Trading Transforms of Non-weighted Simple Games

arXiv:2101.07621v2 [cs.GT] 29 May 2021

and Integer Weights of Weighted Simple Games ∗

† ‡
Akihiro Kawana Tomomi Matsui

June 1, 2021

Abstract
This study investigates simple games. A fundamental research
question in this field is to determine necessary and sufficient conditions
for a simple game to be a weighted majority game. Taylor and Zwicker
(1992) showed that a simple game is non-weighted if and only if there
exists a trading transform of finite size. They also provided an upper
bound on the size of such a trading transform, if it exists. Gvozdeva
and Slinko (2011) improved that upper bound; their proof employed a
property of linear inequalities demonstrated by Muroga (1971). In this
study, we provide a new proof of the existence of a trading transform
when a given simple game is non-weighted. Our proof employs Farkas’
lemma (1894), and yields an improved upper bound on the size of a
trading transform.
We also discuss an integer-weight representation of a weighted sim-
ple game, improving the bounds obtained by Muroga (1971). We show
that our bound on the quota is tight when the number of players is
less than or equal to five, based on the computational results obtained
by Kurz (2012).
Furthermore, we discuss the problem of finding an integer-weight
representation under the assumption that we have minimal winning
coalitions and maximal losing coalitions. In particular, we show a
performance of a rounding method.
Lastly, we address roughly weighted simple games. Gvozdeva and
Slinko (2011) showed that a given simple game is not roughly weighted
if and only if there exists a potent certificate of non-weightedness.
∗
preliminary version of this paper was presented at Seventh International Workshop
on Computational Social Choice (COMSOC-2018), Rensselaer Polytechnic Institute, Troy,
NY, USA, 25-27 June, 2018.
†
Graduate School of Engineering, Tokyo Institute of Technology
‡
Graduate School of Engineering, Tokyo Institute of Technology

1
 We give an upper bound on the length of a potent certificate of non-
weightedness. We also discuss an integer-weight representation of a
roughly weighted simple game.

1 Introduction
A simple game consists of a pair G = (N, W), where N is a finite set of
players, and W ⊆ 2N is an arbitrary collection of subsets of N . Throughout
this paper, we denote |N | by n. Usually, the property

(monotonicity): if S ′ ⊇ S ∈ W, then S ′ ∈ W, (1)

is assumed. Subsets in W are called winning coalitions. We denote 2N \ W

by L, and subsets in L are called losing coalitions. A simple game (N, W)
is said to be weighted if there exists a weight vector w ∈ RN and q ∈ R
satisfying the following property:
X
(weightedness): for any S ⊆ N, S ∈ W if and only if wi ≥ q. (2)
i∈S

Previous research established the necessary and sufficient conditions that

guarantee the weightedness of a simple. [Elgot, 1961] and [Chow, 1961] in-
vestigated the theory of threshold logic and showed the condition of the
weightedness in terms of asummability. [Muroga, 1971] proved the suffi-
ciency of asummability based on the theory of linear inequality systems
and discussed some variations of their results in cases of a few variables.
[Taylor and Zwicker, 1992, Taylor and Zwicker, 1999] obtained necessary and
sufficient conditions independently in terms of a trading transform. A trad-
ing transform of size j is a coalition sequence (X1 , X2 , . . . , Xj ; Y1 , Y2 , . . . , Yj ),
which may contain repetitions of coalitions, satisfying the condition ∀p ∈ N ,
|{i | p ∈ Xi }| = |{i | p ∈ Yi }|. A simple game is called k-trade robust if there
is no trading transform of size j satisfying 1 ≤ j ≤ k, X1 , X2 , . . . , Xj ∈ W,
and Y1 , Y2 , . . . , Yj ∈ L. A simple game is called trade robust if it is k-trade
robust for all positive integers k.
Taylor and Zwicker showed that a given simple game G with n players is
n
weighted if and only if G is 22 -trade robust. In 2011, [Gvozdeva and Slinko, 2011]
showed that a given simple game G is weighted if and only if G is (n+1)nn/2 -
trade robust. [Freixas and Molinero, 2009b] proposed a variant of trade ro-
bustness, called invariant-trade robustness, which determines whether a sim-
ple game is weighted. The relations between the results in threshold logic
and simple games are clarified in [Freixas et al., 2016, Freixas et al., 2017].

2
 In Section 2, we show that a given simple game G is weighted if and
only if G is αn+1 -trade robust, where αn+1 denotes the maximal value of
determinants of (n + 1) × (n + 1) 0–1 matrices. It is well-known that αn+1 ≤
n+2
(n + 2) 2 (1/2)(n+1) .
Our definition of a weighted simple game allows for an arbitrary real
number of weights. However, any weighted simple game can be represented
by integer weights (e.g., see [Freixas and Molinero, 2009a]). An integer-
weight representation of a weighted simple game consists of an integer vec-
tor w ∈ ZN and some q ∈ Z satisfying the weightedness property (2).
[Isbell, 1956] found an example of a weighted simple game with 12 players
without a unique minimum-sum integer-weight representation. Examples for
9, 10, or 11 players are given in [Freixas and Molinero, 2009a, Freixas and Molinero, 2010].
In the field of threshold logic, examples of threshold functions requiring large
weights are discussed by [Myhill and Kautz, 1961, Muroga, 1971, Håstad, 1994].
Some previous studies enumerate (minimal) integer-weight representations
of simple games with a small number of players (e.g., [Muroga et al., 1962,
Winder, 1965, Muroga et al., 1970, Krohn and Sudhölter, 1995]). In the case
of n = 9 players, refer to [Kurz, 2012]. In general, [Muroga, 1971] (Proof of
Theorem 9.3.2.1) showed that (under the monotonicity property (1) and
∅ 6∈ W ∋ N ) every weighted simple game has an integer-weight repre-
n+1
sentation satisfying 0 ≤ wi ≤ αn ≤ (n + 1) 2 (1/2)n (∀i ∈ N ) and
n+1
0 ≤ q ≤ nαn ≤ n(n+1) 2 (1/2)n simultaneously. Here, αn denotes the max-
imal value of determinants of n×n 0–1 matrices. [Wang and Williams, 1991]
discussed Boolean functions that require more general surfaces to sepa-
rate their true vectors from false vectors. [Hansen and Podolskii, 2015] in-
vestigates the complexity of computing Boolean functions by polynomial
threshold functions. [Freixas, 2021] discusses a point-set-additive pseudo-
weighting for a simple game, which assigns weights directly to coalitions.
In Section 3, we slightly improve Muroga’s result and show that ev-
ery weighted simple game (satisfying ∅ 6∈ W ∋ N ) has an integer-weight
representation
P (q; w ⊤ ) satisfying |wi | ≤ αn (∀i ∈ N ), |q| ≤ αn+1 , and
1 ≤ i∈N wi ≤ 2αn+1 − 1 simultaneously. Based on the computational
results of [Kurz, 2012], we also demonstrate the tightness of our bound on
the quota when n ≤ 5.
For a family of minimal winning coalitions, [Peled and Simeone, 1985]
proposed a polynomial-time algorithm for checking the weightedness of a
given simple game. They also showed that for weighted simple games repre-
sented by minimal winning coalitions, all maximal losing coalitions can be
computed in polynomial time. When we have minimal winning coalitions

