (Ai,ji,j , Pi,ji,j )} of singular points at the node Ni,j . By the previous
construction these are exactly all the singular points associated to this node. Remark that
Li,j ≤ Li+1,j + Li+1,j+1 − 2.
The previous argument can be applied at every step i = n − 1, ..., 0 and it holds for all
j = 1, ..., i − 1. At the nodes Ni,i , Ni,0 , there is only a singular point whose price is given by
1 1 1 1 1 L
Pi,0 = e−r∆T [πPi+1,0 + (1 − π)Pi+1,1 ], Pi,i = e−r∆T [πPi+1,i+1 i+1,i
+ (1 − π)Pi+1,i ]; (12)

so that we get a complete description of the price function vi,j (A) at each node of the tree.
1
The value P0,0 is exactly the binomial price relative to the tree with n steps of fixed strike
European Asian call option. In fact, the method provides the price corresponding to every
possible average at each node, in particular to the averages which are effectively realized on the
binomial tree.

2.3 Fixed strike American Asian options

Consider now the American case. At maturity we have the same situation as in the European
case. The price function is vn,j (A) = (A − K)+ for A ∈ [Amin max
n,j , An,j ], and it is characterized by
the same singular points.
Consider the step i = n − 1. At the node Ni,j we first compute, by using the procedure
described in the previous section, the singular points associated to this node, obtaining in this
c
way the continuation value function vi,j (A).

10
 Taking into account of the American feature, the price function vi,j (A) is obtained by
comparing the continuation value with the early exercise:
c
vi,j (A) = max{vi,j (A), A − K}.

Let us remark that vi,j (A), A ∈ [Amin max

i,j , Ai,j ], is still a piecewise linear convex function. For
this reason we can characterize it again by its singular points.
In order to compute the singular points associated to the American case we first remark
c
that the slopes characterizing the piecewise linear convex function vi,j (A) are all smaller than 1.
c
This follows by virtue of equations (7), (8) and by differentiating vi,j (A) in the open intervals
l l+1
(Ai,j , Ai,j ), l = 1, ..., Li,j . Therefore there are two possible cases:

1. Amax c max c
i,j − K ≤ vi,j (Ai,j ) then vi,j ≡ vi,j , so the singular points do not change;

2. Amax c max
i,j − K > vi,j (Ai,j ). Here we have two subcases:

• Amin c min min max

i,j − K ≥ vi,j (Ai,j ) then vi,j (A) = A − K for all A ∈ [Ai,j , Ai,j ], so the set of
singular points consists only on two points

(Amin min max max

i,j , Ai,j − K), (Ai,j , Ai,j − K);

• Amin c min
i,j − K < vi,j (Ai,j ) then there is an unique average A where the continuation
value is equal to the early exercise. Let j0 be the largest index such that Aji,j0 < A.
The new set of singular points becomes (see also Fig.4):

{(A1i,j , Pi,j
1
), ..., (Aji,j0 , P (Aji,j0 )), (A, A − K), (Amax max
i,j , Ai,j − K)}.

The same argument can be applied at every step i = n − 2, ..., 0. This allows us to compute
1
P0,0 which provides the exact American binomial price relative to the tree with n steps.
Remark 3. The number of singular points associated to a node could decrease in the American
case, so the American procedure could be faster than the European one.
Remark 4. In the case of Asian put option the procedure is similar.
Remark 5. In the floating strike case the procedure is modified as follows: at maturity the
singular points depend not more on the strike K but on the underlying value at each node Si,j .
Therefore the new singular points are obtained by replacing K by Si,j . The backward procedure
is the same as before, just taking into account properly the new intrinsic values.

Figure 4: American case: the point A has been inserted, A4 and A5 have been removed.

11
2.4 Upper and lower bound
In the previous subsections we have introduced a new method in order to evaluate the exact
binomial price in a discrete setting of an European or American Asian option.
As Li,j ≤ Li+1,j + Li+1,j+1 − 2, the resulting algorithm can be of exponential complexity as
the standard binomial technique.
The main advantage of our technique is that it allows us to obtain easily both an upper and
a lower bound of the binomial price, drastically reducing the amount of computational time
and memory requirements. Moreover a further appeal is given by the possibility to obtain an
a-priori control of the distance of the estimates from the exact binomial price. Actually all
these results are simple consequences of the previous Lemma 1 and Lemma 2.
More precisely, in order to get an upper bound of the exact binomial price, we just remove
some singular points at each node. Lemma 1 ensures that the value obtained in such a way is
an upper estimate of the exact binomial price.
There are several possible criteria to remove the singular points. Here we propose the
following:
Consider the set of singular points C = {(A1i,j , Pi,j1
), ..., (ALi,j , Pi,j
L
)} (L = Li,j ), associated to
′
the node Ni,j and the corresponding price value function vi,j (A). Let vi,j (A) be the price value
l l
function obtained by removing a point (Ai,j , Pi,j ) from C. We have
′
|vi,j (A) − vi,j (A)| ≤ ǫl , ∀A ∈ [Amin max
i,j , Ai,j ] (13)

where
l+1 l−1
′ Pi,j − Pi,j
ǫl = vi,j (Ali,j ) − vi,j (Ali,j ) = l l−1 l−1 l
l−1 (Ai,j − Ai,j ) + Pi,j − Pi,j . (14)
Al+1
i,j − Ai,j