3
and maximal losing coalitions, there exists a linear inequality system whose
solution gives a weight vector w ∈ RN and q ∈ R satisfying property (2).
However, it is less straightforward to find an integer-weight representation as
the problem transforms from linear programming to integer programming.
In Section 4, we address the problem of finding an integer-weight rep-
resentation under the assumption that we have minimal winning coalitions
and maximal losing coalitions. We show that an integer-weight represen-
tation is obtained by carefully rounding √ a solution
√ of the linear inequality
system multiplied by at most (2 − 2)n + ( 2 − 1).
A simple game G = (N, W) is called roughly weighted if there exist a
non-negative vector w ∈ RN + and a real number q ∈ R, P not all equal to
zero ((q; w ⊤ ) 6= 0⊤P ), such that for any S ⊆ N condition i∈S wi < q im-
plies S 6∈ W, and i∈S wi > q implies S ∈ W. We say that (q; w ⊤ ) is a
rough voting representation for G. Roughly weighted simple games were ini-
tially introduced by [Baugh, 1970]. [Muroga, 1971] (p. 208) studied them
under the name of pseudothreshold functions. [Taylor and Zwicker, 1999]
discussed roughly weighted simple games and constructed several examples.
[Gvozdeva and Slinko, 2011] developed a theory of roughly weighted simple
games. A trading transform (X1 , X2 , . . . , Xj ; Y1 , Y2 , . . . , Yj ) with all coali-
tions X1 , X2 , . . . , Xj winning and Y1 , Y2 , . . . , Yj losing is called a certificate of
non-weightedness. This certificate is said to be potent if the grand coalition
N is among X1 , X2 , . . . , Xj and the empty coalition is among Y1 , Y2 , . . . , Yj .
[Gvozdeva and Slinko, 2011] showed that under the the monotonicity prop-
erty (1) and ∅ 6∈ W ∋ N , a given simple game G is not roughly weighted if
and only if there exists a potent certificate of non-weightedness whose length
is less than or equal to (n+1)nn/2 . Further research on roughly weighted sim-
ple games appears in [Gvozdeva et al., 2013, Freixas and Kurz, 2014, Hameed and Slinko, 2015].
In Section 5, we show that (under the the monotonicity property (1) and
∅ 6∈ W ∋ N ) the length of a potent certificate of non-weightedness is less
than or equal to 2αn+1 , if it exists. We also show that a roughly weighted
simple game (satisfying ∅ 6∈ W ∋ N ) has an integer vector (q; w ⊤ ) of rough
votingP representation satisfying 0 ≤ wi ≤ αn−1 (∀i ∈ N ), 0 ≤ q ≤ αn and
0 ≤ i∈N wi ≤ 2αn .

2 Trading Transforms of Non-weighted Simple Games

In this section, we discuss the size of a trading transform that guarantees
the non-weightedness of a given simple game. Throughout this section, we
do not need to assume the monotonicity property (1). First, we introduce a

4
linear inequality system for determining the weightedness of a given simple
game. For any nonempty family of player subsets ∅ =6 N ⊆ 2N , we introduce
a 0–1 matrix A(N ) = (a(N )Si ) whose rows are indexed by subsets in N and
columns are indexed by players in N defined by

1 (if i ∈ S ∈ N ),
a(N )Si =
0 (otherwise).

A given simple game G = (N, W) is weighted if and only if the following

linear inequality system is feasible:
 
  w
A(W) 1 0 
P1: −q  ≥ 0,
−A(L) −1 −1
ε
ε > 0,

where 0 (1) denotes a zero vector (all-one vector) of an appropriate dimen-

sion.
Farkas’ Lemma [Farkas, 1902] states that P1 is infeasible if and only if
the following system is feasible:
   
A(W)⊤ −A(L)⊤   0
x
D1:  1⊤ −1⊤  =  0 ,
⊤ ⊤ y
0 −1 −1
x ≥ 0, y ≥ 0.

For simplicity, we denote D1 by A1 z = c, z ≥ 0, where

   
A(W)⊤ −A(L)⊤   0
x
A1 =  1⊤ −1⊤  , z = , and c =  0  .
⊤ ⊤ y
0 −1 −1

Subsequently, we assume that D1 is feasible. Let A f1 z = e

c be a linear
equality system obtained from A1 z = c by repeatedly removing redundant
equalities. A column submatrix B b of Af1 is called a basis matrix if B
b is a
square invertible matrix. Variables corresponding to the columns of B b are
called basic variables, and JBb denotes an index set of basic variables. A
b is a vector z defined by
basic solution with respect to B

zbi (i ∈ JBb ),
zi =
0 (i 6∈ JBb ),

5
where z b is a vector of basic variables satisfying z
b=B b −1e c. It is well-known
that if a linear inequality system D1 is feasible, then it has a basic feasible
solution.
Let z ′ be a basic feasible solution of D1 with respect to a basis matrix B.
By Cramer’s rule, zi′ = det(Bi )/det(B) for each i ∈ JB , where Bi is a matrix
formed by replacing i-th column of B by e c. Because Bi is an integer matrix,
det(B)zi′ = det(Bi ) is an integer for any i ∈ JB . Let (x′⊤ , y ′⊤ )⊤ be a vector
corresponding to z ′ , and (x∗ ⊤ , y ∗ ⊤ ) = |det(B)|(x′⊤ , y ′⊤ ). Cramer’s rule
states that both x∗ and y ∗ are integer vectors. The pair of vectors x∗ and
y ∗ satisfies the following conditions:

A(W)⊤ x∗ − A(L)⊤ y ∗ = |det(B)|(A(W)⊤ x′ − A(L)⊤ y ′ ) = |det(B)|0 = 0,

X X
x∗S − yS∗ = |det(B)|(1⊤ x′ − 1⊤ y ′ ) = |det(B)|0 = 0,
S∈W S∈L
X
yS∗ = |det(B)|1⊤ y ′ = |det(B)|,
S∈L
x∗ = |det(B)|x′ ≥ 0, and y ∗ = |det(B)|y ′ ≥ 0.