Therefore, given a real number h > 0 we choose to remove the point (Ali,j , Pi,j l
) if ǫl < h.
Repeating this procedure sequentially at each node of the tree, avoiding the elimination of two
consecutive singular points, we can conclude that the obtained upper estimate differs from the
exact binomial value at most for nh.
The algorithm for the computation of the lower bound is similar and follows by Lemma 2.
l−1 l−1
Removing the points (Ai,j , Pi,j ), (Ali,j , Pi,j
l
), l = 2, ..., L − 2, and adding the point (x, y) (see
Lemma 2) the difference between the values of the associated piecewise linear functions is less
or equal to δl , where
l l−1
Pi,j − Pi,j l−1 l−1
δl = l l−1 (x − Ai,j ) + Pi,j − y. (15)
Ai,j − Ai,j
This replacement will take place only if δl < h. Inductively and using the scheme proposed in
the next remark, we get that the obtained lower estimate differs again from the exact binomial
value at most for nh.
Remark 6. In the case of the lower estimate, in order to obtain an error smaller than h at
every step we propose the following algorithm:
we start considering the points (A1i,j , Pi,j
1
), (A2i,j , Pi,j
2
), (A3i,j , Pi,j
3
), (A4i,j , Pi,j
4
). If δ3 < h then
2 2 3 3
we add the point (x, y) and delete (Ai,j , Pi,j ), (Ai,j , Pi,j ). Moreover the procedure will continue
considering the new four points (x, y), (A4i,j , Pi,j
4
), (A5i,j , Pi,j
5
), (A6i,j , Pi,j
6
). On the other hand if

12
δ3 ≥ h then we don’t remove points and the procedure will continue considering the new four
points (A2i,j , Pi,j
2
), (A3i,j , Pi,j
3
), (A4i,j , Pi,j
5
), (A5i,j , Pi,j
5
). We repeat this procedure completing the
sequence of singular points of the node Ni,j .
Remark 7. Jiang and Dai [8] proved the convergence of the exact binomial algorithm for Euro-
pean/ American path-dependent options. In particular they proved that the rate of convergence
of the exact binomial algorithm to the continuous value is O(∆T ).
The possibility of obtaining estimates of the exact binomial price with an error control allows
us to prove easily the convergence of our method to the continuous value. Choosing h depending
on n and so that nh(n) → 0 we have that the corresponding sequences of upper and lower
estimates converge to the continuous price value. Moreover, choosing h(n) = O( n12 ), we are
able to guarantee that the order of convergence is O(∆T ).
Remark 8. The key issue in assessing the complexity of our algorithm is in the upper and
lower bound computation. A theoretical complexity analysis combined with the above upper and
lower bounds is out of reach. In fact, the control of the error with respect of the exact binomial
algorithm does not permit us to control of the number of singular points. Nevertheless, the
numerical results in section 4.2 indicate that the present method is very competitive in practice.

2.5 Sketch of the algorithm in the American Asian case

Let us finally summarize the algorithm in order to obtain an upper and a lower bound of the
exact binomial price for a fixed strike American Asian call option with an error smaller that
nh (h > 0).

• STEP n
- Compute the singular points at maturity by using (5) and (6).

• STEP i, for i = n − 1, ..., 0

1 1
- Evaluate Pi,0 , Pi,i by comparing the continuation values given in (12) with the early
exercise.
- For each node Ni,j , j = 1, ..., i−1, compute the set of the singular points by the following
steps:

1. for each average Ali+1,j , l = 1, ..., Li+1,j compute Bl by (9),

2. for Bl ∈ [Amin max c
i,j , Ai,j ] compute vi,j (Bl ) by (10),

3. for each average Ali+1,j+1 , l = 1, ..., Li+1,j compute Cl by (11),

4. for Cl ∈ [Amin max c
i,j , Ai,j ] compute vi,j (Cl ),

5. sort the set of the singular averages Bl and Cl ∈ [Amin max

i,j , Ai,j ] obtaining the set of
Li,j singular points associated to the node Ni,j ,
6. compute the American price according to Case 1 or 2 of Section 2.3 getting a new
set of singular points with a new cardinality denoted, for simplicity, again by Li,j ,
7. compute upper and lower bounds with error smaller than h:

13
 upper bound: remove sequentially all the singular points (Ali,j , Pi,j l
), l = 2, ..., Li,j −
1, for which ǫl < h (see (14)) avoiding the elimination of two consecutive singular
points, obtaining a new set with a new cardinality denoted again by Li,j ,
lower bound: for each l, l = 2, ..., Li,j − 2, for which δl < h (see (15)), remove the
l−1 l−1
points (Ai,j , Pi,j ), (Ali,j , Pi,j
l
) and add the point (x, y) given by the intersection
l−2 l−2 l−1 l−1
between the two straight lines joining (Ai,j , Pi,j ), (Ai,j , Pi,j ) and (Ali,j , Pi,j
l
),
l+1 l+1
(Ai,j , Pi,j ), respectively (following the scheme described in Remark 6). We
obtain again a new set of singular points with a new cardinality Li,j .
1
P0,0 is the upper [lower] estimate of the exact binomial price with error smaller that nh.