Next, we construct a trading transform corresponding to the pair of x∗ and

y ∗ . Let X = (X1 , X2 , . . . , X|det(B)| ) be a sequence of winning coalitions,
where each winning coalition S ∈ W appears in X x∗S -times. Similarly, we
introduce a sequence Y = (Y1 , Y2 , . . . , Y|det(B)| ), where each losing coalition
S ∈ L appears in Y yS∗ -times. The above equalities imply that (X ; Y) is a
trading transform of size |det(B)|. Therefore, we have shown that if D1 is
feasible, then a given simple game G = (N, W) is not |det(B)|-trade robust.
Finally, we provide an upper bound on |det(B)|. Let αn be the maximum
of the determinants of n × n 0–1 matrices. For any n × n 0–1 matrix M, it is
easy to show that det(M ) ≥ −αn by swapping two rows of M (when n ≥ 2).
If a column of B is indexed by a component of x (i.e., indexed by a winning
coalition), then each component of the column is either 0 or 1. Otherwise,
a column (of B) is indexed by a component of y (i.e., indexed by a losing
coalition) whose components are either 0 or −1. Now, we apply elementary
matrix operations to B (see Figure 1). For each column of B indexed by
a component of y, we multiply the column by (−1). The resulting matrix,
denoted by B ′ , is a 0–1 matrix satisfying |det(B)| = |det(B ′ )|.
As B is a submatrix of A1 , the number of rows (columns) of B, denoted
by n′ , is less than or equal to n + 2. When n′ < n + 2, we obtain the
desired result: |det(B)| = |det(B ′ )| ≤ αn′ ≤ αn+1 . If n′ = n + 2, then B
has a row vector corresponding to equality 1⊤ x − 1⊤ y = 0, which satisfies
the condition that each component is either 1 or −1, and thus B ′ has an

6
 0 0 1 1 0 −1 0 0 1 1 0 1
0 1 0 1 0 0 0 1 0 1 0 0
1 0 0 1 0 −1 1 0 0 1 0 1
1 1 1 0 −1 −1 1 1 1 0 1 1
1 1 1 1 −1 −1 1 1 1 1 1 1
0 0 0 0 −1 −1 0 0 0 0 1 1
B B′

Figure 1: Example of elementary matrix operations for D1.

all-one row vector. Lemma 2.1 (c1) appearing below states that |det(B)| =
|det(B ′ )| ≤ αn′ −1 ≤ αn+1 .

Lemma 2.1. Let M be an n × n 0–1 matrix, where n ≥ 2.

(c1) If a row (column) vector of M is the all-one vector, then |det(M )| ≤ αn−1 .

(c2) If a row (column) vector of M is a 0–1 vector consisting of a unique

0-component and n − 1 1-components, then |det(M )| ≤ 2αn−1 .

Proof of (c1). Assume that the first column of M is the all-one vector. We
apply the following elementary matrix operations to M (see Figure 2). For
each column of M except the first column, if the first component is equal to
1, then we multiply the column by (−1) and add the all-one column vector.
The obtained matrix, denoted by M ′ , is an n × n 0–1 matrix satisfying
|det(M )| = |det(M ′ )|, and the first row is a unit vector. Thus, it is obvious
that |det(M ′ )| ≤ αn−1 .

1 1 0 1 0 1 0 0 0 0
1 1 1 1 0 1 0 1 0 0
1 0 1 0 0 1 1 1 1 0
1 1 1 0 1 1 0 1 1 1
1 0 0 1 1 1 1 0 0 1
M M′

Figure 2: Example of elementary matrix operations for (c1).

Proof of (c2). Assume that the first column vector of M , denoted by a,

contains exactly one 0-component. Obviously, e = 1 − a is a unit vector.
Let M1 and Me be a pair of matrices obtained from M with the first column

7
replaced by 1 and e, respectively. Then, it is easy to prove that

|det(M )| = |det(M1 ) − det(Me )| ≤ |det(M1 )| + |det(Me )| ≤ 2αn−1 .

QED
From the above discussion, we obtain the following theorem (without
the assumption of the monotonicity property (1)).

Theorem 2.2. A given simple game G = (N, W) with n players is weighted

if and only if G is αn+1 -trade robust, where αn+1 is the maximum of deter-
minants of (n + 1) × (n + 1) 0–1 matrices.

Proof. If a given simple game is not αn+1 -trade robust, then it is not trade
robust and, thus, not weighted, as shown by [Taylor and Zwicker, 1992,
Taylor and Zwicker, 1999]. We have discussed the inverse implication: if
a given simple game G is not weighted, then the linear inequality system P1
is infeasible. Farkas’ lemma [Farkas, 1902] implies that D1 is feasible. From
the above discussion, we have a trading transform (X1 , . . . , Xj ; Y1 , . . . Yj )
satisfying j ≤ αn+1 , X1 , . . . , Xj ∈ W, and Y1 , . . . , Yj ∈ L. QED
Applying the Hadamard’s evaluation [Hadamard, 1893] of the determi-
nant, we obtain Theorem 2.3.
n+1
Theorem 2.3. For any positive integer n, αn ≤ (n + 1) 2 (1/2)n .

The exact values of αn for small positive integers n appear in “The On-
Line Encyclopedia of Integer Sequences (A003432)” [Sloane et al., 2018] and
Table 1.

3 Integer Weights of Weighted Simple Games

This section reviews the integer-weight representations of weighted simple
games. Throughout this section, we do not need to assume the monotonicity
property (1), except in Table 1.

Theorem 3.1. Assume that a given simple game G = (N, W) satisfies

∅ 6∈ W ∋ N . If a given simple game G is weighted, then there exists an
integer-weight representation
P (q; w ⊤ ) of G satisfying |wi | ≤ αn (∀i ∈ N ),
|q| ≤ αn+1 , and 1 ≤ i∈N wi ≤ 2αn+1 − 1.

8
Proof. It is easy to show that a given simple game G = (N, W) is weighted
if and only if the following linear inequality system is feasible:

P2: A(W)w ≥ q1,

A(L)w ≤ q1 − 1,
1⊤ w ≤ u − 1.

We define
     
A(W) 1 0 w 0
A2 =  −A(L) −1 0  , v =  −q  , d =  1  ,
−1⊤ 0 1 u 1

and denote the inequality system P2 by A2 v ≥ d.

Subsequently, we assume that P2 is feasible. A non-singular submatrix
b of A2 is called a basis matrix. Variables corresponding to columns of B
B b
are called basic variables, and JBb denotes an index set of basic variables.
Let dBb be a subvector of d corresponding to rows of B.b A basic solution
with respect to Bb is a vector v defined by

vbi (i ∈ JBb ),
vi =
0 (i 6∈ JBb ),

where v b is a vector of basic variables satisfying v

b=B b −1 d b . It is well-known
B
that if a linear inequality system P2 is feasible, there exists a basic feasible
solution.
Let (w′⊤ , −q ′ , u′ )⊤ be a basic feasible solution of P2 with respect to a
basis matrix B. Assumption ∅ 6∈ W implies that 0 ≤ q ′ − 1 and, thus,
−q ′ 6= 0. As N ∈ W, we have inequalities u′ − 1 ≥ 1⊤ w ′ ≥ q ′ ≥ 1, which
imply that u′ 6= 0. The definition of a basic solution implies that −q and
u are basic variables with respect to the basis matrix B. Thus, B has
columns corresponding to basic variables −q and u. A column of B indexed
by u is called the last column. As B is invertible, the last column of B is
not the zero vector, and thus B includes a row corresponding to inequality
1⊤ w ≤ u − 1, which is called the last row (see Figure 3). Here, the number
of rows (columns) of B, denoted by n′ , is less than or equal to n + 2.
For simplicity, we denote the basic feasible solution (w ′⊤ , −q ′ , u′ )⊤ by
v . By Cramer’s rule, vi′ = det(Bi )/det(B) for each i ∈ JB , where Bi is
′

obtained from B with a column corresponding to variable vi replaced by dB .

Because Bi is an integer matrix, det(B)vi′ = det(Bi ) is an integer for any

9
i ∈ JB . Cramer’s rule states that (w ∗ ⊤ , −q ∗ , u∗ ) = |det(B)|(w ′⊤ , −q ′ , u′ ) is
an integer vector satisfying the following conditions:

A(W)w ∗ = |det(B)|A(W)w ′ ≥ |det(B)|q ′ 1 = q ∗ 1,

A(L)w ∗ = |det(B)|A(L)w ′ ≤ |det(B)|(q ′ 1 − 1) ≤ q ∗ 1 − 1, and
1⊤ w ∗ = |det(B)|1⊤ w′ ≤ |det(B)|(u′ − 1) ≤ u∗ − 1.

From the above, (q ∗ ; w ∗ ⊤ ) is an integer-weight representation of G. As

N ∈ W, we obtain 1⊤ w ∗ ≥ q ∗ = |det(B)|q ′ ≥ 1.

w1 w2 w3 w4 −q u
1 1 1 0 1 0
0 1 1 1 1 0
0 −1 −1 0 −1 0
−1 0 0 −1 −1 0
0 −1 0 −1 −1 0
−1 −1 −1 −1 0 1
B

w1 w2 w3 w4 −q u w1 w2 w3 w4 −q w1 w2 w3 w4 −q
1 1 1 0 0 0 1 1 1 0 0 1 1 1 0 0
0 1 1 1 0 0 0 1 1 1 0 0 1 1 1 0
0 −1 −1 0 1 0 0 −1 −1 0 1 0 1 1 0 1
−1 0 0 −1 1 0 −1 0 0 −1 1 1 0 0 1 1
0 −1 0 −1 1 0 0 −1 0 −1 1 0 1 0 1 1
−1 −1 −1 −1 1 1
Bq Bq′ Bq′′

w1 w2 w3 w4 −q u w1 w2 w3 w4 −q w1 w2 w3 w4 −q
1 0 1 0 1 0 1 0 1 0 1 1 0 1 0 1
0 0 1 1 1 0 0 0 1 1 1 0 0 1 1 1
0 1 −1 0 −1 0 0 1 −1 0 −1 0 1 1 0 1
−1 1 0 −1 −1 0 −1 1 0 −1 −1 1 1 0 1 1
0 1 0 −1 −1 0 0 1 0 −1 −1 0 1 0 1 1
−1 1 −1 −1 0 1
B2 B2′ B2′′

w1 w2 w3 w4 −q u w1 w2 w3 w4 −q u
1 1 1 0 1 0 1 1 1 0 1 0
0 1 1 1 1 0 0 1 1 1 1 0
0 −1 −1 0 −1 1 0 1 1 0 1 1
−1 0 0 −1 −1 1 1 0 0 1 1 1
0 −1 0 −1 −1 1 0 1 0 1 1 1
−1 −1 −1 −1 0 1 1 1 1 1 0 1
Bu ′
Bu

Figure 3: Examples of elementary matrix operations for P2.

Now, we discuss the magnitude of |q ∗ | = |det(Bq )|, where Bq is obtained

10
from B with a column corresponding to variable −q replaced by dB . As the
last column of Bq is a unit vector, we delete the last column and the last row
from Bq and obtain a matrix Bq′ satisfying det(Bq ) = det(Bq′ ). We apply
the following elementary matrix operations to Bq′ . First, we multiply the
column corresponding to variable −q (which is equal to dB ) by (−1). Next,
we multiply the rows indexed by losing coalitions by (−1). The resulting
matrix, denoted by Bq′′ , is 0–1 valued and satisfies the following condition:

|q ∗ | = |det(Bq )| = |det(Bq′ )| = |det(Bq′′ )| ≤ αn′ −1 ≤ αn+1 .

Next, we show that |wi∗ | ≤ αn (i ∈ N ). If wi∗ 6= 0, then wi is a basic

variable that satisfies |wi∗ | = |det(Bi )|, where Bi is obtained from B but
the column corresponding to variable wi is replaced by dB . In a manner
similar to that above, we delete the last column and the last row from Bi
and obtain a matrix Bi′ satisfying det(Bi ) = det(Bi′ ). Next, we multiply a
column corresponding to variable wi by (−1). We multiply rows indexed by
losing coalitions by (−1) and obtain a 0–1 matrix Bi′′ . Matrix Bi contains
a column corresponding to the original variable −q, which contains values 1
or −1. Thus, matrix Bi′′ contains a column vector that is equal to an all-one
vector. Lemma 2.1 (c1) implies that

|wi∗ | = |det(Bi )| = |det(Bi′ )| = |det(Bi′′ )| ≤ αn′ −2 ≤ αn .

Lastly, we discuss the value of |u∗ | = |det(Bu )|, where Bu is obtained

from B but the last column (column indexed by variable u) is replaced by
dB . In a manner similar to that above, we multiply the last column by
(−1), multiply the rows indexed by losing coalitions by (−1), and multiply
the last row by (−1). The resulting matrix, denoted by Bu′ , is a 0–1 matrix
in which the last row contains exactly one 0-component (indexed by variable
−q). Lemma 2.1 (c2) implies that

|u∗ | = |det(Bu )| = |det(Bu′ )| ≤ 2αn′ −1 ≤ 2αn+1 ,

and thus 1⊤ w ∗ ≤ u∗ − 1 ≤ |u∗ | − 1 ≤ 2αn+1 − 1. QED

[Kurz, 2012] exhaustively generated all weighted voting games satisfying
the monotonicity property (1) for up to nine voters. Table 1 shows max-
ima of the exact values of minimal integer-weight representations obtained
by [Kurz, 2012], Muroga’s bounds in [Muroga, 1971], and our upper bounds.
The table shows that our bound on the quota is tight when n ≤ 5.

11
 Table 1: Exact values of integer weights representations.
n 1 2 3 4 5 6 7 8 9 10 11
αn † 1 1 2 3 5 9 32 56 144 320 1458
max min max wi ‡ 1 1 2 3 5 9 18 42 110
(N,W) [q;w] i
Muroga’s bound (αn )• 1 1 2 3 5 9 32 56 144 320 1458
max min q ‡ 1 2 3 5 9 18 40 105 295
(N,W) [q;w]
Our bound (αn+1 ) 1 2 3 5 9 32 56 144 320 1458
Muroga’s bound (nαn )• 1 2 6 12 25 54 224 448 1296 3200 16038
P
max min i wi ‡ 1 2 4 8 15 33 77 202 568
(N,W) [q;w]
Our bound (2αn+1 − 1) 1 3 5 9 17 63 111 287 639 2915
† [Sloane et al., 2018], ‡ [Kurz, 2012], • [Muroga, 1971].