3 Lookback American options

We can apply the same procedure described in the Asian case, to the lookback options. Actually,
in this case the algorithm admits several simplifications.
Consider a fixed strike American lookback call option. At the nodes Nn,j , j = 0, ..., n
min
(at maturity) the maximum of the underlying varies between a minimum value Mn,j and a
max
maximum value Mn,j given by
min max
Mn,j = max{Sn,j , s0 }, Mn,j = s 0 uj .
min max
For all M ∈ [Mn,j , Mn,j ] the price of the option can be continuously defined by vn,j (M ) =
(M − K)+ .
As in the Asian case, the function vn,j (M ) is piecewise linear and its singular points are
valuable by relations (5), (6) where M replaces A.
Consider now the step i, 0 ≤ i ≤ n − 1. At the node Ni,j we can evaluate recursively the
minimum and the maximum value of the maximum of the underlying by the relations
min min max max
Mi,j = max{s0 , Mi+1,j+1 /u}, Mi,j = Mi+1,j . (16)
Lemma 5. At each node Ni,j , i = 0, ..., n, j = 0, ..., i, vi,j (M ) is a piecewise linear and convex
min max
function on the interval [Mi,j , Mi,j ].
Proof. The claim is true at step i = n. Consider the step i = n − 1. We extend the function
min max min
vi+1,j+1 to the interval [Mi+1,j+1 /u, Mi+1,j+1 ] setting vi+1,j+1 (M ) = vi+1,j+1 (Mi+1,j+1 ) for M ∈
min min c
[Mi+1,j+1 /u, Mi+1,j+1 ). With such an extension the continuation value price function vi,j (M ),
becomes
c
vi,j (M ) = e−r∆T [πvi+1,j+1 (M ) + (1 − π)vi+1,j (M )]. (17)
As vi+1,j+1 (M ) and vi+1,j (M ) are piecewise linear and convex we can conclude that the same
c c
holds true for vi,j (M ). Moreover vi,j (M ) = max{vi,j (M ), M − K}, therefore vi,j (M ) is still
piecewise linear and convex. Inductively we have the claim. ♦
By the previous lemma, the price of an American lookback option can be obtained by
computing only the singular points of the price function at each node. For this purpose we could
use an algorithm similar to the one described in the Asian case. So the procedure for American
1 1 L L
lookback options consists in evaluating first the singular points (Mi,j , Pi,j ), ..., (Mi,j , Pi,j ) of
c
vi,j (M ). Then we can get the singular points of vi,j (M ) in an easy way:

14
 max c max c
• if Mi,j −K ≤ vi,j (Mi,j ) then the sets of singular points of vi,j (M ) and vi,j (M ) coincide;
min c min
• if Mi,j − K ≥ vi,j (Mi,j ) then the set of singular points is composed only by two points:
min min max max
(Mi,j , Mi,j − K), (Mi,j , Mi,j − K);
min c min max c max
• if Mi,j − K < vi,j (Mi,j ) and Mi,j − K > vi,j (Mi,j ) then there is an unique critical
min max
value M i,j ∈ (Mi,j , Mi,j ) where the continuation value coincides with the early exercise
value. Then the set of singular points of vi,j is composed by all the singular points
c min
of vi,j whose singular value belongs to [Mi,j , M i,j ), with the addition of the points:
max max
(M i,j , M i,j − K), (Mi,j , Mi,j − K).