4 Rounding Method
This section addresses the problem of finding integer-weight representations.
In this section, we assume the monotonicity property (1). In addition, a
weighted simple game is given by a triplet (N, W m , LM ), where W m and
LM denote the set of minimal winning coalitions and the set of maximal
losing coalitions, respectively. We also assume that the empty set is a losing
coalition, N is a winning coalition, and every player in N is not a null
player. Thus, there exists an integer-weight representation in which q ≥ 1
and wi ≥ 1 (∀i ∈ N ).
We discuss a problem for finding an integer-weight representation, which
is formulated by the following integer programming problem:

Q: find a vector (q; w)

X
satisfying wi ≥ q (∀S ∈ W m ), (3)
i∈S
X
wi ≤ q − 1 (∀S ∈ LM ), (4)
i∈S
q ≥ 1, wi ≥ 1 (∀i ∈ N ), (5)
q ∈ Z, wi ∈ Z (∀i ∈ N ). (6)

A linear relaxation problem Q is obtained from Q by dropping the integer

constraints (6).
Let (q ∗ ; w∗ ⊤ ) be a basic feasible solution of the linear inequality sys-
tem Q. Our proof in the previous section showed that |det(B ∗ )|(q ∗ ; w∗ ⊤ )

12
gives a solution of Q (i.e., an integer-weight representation), where B ∗ de-
notes a corresponding basis matrix of Q. When |det(B ∗ )| > n, there ex-
ists a simple method for generating a smaller integer-weight representation.
For any weight vector w = (w1 , w2 , . . . , wn )⊤ , we denote the integer vector
(⌊w1 ⌋, ⌊w2 ⌋, . . . , ⌊wn ⌋)⊤ by ⌊w⌋. Given a solution (q ∗ ; w ∗ ⊤ ) of Q, we intro-
duce an integer vector w′ = ⌊nw∗ ⌋ and an integer q ′ = ⌊n(q ∗ − 1)⌋ + 1. For
any minimal winning coalition S ∈ W m , we have that
X X X
wi′ > (nwi∗ − 1) ≥ n wi∗ − n ≥ nq ∗ − n = n(q ∗ − 1) ≥ ⌊n(q ∗ − 1)⌋,
i∈S i∈S i∈S
X
wi′ ≥ ⌊n(q ∗ − 1)⌋ + 1 = q ′ .
i∈S

Each maximal losing coalition S ∈ LM satisfies

X X
wi′ ≤ nwi∗ ≤ n(q ∗ − 1),
i∈S i∈S
X
wi′ ≤ ⌊n(q ∗ − 1)⌋ = q ′ − 1.
i∈S

Thus, the pair of w′ and q ′ gives an integer-weight representation satisfying

(q ′ ; w′⊤ ) ≤ n(q ∗ ; w ∗⊤ ). In the remainder of this section, we show that there
exists an√integer-weight
√ representation (vector) that is less than or equal
to ((2 − 2)n + ( 2 − 1))(q ∗ ; w∗ ⊤ ) < (0.5858n + 0.4143)(q ∗ ; w∗ ⊤ ) for any
solution (q ∗ ; w∗ ⊤ ) of Q.
∗⊤
√
Theorem
√ 4.1. Let (q ∗ ; w √ ) be a√solution of Q. We define ℓ1 = (2− 2)n−
( 2 − 1) and u1 = (2 − 2)n + ( 2 − 1). Then, there exists a real number
λ• ∈ [ℓ1 , u1 ] so that the pair Q = ⌊λ• (q ∗ − 1)⌋ + 1 and W = ⌊λ• w∗ ⌋ gives a
feasible solution of Q (i.e., an integer-weight representation).
Proof. For any positive real λ, it is easy to see that each maximal losing
coalition S ∈ LM satisfies
X X
⌊λwi∗ ⌋ ≤ λwi∗ ≤ λ(q ∗ − 1),
i∈S i∈S
X
⌊λwi∗ ⌋ ≤ ⌊λ(q ∗ − 1)⌋.
i∈S

To discuss the weights

P of minimal winning coalitions, we introduce a
function g(λ) = λ− i∈N (λwi∗ −⌊λwi∗ ⌋). In the second part of this proof, we
show that if we choose Λ ∈ [ℓ1 , u1 ] uniformly at random, then E[g(Λ)] ≥ 0.

13
This implies that ∃λ• ∈ [ℓ1 , u1 ] satisfying g(λ• ) > 0, because g(λ) is right-
continuous, piecewise linear, and not a constant function. When g(λ• ) > 0,
each minimal winning coalition S ∈ W m satisfies
X X X X
λ• > (λ• wi∗ − ⌊λ• wi∗ ⌋) ≥ (λ• wi∗ − ⌊λ• wi∗ ⌋) = λ• wi∗ − ⌊λ• wi∗ ⌋,
i∈N i∈S i∈S i∈S
(7)
which implies
!
X X X
•
⌊λ wi∗ ⌋ > •
λ wi∗ − λ• =λ •
wi∗ −1 ≥ λ• (q ∗ − 1) ≥ ⌊λ• (q ∗ − 1)⌋,
i∈S i∈S i∈S

and thus X
⌊λ• wi∗ ⌋ ≥ ⌊λ• (q ∗ − 1)⌋ + 1.
i∈S

Finally, we show that E[g(Λ)] ≥ 0 if we choose Λ ∈ [ℓ1 , u1 ] uniformly at

random. It is obvious that
Z u1
∗ ∗
X ℓ 1 + u1
X ℓ (λwi − ⌊λwi ⌋)dλ
E[g(Λ)] = E[Λ] − E[(Λwi∗ − ⌊Λwi∗ ⌋)] = − 1
2 u1 − ℓ1
i∈N i∈N
Z u1
∗ ∗
√ X ℓ (λwi − ⌊λwi ⌋)dλ
= (2 − 2)n − 1
.
u1 − ℓ1
i∈N

Let us discuss the last term appearing above. By substituting µ for λwi∗ , we
obtain

Z u1
Z u1 wi∗
(λwi∗ − ⌊λwi∗ ⌋)dλ (µ − ⌊µ⌋)dµ
ℓ1 ℓ1 wi∗
=
u1 − ℓ 1 wi∗ (u1 − ℓ1 )
Z 0 Z 0
(µ − ⌊µ⌋)dµ (µ − ⌊µ⌋)dµ
−wi∗ (u1 −ℓ1 ) −x
≤ = ,
wi∗ (u1 − ℓ1 ) x
where
√ the last equality is obtained by setting x = wi∗ (u1 − ℓ1 ).
√ As u1 − ℓ1 =
2( 2 − 1) and wi ≥ 1, it is clear that x = wi∗ (u1 − ℓ1 ) ≥ 2( 2 − 1). Here,
∗
Z 0
(µ − ⌊µ⌋)dµ
we introduce a function f (x) = −x . According to numerical
x

14
 √
calculations
√ (see Figure 4), inequality x ≥ 2( 2 − 1) implies that f (x) ≤
2 − 2.