It is important to note that the particular structure of the tree in the lookback case allows us
to obtain a simpler and more efficient procedure for the valuation of the singular points of vi,j .
This procedure, described in the next Proposition 1, is based on the possibility of computing
the singular points in a direct way avoiding the sorting procedure. For this purpose we first
need some properties which are strictly related to the lookback case:
min max
Lemma 6. The price value function vi,j (M ), M ∈ [Mi,j , Mi,j ] has the following properties:
min max min
a) if K ∈ (Mi,j , Mi,j ) then vi,j (M ) is constant in [Mi,j , K],
min max
b) if M ∈ [Mi,j , Mi,j−1 ] and vi,j (M ) = M − K then vi,j−1 (M ) = M − K,
min max
c) if M ∈ [Mi+1,j+1 , Mi,j ] and vi+1,j+1 (M ) = M − K then vi,j (M ) = M − K,
d) assume that x1 = s0 ul , x2 ∈ (s0 ul , s0 ul+1 ), x3 = s0 ul+1 are singular values of vi,j . If we
delete the singular point (x2 , vi,j (x2 )) then v0,0 (s0 ) does not change.
Proof. The first two properties follow easily by induction on the tree. Property (c) follows by
(b).
The claim of (d) follows by the fact that the value of the option at the nodes Ni,0 , Ni,i ,
i = 0, ..., n − 1, depends only on the values that vi+1,j assumes at the nodal stock values of the
tree. ♦
By Lemma 6(d) it follows that every singular value which lies between two consecutive nodal
stock values and which are singular values as well, can be removed. This implies that we may
delete the critical value M i,j , during the backward procedure, if it lies between two consecutive
nodal singular values.
In the next proposition we shall see that the set of internal singular values of vi,j at each
node can be reduced to a sequence of consecutive nodal singular values which are singular values
of vi+1,j+1 as well, with the eventual addition of K. M i,j lies always between two consecutive
nodal singular values, so that it is not necessary to compute it in the backward procedure.
Proposition 1. Consider the price value function vi,j and denote by l0 the smallest index l
such that s0 ul > max{K, Mi,j min
}. The set of singular values of vi,j can be reduced to:
min max min max
Mi,j , Mi,j , K if K ∈ (Mi,j , Mi,j ) and a set (eventually empty) of consecutive nodal stock
l0 l0 +1 l0 +k
values {s0 u , s0 u , ..., s0 u } which are singular values of vi+1,j+1 as well.
M max
Moreover if M = s0 ul0 +k < i,j u
, then vi,j (M ) = M − K.
Proof. See Appendix.

15
Remark 9. As in the case of Asian options, our procedure allows us to obtain an upper and a
lower bound of the price in a simple way. However in this case the singular points are very few
and their distance is much more relevant than in the Asian case.
For this reason is not useful to compute upper and lower bounds unless we need to consider
an extremely large number of time steps.

3.1 Sketch of the algorithm in the American lookback case

Let us summarize the algorithm in order to obtain the exact binomial price for a fixed strike
American lookback call option.

• STEP n
- Compute the singular points at maturity by using (5) and (6) where M replaces A.

• STEP i, for i = n − 1, ..., 0

1 1
- compute Pi,0 , Pi,i by comparing the continuation values given in (12) with the early
exercise,
- for each node Ni,j , j = 1, ..., i−1, compute the set of the singular points by the following
steps:
c min c max
1. evaluate vi,j (Mi,j ), vi,j (Mi,j ),
c min mim min min
if vi,j (Mi,j ) ≤ Mi,j −K then there are only two singular points: (Mi,j , Mi,j −K),
max max min min
(Mi,j , Mi,j −K) and the computation is concluded; otherwise insert (Mi,j , vi,j (Mi,j )),
max max
(Mi,j , vi,j (Mi,j )),
min max
2. if K ∈ (Mi,j , Mi,j ) then insert (K, vi,j (K)),
max c
3. for each singular value M of the node Ni+1,j+1 belonging to (K, Mi,j ) add (M, vi,j (M )).
c max max c
If vi,j (Mi,j ) ≥ Mi,j −K then vi,j and vi,j coincide so the computation is concluded.
Otherwise (in this case M i,j exists) remove all the singular points with singular value
min max
internal to [Mi,j , Mi,j ] and singular price given by the early exercise, except from
the one which has the smallest singular value.

4 Numerical Comparisons
In this section we will illustrate numerically the efficiency of the singular points method, pre-
viously introduced, for pricing fixed Asian and lookback options in the American case.
We will first compare our algorithm with the most efficient tree methods for the fixed strike
American Asian call options. Then we will study the behavior of convergence to the continuous
price. Our comparison will include the PDE-based methods. Finally we will consider lookback
options.
All the computations were performed in double precision on a PC equipped with a processor
Centrino at 1.6 Ghz and 512 Mb of RAM.

16
4.1 Fixed strike American Asian call options: comparison with the
tree methods
In order to check the behavior of the singular points algorithm, we will compare:
1. the exact CRR binomial method;
2. the Hull-White method (HW) with h = 0.005 (see [7]);
3. the linear interpolation forward shooting grid method (FSG) of Barraquand-Pudet choos-
ing ρ = 0.1 (see [2],[11]);
4. the Chalasani et al. method (CJEV) providing an upper and a lower bound (see [3]);
5. the singular points method providing an upper and a lower bound with a level of error
smaller than nh, with two different choices of h: h = 10−4 (SP1 ), h = 10−5 (SP2 ).
We will assume that the initial value of the stock price is s0 = 100, the maturity T = 1,
the force of interest rate r = 0.1, the continuous dividend yield q = 0.03. We will consider two
choices for volatility σ = 0.2, σ = 0.4 and two choices for the strike K = 90 and K = 110.
We will consider various time steps n = 25, 50, 100, 200, 400, 800 and we will report the
price estimates and the corresponding time of computation (in brackets). The exact binomial
method is available only for n = 25, while CJEV is available only for n = 25, 50, 100 because
of insufficient memory capacity. In the case of CJEV the global computational time in order
to obtain the upper and lower estimates (by means of a single procedure) has been reported.
As regards the SP methods, the two estimates are obtained separately.