0.8
0.75
0.7
f(x)

0.65
0.6
0.55
0.5
0.45
0 1 2 3 4 5
x

Z 0
(µ − ⌊µ⌋)dµ
−x
Figure 4: Plot of function f (x) = .
x

From the above, we obtain the desired result

√ X √ √ √
E[g(Λ)] ≥ (2 − 2)n − (2 − 2) = (2 − 2)n − (2 − 2)n = 0.
i∈N

QED

5 Roughly Weighted Simple Games

In this section, we discuss roughly weighted simple games. First, we show
an upper bound of the length of a potent certificate of non-weightedness.

Theorem 5.1. Assume that a given simple game G = (N, W) satisfies ∅ 6∈

W ∋ N and the monotonicity property (1). If a given simple game G is not
roughly weighted, then there exists a potent certificate of non-weightedness
whose length is less than or equal to 2αn+1 .

Proof. Let us introduce a linear inequality system:

  
A(W) 1 w
P3: ≥ 0,
−A(L) −1 −q
1⊤ w > 0.

15
First, we show that if P3 is feasible, then a given simple game is roughly
weighted. Let (q ′ ; w′⊤ ) be a feasible solution of P3. We introduce a new
voting weight wi′′ = max{wi′ , 0} for each i ∈ N . We show that (q ′ ; w ′′⊤ ) is a
rough voting representation. As 1⊤ w′ > 0, vector w′ includes Pat least one
positive component, and thus w ′′ 6= 0. If a coalition S satisfies w ′′ < q ′ ,
P P i∈S i
then q ′ > i∈S wi′′ ≥ i∈S w P
′ , and thus S is losing. Consider the case in
i
which a coalition SPsatisfies P ′′ ′ Let S ′ = {i ∈ S | w ′ > 0}. It is
i∈S wi > q .P i
obvious that q < i∈S wi = i∈S ′ wi = i∈S ′ wi′ and thus S ′ is winning.
′ ′′ ′′

The monotonicity property (1) and S ′ ⊆ S imply that S is winning.

From the above discussion, it is obvious that if a given simple game is
not roughly weighted, then P3 is infeasible. Farkas’ Lemma [Farkas, 1902]
states that
    
A(W)⊤ −A(L)⊤ x −1
D3: = ,
1⊤ −1⊤ y 0
x ≥ 0, y ≥ 0,

is feasible if and only if P3 is infeasible. By introducing an artificial non-

negative variable u ≥ 0 and equality 1⊤ x = u − 1, we obtain a linear
inequality system:
    
A(W)⊤ −A(L)⊤ 0 x −1
D3+ :  1⊤ −1⊤ 0  y  =  0 ,
1 ⊤ 0 ⊤ −1 u −1
x ≥ 0, y ≥ 0, u ≥ 0.

It is obvious that D3 is feasible if and only if D3+ is feasible.

Next, we construct a trading transform from a basic feasible solution
of D3+ . Let A c′ be a linear equality system obtained from D3+ by
f3 z = e
repeatedly removing redundant equalities. As D3+ is feasible, there exists a
basic feasible solution, denoted by z ′ , and a corresponding basis matrix B
of Af3 . From Cramer’s rule, z ′ 6= 0 implies that z ′ = det(BS )/det(B) for
S S
each S ⊆ N, where BS is obtained from B with a column corresponding to
variable zS replaced by e c′ . Obviously, |det(B)|z ′ is a non-negative integer
vector. We denote by (x , y ′⊤ , u′ )⊤ the basic feasible solution z ′ . We recall
′⊤

that ∅ 6∈ W ∋ N and introduce a pair of non-negative integer vectors (x∗ , y ∗ )

defined as follows:

∗ |det(B)|x′S (if S ∈ W \ {N }),
xS =
|det(B)|(x′N + 1) (if S = N ),

|det(B)|yS′ (if S ∈ N \ {∅}),
yS∗ =
|det(B)|(y∅′ + 1) (if S = ∅).

16
Subsequently, χ(S) denotes the characteristic vector of a coalition S. It is
easy to see that pair (x∗ , y ∗ ) satisfies
X X
A(W)⊤ x∗ − A(L)⊤ y ∗ = χ(S)x∗S − χ(S)yS∗
S∈W S∈L
X X
= χ(S)|det(B)|x′S + χ(N )|det(B)| − χ(S)|det(B)|yS′ − χ(∅)|det(B)|
S∈W S∈L
! !
X X
= |det(B)| χ(S)x′S − χ(S)yS′ + χ(N ) − χ(∅)
S∈W S∈L
  
= |det(B)| A(W)⊤ x′ − A(L)⊤ y ′ + 1 − 0 = |det(B)|(−1 + 1 − 0) = 0

and
X X X X
x∗S − yS∗ = |detB|x′S + |det(B)| − |det(B)|yS′ − |det(B)|
S∈W S∈L S∈W S∈L
! !
X X
= |det(B)| x′S − yS′ +1−1 = |det(B)|(0 + 1 − 1) = 0.
S∈W S∈L

Next, we can construct a trading transform (X ; Y) corresponding to the

pair of x∗ and y ∗ by analogy with the proof of Theorem 2.2. Both x∗N and
y∅∗ are positive and ∅ 6∈ W ∋ N ; therefore (X ; Y) is a potent certificate of
non-weightedness.
Lastly, we discuss the length of (X ; Y). The number of rows (columns)
of B, denoted by n′ , is less than or equal to n+2. The basic feasible solution
z ′ satisfies that u′ = 1 + 1⊤ x′ ≥ 1 > 0, and thus Cramer’s rule states that
det(B)u′ = det(Bu ) (Figure 5 shows an example). We multiply columns
of Bu corresponding to components in (y ⊤ , u) by (−1) and obtain a 0–1
matrix Bu′ satisfying |det(Bu )| = |det(Bu′ )|. As ec′ includes at most one 0-
′
component, Lemma 2.1 implies that |det(Bu )| ≤ 2αn′ −1 ≤ 2αn+1 . Thus, the
length of (X ; Y) satisfies
X X
x∗S = |det(B)|x′S + |det(B)| = |det(B)|(1⊤ x′ + 1)
S∈W S∈W
= |det(B)|(u′ − 1 + 1) = |det(B)|u′ = |det(B)u′ |
= |det(Bu )| = |det(Bu′ )| ≤ 2αn+1 .
QED
In the rest of this section, we discuss integer voting weights and a quota
of a rough voting representation. We say that a player i ∈ N is a passer if
and only if every coalition S ∋ i is winning.