n HW FSG CJEV SP1 SP2 Exact BIN

down up down up down up
σ = 0.2 25 14.60279 14.25496 14.24607 14.24723 14.24610 14.24628 14.24615 14.24617 14.24616
(0.028) (0.032) (0.012) (0.008) (0.008) (0.011) (0.009)
50 14.49228 14.48210 14.47778 14.47887 14.47767 14.47830 14.47788 14.47793 -
(0.15) (0.20) (0.20) (0.03) (0.03) (0.07) (0.05)
100 14.64524 14.62562 14.62141 14.62220 14.62095 14.62258 14.62150 14.62165 -
(0.78) (1.64) (3.02) (0.11) (0.09) (0.28) (0.20)
200 14.74383 14.71127 - - 14.70601 14.70954 14.70727 14.70762 -
(4.70) (13.20) (0.48) (0.34) (1.33) (0.95)
400 14.80389 14.76035 - - 14.75368 14.76072 14.75641 14.75715 -
(37.94) (105.82) (2.08) (1.45) (6.19) (4.31)
800 14.83690 14.78716 - - 14.77751 14.79049 14.78318 14.78464 -
(201.86) (836,80) (9.94) (6.73) (30.19) (20.66)
σ = 0.4 25 18.32871 18.32871 17.84535 17.85148 17.84666 17.84687 17.84672 17.84674 17.84672
(0.046) (0.031) (0.013) (0.009) (0.008) (0.013) (0.011)
50 18.21421 18.21659 18.20337 18.20847 18.20428 18.20493 18.20448 18.20454 -
(0.28) (0.20) (0.20) (0.04) (0.03) (0.09) (0.06)
100 18.46687 18.46077 18.44834 18.45253 18.44918 18.45091 18.44975 18.44992 -
(1.78) (1.64) (3.01) (0.16) (0.12) (0.41) (0.30)
200 18.62676 18.60731 - - 18.59546 18.59922 18.59676 18.59714 -
(9.51) (13.23) (0.80) (0.55) (2.28) (1.56)
400 18.72696 18.69170 - - 18.67888 18.68645 18.68168 18.68247 -
(50.70) (104.60) (4.33) (3.00) (13.28) (9.09)
800 18.78597 18.73771 - - 18.72242 18.73669 18.72822 18.72979 -
(302.72) (825,66) (31.34) (21.11) (96.95) (64.91)

Table 1: Fixed strike American Asian call options with T = 1, s0 = 100, r = 0.1, q = 0.03 and
K = 90

17
 n HW FSG CJEV SP1 SP2 Exact BIN
down up down up down up
σ = 0.2 25 2.21580 2.21376 2.20952 2.21154 2.20973 2.21000 2.20982 2.20984 2.20983
(0.031) (0.031) (0.013) (0.008) (0.006) (0.011) (0.009)
50 2.25529 2.24769 2.24348 2.24475 2.24353 2.24451 2.24384 2.24393 -
(0.14) (0.20) (0.20) (0.03) (0.02) (0.06) (0.05)
100 2.28191 2.26623 2.26213 2.26290 2.26169 2.26416 2.26245 2.26269 -
(0.78) (1.64) (3.03) (0.11) (0.08) (0.25) (0.17)
200 2.29734 2.27597 - - 2.27054 2.27562 2.27220 2.27275 -
(4.70) (13.07) (0.44) (0.31) (1.22) (0.84)
400 2.30536 2.28080 - - 2.27359 2.28331 2.27705 2.27815 -
(37.92) (104.59) (1.95) (1.34) (5.73) (3.98)
800 2.30944 2.28292 - - 2.27221 2.28977 2.27919 2.28120 -
(201.59) (828,03) (9.50) (6.47) (28.95) (19.59)
σ = 0.4 25 6,78517 6.79136 6.77940 6.78672 6.78106 6.78131 6.78115 6.78117 6.78116
(0.047) (0.031) (0.013) (0.009) (0.008) (0.011) (0.009)
50 6.88872 6.89089 6.87917 6.88433 6.88086 6.88184 6.88118 6.88127 -
(0.28) (0.20) (0.20) (0.04) (0.03) (0.08) (0.06)
100 6.95445 6.94958 6.93817 6.94173 6.93943 6.94190 6.94024 6.94048 -
(1.78) (1.63) (3.03) (0.16) (0.11) (0.39) (0.26)
200 6.99555 6.98150 - - 6.97075 6.97591 6.97249 6.97302 -
(9.47) (13.12) (0.78) (0.53) (2.22) (1.52)
400 7.01909 6.99776 - - 6.98572 6.99571 6.98935 6.99047 -
(50.53) (104.75) (4.27) (2.95) (13.23) (8.89)
800 7.03155 7.00538 - - 6.99025 7.00849 6.99772 6.99980 -
(301.80) (827,48) (31.23) (20.98) (95.44) (64.20)