17
 u
0 0 1 1 0 −1 0
0 1 0 1 0 0 0
1 0 0 1 0 −1 0
1 1 1 0 −1 −1 0
1 1 1 1 −1 −1 0
1 1 1 1 0 0 −1
B

c′
e
0 0 1 1 0 −1 −1 0 0 1 1 0 1 1
0 1 0 1 0 0 −1 0 1 0 1 0 0 1
1 0 0 1 0 −1 −1 1 0 0 1 0 1 1
1 1 1 0 −1 −1 −1 1 1 1 0 1 1 1
1 1 1 1 −1 −1 0 1 1 1 1 1 1 0
1 1 1 1 0 0 −1 1 1 1 1 0 0 1
Bu Bu′

Figure 5: Example of elementary matrix operations for D3+ .

Theorem 5.2. Assume that a given simple game G = (N, W) satisfies

∅ 6∈ W ∋ N. If a given simple game G is roughly weighted, then there exists
an integer vector (q; w⊤ ) of the rough voting
P representation satisfying 0 ≤
wi ≤ αn−1 (∀i ∈ N ), 0 ≤ q ≤ αn , and 1 ≤ i∈N wi ≤ 2αn .

Proof. First, we show that if a given game is roughly weighted, then either
   
A(W) 0   1
w
P4:  −A(L) 0  ≥  −1  , w ≥ 0, u ≥ 0,
⊤ u
−1 1 0

is feasible or there exists at least one passer. Suppose that a given simple
game has a rough voting representation (q; w ⊤ ). If q > 0, then (1/q)w
becomes a feasible solution of P4 by setting u to a sufficiently large positive
number. Consider the case q ≤ 0. Assumption ∅ 6∈ W implies that 0 ≤ q,
and thus we obtain q = 0. Properties (q, w ⊤ ) 6= 0⊤ and w ≥ 0 imply that
∃i◦ ∈ N, wi◦ > 0, i.e., a given game G has a passer i◦ .
When a given game G has a passer i◦ ∈ N , then there exists a rough

18
voting representation (q ◦ ; w ◦ ⊤ ) defined by

1 (i = i◦ ), ◦
wi◦ = q = 0,
0 (i 6= i◦ ),

which produces the desired result.

Lastly, we consider the case in which P4 is feasible. It is well-known that
when P4 is feasible, there exists a basic feasible solution. Let (w′⊤ , u′ )⊤ be
a basic feasible solution of P4 and B be a corresponding basis matrix. It is
easy to see that (1; w ′⊤ ) is a rough voting representation of G. Assumption
N ∈ W implies the positivity of u′ because u′ ≥ 1⊤ w′ ≥ 1. Then, variable
u is a basic variable, and thus B includes a column corresponding to u,
which is called the last column. The non-singularity of B implies that a
column corresponding to u is not the zero vector, and thus B includes a row
corresponding to the inequality 1⊤ w ≤ u, which is called the last row (see
Figure 6). The number of rows (columns) of basis matrix B, denoted by n′ ,
is less than or equal to n + 1.
Cramer’s rule states that (q ∗ , w ∗ ⊤ , u∗ ) = |det(B)|(1, w ′⊤ , u′ ) is a non-
negative integer vector. It is easy to see that (q ∗ , w ∗ ⊤ , u∗ ) satisfies

A(W)w ∗ = |det(B)|A(W)w ′ ≥ |det(B)|1 = q ∗ 1,

A(L)w ∗ = |det(B)|A(L)w ′ ≤ |det(B)|1 = q ∗ 1, and
1⊤ w ∗ = |det(B)|1⊤ w′ ≤ |det(B)|u′ = u∗ .

From the above, (q ∗ ; w ∗ ⊤ ) is an integer vector of a rough voting represen-

tation. Assumption N ∈ W implies that 1⊤ w ∗ ≥ q ∗ = |det(B)| ≥ 1.
Let d′B be a subvector of the right-hand-side vector of an inequality sys-
tem in P4 corresponding to rows of B. Cramer’s rule states that det(B)u′ =
det(Bu ), where Bu is obtained from B but the column corresponding to a
basic variable u is replaced by d′B (see Figure 6). We multiply rows of Bu
that correspond to losing coalitions by (−1) and multiply the last row by
(−1). The resulting matrix, denoted by Bu′ , is a 0–1 matrix whose last row
includes exactly one 0-component (indexed by u). Lemma 2.1 (c2) implies
that |det(Bu′ )| ≤ 2αn′ −1 ≤ 2αn . Thus, we obtain that

1⊤ w ∗ ≤ u∗ ≤ |u∗ | = |det(B)u′ | = |det(Bu )| = |det(Bu′ )| ≤ 2αn .

By analogy with the proof of Theorem 3.1, we can prove the desired inequal-
ities: q ∗ = |det(B)| ≤ αn and wi∗ ≤ αn−1 (∀i ∈ N ). QED

19
 w1 w2 w3 w4 w5 u w1 w2 w3 w4 w5 u w1 w2 w3 w4 w5 u
1 1 1 0 1 0 1 1 1 0 1 1 1 1 1 0 1 1
0 1 0 1 1 0 0 1 0 1 1 1 0 1 0 1 1 1
0 −1 −1 0 0 0 0 −1 −1 0 0 −1 0 1 1 0 0 1
−1 0 0 −1 −1 0 −1 0 0 −1 −1 −1 1 0 0 1 1 1
0 −1 0 −1 0 0 0 −1 0 −1 0 −1 0 1 0 1 0 1
−1 −1 −1 −1 −1 1 −1 −1 −1 −1 −1 0 1 1 1 1 1 0
B Bu ′
Bu

Figure 6: Examples of elementary matrix operations for P4.

6 Conclusion
In this paper, we discussed the smallest value of k∗ such that every k∗ -trade
robust simple game would be weighted. We provided a new proof of the
existence of a trading transform when a given simple game is non-weighted.
Our proof yields an improved upper bound on the required length of a
trading transform. We showed that a given simple game G is weighted if
and only if G is αn+1 -trade robust, where αn+1 denotes the maximal value
of determinants of (n + 1) × (n + 1) 0–1 matrices. Applying the Hadamard’s
evaluation [Hadamard, 1893] of the determinant, we obtain k∗ ≤ αn+1 ≤
n+2
(n + 2) 2 (1/2)(n+1) , which improves the existing bound k∗ ≤ (n + 1)nn/2
obtained by [Gvozdeva and Slinko, 2011].
Next, we discussed upper bounds for the maximum possible integer
weights and the quota needed to represent any weighted simple game with n
players. We show that every weighted simple game (satisfying ∅ 6∈ W ∋ N )
has an integer-weight representation
P (q; w ⊤ ) ∈ Z × ZN such that |wi | ≤ αn
(∀i ∈ N ), |q| ≤ αn+1 , and 1 ≤ i∈N wi ≤ 2αn+1 − 1. We demonstrated the
tightness of our bound on the quota when n ≤ 5.
We described a rounding method based on a linear relaxation of an
integer programming problem for finding an integer-weight representation.
We showed that an integer-weight representation is obtained by carefully
rounding
√ a solution
√ of the linear inequality system multiplied by λ• ≤
(2− 2)n+( 2−1) < 0.5858n+0.4143. Our proof of Theorem 4.1 indicates
an existence of a randomized rounding algorithm for finding an appropriate
value λ• . However, from theoretical point of view, Theorem 4.1 only showed
the existence of a real number λ• . Even if there exists an appropriate “ratio-
nal” number λ• , we need to determine the size of the rational number (its
numerator and denominator) to implement a naive randomized rounding
algorithm. Thus, it remains open whether there exists an efficient algo-

20
rithm for finding an integer-weight representation satisfying the properties
in Theorem 4.1.
Lastly, we showed that a roughly weighted simple game (satisfying ∅ 6∈
W ∋ N ) has an integer vector (q; w ⊤ ) of the rough voting Prepresentation
satisfying 0 ≤ wi ≤ αn−1 (∀i ∈ N ), 0 ≤ q ≤ αn , and 1 ≤ i∈N wi ≤ 2αn .
When a given simple game is not roughly weighted, we showed that (under
the the monotonicity property (1) and ∅ 6∈ W ∋ N ) there existed a potent
certificate of non-weightedness whose length is less than or equal to 2αn+1 .

References
[Baugh, 1970] Baugh, C. R. (1970). Pseudo-threshold logic: A generalization
of threshold logic. PhD thesis, University of Illinois at Urbana-Champaign.
[Chow, 1961] Chow, C.-K. (1961). On the characterization of threshold
functions. In 2nd Annual Symposium on Switching Circuit Theory and
Logical Design (SWCT 1961), pages 34–38. IEEE.
[Elgot, 1961] Elgot, C. C. (1961). Decision problems of finite automata de-
sign and related arithmetics. Transactions of the American Mathematical
Society, 98(1):21–51.
[Farkas, 1902] Farkas, J. (1902). Theorie der einfachen Ungleichungen.
Journal für die reine und angewandte Mathematik, 1902(124):1–27.
[Freixas, 2021] Freixas, J. (2021). A characterization of weighted simple
games based on pseudoweightings. Optimization Letters, 15:1371—-1383.
[Freixas et al., 2016] Freixas, J., Freixas, M., and Kurz, S. (2016).
Characterization of threshold functions: state of the art, some
new contributions and open problems. Available at SSRN.
https://ssrn.com/abstract=2740475(March1,2016).
[Freixas et al., 2017] Freixas, J., Freixas, M., and Kurz, S. (2017). On
the characterization of weighted simple games. Theory and Decision,
83(4):469–498.
[Freixas and Kurz, 2014] Freixas, J. and Kurz, S. (2014). On α-roughly
weighted games. International Journal of Game Theory, 43(3):659–692.
[Freixas and Molinero, 2009a] Freixas, J. and Molinero, X. (2009a). On the
existence of a minimum integer representation for weighted voting sys-
tems. Annals of Operations Research, 166(1):243–260.

21
[Freixas and Molinero, 2009b] Freixas, J. and Molinero, X. (2009b). Simple
games and weighted games: a theoretical and computational viewpoint.
Discrete Applied Mathematics, 157(7):1496–1508.

[Freixas and Molinero, 2010] Freixas, J. and Molinero, X. (2010). Weighted

games without a unique minimal representation in integers. Optimisation
Methods & Software, 25(2):203–215.

[Gvozdeva et al., 2013] Gvozdeva, T., Hemaspaandra, L. A., and Slinko, A.

(2013). Three hierarchies of simple games parameterized by “resource”
parameters. International Journal of Game Theory, 42(1):1–17.

[Gvozdeva and Slinko, 2011] Gvozdeva, T. and Slinko, A. (2011). Weighted

and roughly weighted simple games. Mathematical Social Sciences,
61(1):20–30.

[Hadamard, 1893] Hadamard, J. (1893). Résolution d’une question relative

aux déterminants. Bull. des Sciences Math., 2:240–246.

[Hameed and Slinko, 2015] Hameed, A. and Slinko, A. (2015). Roughly

weighted hierarchical simple games. International Journal of Game The-
ory, 44(2):295–319.

[Hansen and Podolskii, 2015] Hansen, K. A. and Podolskii, V. V. (2015).

Polynomial threshold functions and Boolean threshold circuits. Informa-
tion and Computation, 240:56–73.

[Håstad, 1994] Håstad, J. (1994). On the size of weights for threshold gates.
SIAM Journal on Discrete Mathematics, 7(3):484–492.

[Isbell, 1956] Isbell, J. R. (1956). A class of majority games. The Quarterly

Journal of Mathematics, 7(1):183–187.

[Krohn and Sudhölter, 1995] Krohn, I. and Sudhölter, P. (1995). Directed

and weighted majority games. Zeitschrift für Operations Research,
42(2):189–216.

[Kurz, 2012] Kurz, S. (2012). On minimum sum representations for

weighted voting games. Annals of Operations Research, 196(1):361–369.

[Muroga, 1971] Muroga, S. (1971). Threshold Logic and its Applications.

Wiley, New York.

22
[Muroga et al., 1962] Muroga, S., Toda, I., and Kondo, M. (1962). Majority
decision functions of up to six variables. Mathematics of Computation,
16(80):459–472.

[Muroga et al., 1970] Muroga, S., Tsuboi, T., and Baugh, C. R. (1970).
Enumeration of threshold functions of eight variables. IEEE Transactions
on Computers, 100(9):818–825.

[Myhill and Kautz, 1961] Myhill, J. and Kautz, W. H. (1961). On the size
of weights required for linear-input switching functions. IRE Transactions
on Electronic Computers, EC-10(2):288–290.

[Peled and Simeone, 1985] Peled, U. N. and Simeone, B. (1985).

Polynomial-time algorithms for regular set-covering and threshold syn-
thesis. Discrete Applied Mathematics, 12(1):57–69.

[Sloane et al., 2018] Sloane, N. J. et al. (2018). The on-line encyclopedia of

integer sequences (A003432). Published electronically,.

[Taylor and Zwicker, 1992] Taylor, A. D. and Zwicker, W. S. (1992). A

characterization of weighted voting. Proceedings of the American mathe-
matical society, 115(4):1089–1094.

[Taylor and Zwicker, 1999] Taylor, A. D. and Zwicker, W. S. (1999). Sim-

ple Games: Desirability Relations, Trading, Pseudoweightings. Princeton
University Press.

[Wang and Williams, 1991] Wang, C. and Williams, A. (1991). The thresh-
old order of a Boolean function. Discrete Applied Mathematics, 31(1):51–
69.

[Winder, 1965] Winder, R. O. (1965). Enumeration of seven-argument

threshold functions. IEEE Transactions on Electronic Computers, EC-
14(3):315–325.

23