Table 2: Fixed strike American Asian call options with T = 1, s0 = 100, r = 0.1, q = 0.03 and
K = 110

σ = 0.2, K = 90 σ = 0.4, K = 90 σ = 0.2, K = 110 σ = 0.4, K = 110

n CJEV SP1 SP2 CJEV SP1 SP2 CJEV SP1 SP2 CJEV SP1 SP2
25 .00016 .00018 .00002 .00613 .00021 .00002 .00202 .00027 .00002 .00732 .00025 .00002
50 .00109 .00063 .00005 .00510 .00065 .00006 .00127 .00098 .00009 .00516 .00098 .00009
100 .00079 .00163 .00015 .00419 .00173 .00017 .00077 .00247 .00024 .00356 .00247 .00024
200 - .00353 .00035 - .00376 .00038 - .00508 .00055 - .00516 .00053
400 - .00704 .00074 - .00757 .00079 - .00972 .00110 - .01001 .00112
800 - .01298 .00146 - .01427 .00157 - .01756 .00201 - .01824 .00208

Table 3: Difference between the upper and lower estimates for CJEV, SP methods

The numerical results obtained in Table 1 and 2 confirm the reliability of the singular points
method. In comparison with the Chalasani et al. method (see also Table 3) we obtained an
actual improvement in precision for the the upper and lower bounds in a lower CPU times
and without problems of memory. With respect to Hull-White and the forward shooting grid
methods the improvements seem to be significant.

4.2 Analysis of convergence of the approximations to the continuous

value
In this section we will address more thoroughly the complexity and the convergence to the
continuous price value of our algorithm both in the European and the American cases. We will
compare our algorithm with the modified linear interpolation forward shooting grid method
(M-FSG), which guarantees the convergence to the continuous price value (see [11]), and with
two PDE-based methods (see [6] and [10]). We used the Richardson extrapolation in order

18
to speed-up the convergence of the tree methods. In the European case we used the two-
points extrapolation 2Pn − P n2 , whereas in the American case the three points extrapolation
8
P − 2P n2 + 13 P n4 was adopted.
3 n
As regards to the convergence analysis, we will compare the following algorithms:
1. the PDE-based method of d’Halluin et al. (DFL) available for both the European and
the American Asian options(see [6]);
2. the PDE-based method of Vecer available in the European Asian option case (see [10]);
3. the modified linear interpolation forward
√ shooting grid method (M-FSG) of Barraquand-
Pudet (see [2],[11]). We chose ρ = 0.1 and n n grid points in the Asian direction in order to
guarantee the convergence (see the Premia implementation [9]);
4. the modified FSG algorithm with the Richardson extrapolation (M-FSG-Rich);
5. the singular points method (SP) providing an upper bound with a level of error smaller
than nh with h = 0.1 n2
(see Remark 7);
6. the previous singular points upper algorithm combined with the Richardson extrapolation
(SP-Rich).
For the PDE-based method we will use the numerical results provided in [6]. In order to
compare the convergence behavior we consider the convergence ratio R proposed in [6],

P n2 − P n4
R=
Pn − P n2

In Tables 4 and 5 the European Asian call case is considered using low and high volatility. Table
6 refers to the American Asian put case. The PDE-based algorithms of Vecer and d’Halluin et al.
are almost second order in time (see [6]). The singular points and the modified FSG algorithms
exhibit, as expected, first-order convergence. The use of the Richardson extrapolation speeds
up the convergence both in the case of our method and the modified FSG method.
As concern the computational analysis we have to take into account the computational
7
time (see Remark 8). We will compare our algorithm (SP) with: M-SFG of complexity O(n 2 ),
FSG of complexity O(n3 ) and the Vecer method of complexity O(n2 ). Fig.5 offers a number
of steps/time of computation graph using data of Table 4. The comparison indicates that the
present method can effectively be competitive in practice and it seems to be of complexity
O(n3 ). More estensive numerical experiments have confirmed this order of complexity.

n DFL Vecer M-FSG M-FSG-Rich SP SP-Rich

Price R Price R Price R Price R Price R Price R
25 1.857193 n.a. 1.839863 n.a. 1.845628 n.a. n.a. n.a. 1.845841 n.a. n.a. n.a.
50 1.853254 n.a. 1.848642 n.a. 1.848535 n.a. 1.851442 n.a. 1.848745 n.a. 1.851655 n.a.
100 1.852120 3.475 1.839863 3.974 1.850044 1.927 1.851553 n.a. 1.850204 1.996 1.851661 n.a.
200 1.851781 3.338 1.851407 3.979 1.850814 1.960 1.851583 3.616 1.850902 2.087 1.851600 -0.102
400 1.851660 2.815 1.851546 3.987 1.851202 1.983 1.851590 4.603 1.851248 2.016 1.851594 10.992

Table 4: Fixed strike European Asian call options with T = 0.25, s0 = 100, K = 100, r = 0.1, q = 0,
σ = 0.1

19
 n DFL Vecer M-FSG M-FSG-Rich SP SP-Rich
Price R Price R Price R Price R Price R Price R
25 6.010203 n.a. 6.009821 n.a. 5.995543 n.a. n.a. n.a. 5.995682 n.a. n.a. n.a.
50 6.015092 n.a. 6.014848 n.a. 6.005734 n.a. 6.015924 n.a. 6.005903 n.a. 6.016124 n.a.
100 6.016344 3.905 6.016251 3.582 6.011129 1.889 6.016525 n.a. 6.011255 1.910 6.016607 n.a.
200 6.016651 4.085 6.016619 3.816 6.013911 1.939 6.016693 3.559 6.013982 1.962 6.016710 4.673
400 6.016723 4.219 6.016713 3.915 6.015321 1.974 6.016730 4.546 6.015361 1.979 6.016740 3.485

Table 5: Fixed strike European Asian call options with T = 0.25, s0 = 100, K = 100, r = 0.05, q = 0,
σ = 0.5

n DFL M-FSG M-FSG-Rich SP SP-Rich

Price R Price R Price R Price R Price R
25 2.220443 n.a. 2.047501 n.a. n.a. n.a. 2.079846 n.a. n.a. n.a.
50 2.195726 n.a. 2.091589 n.a. n.a n.a. 2.124386 n.a. n.a. n.a.
100 2.188555 3.447 2.118518 1.637 2.148704 n.a. 2.151597 1.637 2.182102 n.a.
200 2.186717 3.903 2.134301 1.706 2.151631 1.948 2.167529 1.708 2.185012 2.054
400 2.186243 3.847 2.143060 1.802 2.152397 3.818 2.176364 1.803 2.185777 3.802

Table 6: Fixed strike American Asian put options with T = 0.25, s0 = 100, K = 100, r = 0.05, q = 0,
σ = 0.15

Figure 5: Number of steps / time of computation table and graphic in log-log scale

4.3 American fixed strike lookback call options

In the lookback case we will simply compare our technique with an optimized version of the
exact binomial method. As already observed there are very few singular points involved in the
computation, so that the valuation of an upper and a lower bound is not significant. Therefore,
the price obtained through the use of the singular points method coincides with the exact
binomial, but with an improvement in the computational time (see Table 7 and 8 where the
parameters of Section 4.1 are used).
Clearly, when compared with the previous literature, improvements are less significant in
the lookback case than in the Asian case. Nevertheless, the data confirm the power of the
method and its relevance in pricing American path-dependent options.

σ n Bin SP σ n Bin SP
0.2 100 27.73002 27.73002 0.4 100 44.31762 44.31762
(0.004) (0.003) (0.004) (0.003)
200 28.02747 28.02747 200 45.00766 45.00766
(0.025) (0.016) (0.026) (0.017)
400 28.24333 28.24333 400 45.50900 45.50900
(0.184) (0.077) (0.183) (0.080)
800 28.39866 28.39866 800 45.87045 45.87045
(1.28) (0.30) (1.47) (0.42)
1600 28.51033 28.51033 1600 46.12961 46.12961
(10.75) (1.55) (12.02) (2.11)

Table 7: Fixed strike American lookback call options with K = 90 for binomial method and SP method

20
 σ n Bin SP σ n Bin SP
0.2 100 11.06517 11.06517 0.4 100 27.28512 27.28512
(0.004) (0.003) (0.004) (0.003)
200 11.27996 11.27996 200 27.85271 14.3245
(0.025) (0.016) (0.024) (0.017)
400 11.43759 11.43759 400 28.26777 28.26777
(0.184) (0.077) (0.183) (0.081)
800 11.55096 11.55096 800 28.57206 28.57206
(1.45) (0.38) (1.46) (0.43)
1600 11.63192 11.63192 1600 28.79142 28.79142
(12.02) (2.00) (11.98) (2.16)

Table 8: Fixed strike American lookback call options with K = 110 for binomial method and SP
method

5 Conclusion
We have introduced a new general binomial framework, called ’singular points method’, for
pricing path-dependent options of European/American type. We have applied it in the case
of Asian and lookback options. The procedure provides upper and lower bounds of the exact
binomial price with a prescribed level of error. The control of the error allows us to immediately
prove the convergence of order O(∆T ) to the continuous value. The method is competitive in
practice and the observed computational complexity is O(n3 ). The numerical results showed
that the singular points method is an improvement on the previous tree methods.

References
[1] Babbs S.: Binomial Valuation of Lookback Options. [Link]. Dynam. Control 24, 1499-
1525 (2000).

[2] Barraquand J., Pudet T.: Pricing of American Path-dependent Contingent Claims. Math-
ematical Finance 6, 17-51 (1996). 2

[3] Chalasani P., Jha S., Egriboyun F., Varikooty A. : A Refined Binomial Lattice for Pricing
American Asian Optons. Review of Derivatives Research 3, 85-105 (1999). 2, 17, 19

[4] Chalasani P., Jha S., Varikooty A. : Accurate Approximations for European Asian Options.
Journal of Computational Finance 1, 11 - 29 (1999). 2, 17
2

[5] Cox J., Ross S.A. and Rubinstein M. : Option Pricing:A simplified appoach Journal of
Financial Economics 7, 229-264 (1979).

[6] V.D’Halluin, [Link], [Link] : A semi-Lagrangian Approach for American Asian

options under jump-diffusionSiam [Link]. 27, 315-345 (2005).

[7] Hull J., White A. : Efficient Procedures for Valuing European and American Path-
dependent Options Journal of derivatives 1, 21-31 (1993).

21
 [8] L. Jiang, [Link] : Convergence of binomial tree methods for European/American Path-
dependent Options SIM Journal on numerical analysis 42-3, 1094-1109 (2005).
[9] PREMIA : An Option Pricer, Mathfi Project (INRIA, CERMICS, UMLV)
[Link]
[10] [Link] : A new PDE approach for pricing arithmetic average Asian option. Journal of
Computational Finance 4 Summer, 103-113 (2001).
[11] Forsyth P.A. Vetzal K.R. and Zvan R. : Convergence of numerical methods for valuing
path-dependent options using interpolation. Review of Derivatives Research 5, 273-314
(2002).

2, 18, 19
2, 17
13

6 Appendix: Proof of Proposition 1

Proof. Consider the case i = n − 1. Take first j ≥ int[ i+1 2
] and j < n − 1 (the case j = n − 1 is
trivial). 19
c min max min max
At the node Ni,j the singular values of vi,j are Mi,j , Mi,j , K if K ∈ (Mi,j , Mi,j ) and
min min min c
eventually uMi,j = Mi+1,j+1 . By Lemma 6(a) uMi,j is a singular value of vi,j if and only if
min
uMi,j ≥ K. 2, 18, 19
c
Take now the value function vi,j . The possible singular points of vi,j are the same of vi,j with
the possible addiction of M i,j (when it exists). If M i,j exists then M i,j ≥ K. If M i,j = K then
min min min
the claim follows, otherwise necessarily uMi,j = Mi+1,j+1 is a singular value and K < uMi,j .
min min min min
Hence vi+1,j+1 (uMi,j ) = uMi,j − K and, by Lemma 6(c), vi,j (uMi,j ) = uMi,j − K. We can
min min
conclude that M i,j ∈ [Mi,j , uMi,j ] and by Lemma 6(d) it can be removed. Hence the claim
holds. 2, 17, 18, 19
In the case j < int[ i+12
] there are no singular values in (K, Mi,j max
) so the claim is trivial.
Consider now the case i < n − 1 and take 0 < j < n − i (the cases j = 0, j = n − i are
trivial).
c
All the singular values of vi,j are either singular values of vi+1,j+1 or singular values of vi+1,j .
min max
We claim that every singular value of vi+1,j belonging to (Mi,j , Mi,j ) is a singular values of
min
vi+1,j+1 as well. In fact if M > min{K, Mi,j } is a singular value of vi+1,j then, by induction,
c
it is a singular value of vi+2,j+1 as well, therefore it is a singular value of vi+1,j+1 . By Lemma
6(b) we can conclude that it is a singular value of vi+1,j+1 as well. Therefore the set of all the
c min max
singular values of vi,j is composed by Mi,j , Mi,j , eventually K and a sequence of consecutive
l0 l0 +1 l0 +k
nodal values {s0 u , s0 u , ..., s0 u } which are singular values of vi+1,j+1 as well.
Take now vi,j . If v i,j (Mi,j max
) ≥ Mi,j max
− K then vi,j ≡ vi,j c
and their singular points co-
M max
incide. If s0 ul0 +k < i,j u
then s0 ul0 +k+1 is not a singular value of vi+1,j+1 . By induction
vi+1,j+1 (s0 ul0 +k ) = s0 ul0 +k − K. By Lemma 6(c) vi,j (s0 ul0 +k ) = s0 ul0 +k − K.

22
 Assume v i,j (Mi,jmax max
) < Mi,j − K. If vi,j (Mi,j min min
) ≤ Mi,j − K then there are no singular
min max min min
points in (Mi,j , Mi,j ) and the claim holds. If vi,j (Mi,j ) > Mi,j − K then M i,j exists
l c
and M i,j ≥ K. Let l1 be the largest index l such that s0 u is a singular point of vi,j and
l max l1 max c
s0 u ∈ (K, Mi,j ). If s0 u = Mi,j /u then the sequence of singular values of vi,j includes
all the nodal stock values from s0 ul0 to Mi,j max
. Denoting by l the smallest index such that
M i,j ≤ s0 u , we have that the singular values s0 ul+1 , ..., s0 ul1 can be removed, hence the claim
l

holds. If s0 ul1 < Mi,jmax

/u by the induction hypothesis vi+1,j+1 (s0 ul1 ) = s0 ul1 − K. By Lemma
6(c) M i,j ≤ s0 ul1 . Again s0 ul+1 , ..., s0 ul1 can be removed and vi,j (s0 ul ) = s0 ul − K, proving the
claim. ♦

23