James Edward Buttersworth (1817-1894) was an English painter.
The notation \(N^{v}\) will be used to denote the set of players in the game with characteristic function \(v\).
Definition. (Pechersky and Sobolev \cite[p.58]{pec95}). A solution is a set-valued map \(F:v\rightarrow F(v)\subset \mathbb{R}^{N^{v}}\) with \(F(v)\neq\emptyset\) for every game \(v\). The set \(F(v)\) is called the solution of the game \(v\). \(\sharp\)
Definition. (Pechersky and Sobolev \cite[p.58]{pec95}). A mapping \(\tau :N^{v_{1}}\rightarrow N^{v_{2}}\) is called an isomorphism of games \(v_{1}\) and \(v_{2}\) if \(\tau\) is one-to-one mapping and \(v_{2}(\tau (S))=v_{1}(S)\) for every \(S\subseteq N^{v_{1}}\). The games \(v_{1}\) and \(v_{2}\) are called isomorphic. \(\sharp\)
If \(\tau\) is an isomorphism, then \(\tau^{*}:\mathbb{R}^{I^{v_{1}}}\rightarrow\mathbb{R}^{I^{v_{2}}}\) denotes the mapping defined by the formula \(\tau^{*}({\bf x})=(x_{\tau (i)})_{i\in I^{v_{1}}}\) for every \({\bf x}\in\mathbb{R}^{I^{v_{1}}}\) (Pechersky and Sobolev \cite[p.58]{pec95}).
For any real-valued function \(\eta\) and evert game \(v\), let \(\eta v\) be the game defined by \((\eta v)(N)=(\eta v)(\emptyset )=0\) and \((\eta v)(S)=\eta (v(S))\) for \(S\neq N,\emptyset\). The \(\eta\)-sum $v_{1}\oplus_{\eta}v_{2}$ of games \(v_{1}\) and \(v_{2}\) is the set of games \(v_{3}\) such that \(v_{3}(N)=0\) and \(\eta v_{3}=\frac{1}{2}(\eta v_{1}+\eta v_{2})\). The \(\eta\)-sum of the games \(v_{1}\) and \(v_{2}\) always exists for any continuous functions \(\eta\) (Pechersky and Sobolev \cite[p.58]{pec95}).
It is convenient to consider sometimes vector \({\bf x}\in\mathbb{R}^{N}\) as a game defined by formula \({\bf x}(S)=\sum_{i\in S}x_{i}\) for all \(S\subseteq N\). We denote by \(Pr_{\mathbb{R}^{S}}(A)\) the projection of the set \(A\) on the subspace \(\mathbb{R}^{S}\) for \(S\subseteq N\). \(A+B=\{{\bf x}+{\bf y}:{\bf x}\in A,{\bf y}\in B\}\). The Shapley value \(\boldsymbol{\phi}\) is given by
\[\phi_{i}(v)=\sum_{\{S:i\not\in S\}}\frac{s!(n-s-1)!}{n!}\cdot(v(S\cup\{i\})-v(S)),\]
where \(s=|S|\) and \(n=|N^{v}|\). (Pechersky and Sobolev \cite[p.59]{pec95}).
We now formulate the axioms which we claim a solution \(F\) to satisfy.
- Axiom PS1: (Anonymity). If \({\bf x}\in F(v_{1})\) and \(\tau\) is an isomorphism of the games \(v_{1}\) and \(v_{2}\), then \(\tau^{*}({\bf x})\in F(v_{2})\).
- Axiom PS2: (Symmetry). If \(v(S)=f(|S|)\) and \(v(N)=0\), then \({\bf 0}\in F(v)\).
- Axiom PS3: (Dummy). Let \(v_{1}\) and \(v_{2}\) be the games such that \(N^{v_{1}}=N\), \(N^{v_{2}}=N\cup\{i_{0}\}\) with \(i_{0}\not\in N\) and \(v_{1}(S)=v_{2}(S)=v_{2}(S\cup\{i_{0}\})\) for all \(S\subseteq N\), then \(x_{i_{0}}=0\) for every \({\bf x}\in F(v)\) and \(Pr_{\mathbb{R}^{N}}(F(v_{2}))=F(v_{1})\).
- Axiom PS4: (Efficiency). If \({\bf x}\in F(v)\), then \({\bf x}(I)=v(I)\).
- Axiom PS5: (Weak covariance). Let \({\bf a}\in\mathbb{R}^{N}\), and two games \(v_{1}\) and \(v_{2}\) are connected by equality \(v_{2}=v_{1}+{\bf a}\). Then \(F(v_{2})=F(v_{1})+{\bf a}\).
- Axiom PS6($\eta$): ($\eta$-closedness of the null games set}. If \({\bf 0}\in F(v_{1})\cap F(v_{2})\), \(v_{3}\in v_{1}\oplus_{\eta}v_{2}\), then \({\bf 0}\in F(v_{3})\).
The Axiom (A6)($\eta$) depends on the function \(\eta\). Therefore, we obtain the whole family of axiom systems, and every system possibly defines some solution \(F\) (Pechersky and Sobolev \cite[p.59-60]{pec95}).
Proposition. (Pechersky and Sobolev \cite[p.60]{pec95}). Let \(v_{k}\), \(k=1,\cdots ,m\) be the games with the same set of players. Suppose that
\[{\bf 0}\in\bigcap_{k=1}^{m}F(v_{k})\mbox{ and }\eta v=\sum_{k=1}^{m}\lambda_{k}\cdot\eta v_{k},\]
where \(\lambda_{k}\) are the binary-rational numbers, i.e., the numbers of the form \(p/2^{q}\), such that \(\lambda_{k}>0\) and \(\sum_{k=1}^{m}\lambda_{k} =1\). Then \({\bf 0}\in F(v)\). \(\sharp\)
Proposition. (Pechersky and Sobolev \cite[p.60]{pec95}). If \({\bf 0}\in F(v_{1})\), \(v_{2}(N)=0\), \(\eta v_{1}=\eta v_{2}\), then \({\bf 0}\in F(v_{2})\). \(\sharp\)
Proposition. (Pechersky and Sobolev \cite[p.60]{pec95}). Let \(v\) be a game such that \(v(N)=0\) and \(v(S\cup\{i_{0}\})=v(S)=f(|S|)\) for some \(i_{0}\in N\) and every \(S\) with \(i_{0}\not\in S\). Then \({\bf 0}\in F(v)\). \(\sharp\)
Proposition. (Pechersky and Sobolev \cite[p.60]{pec95}). If the function \(\eta\) is linear and \(F\) satisfies the Axioms PS1–PS6$(\eta )$, then \(F\) is single-valued for any game \(v\) and therefore \(F\equiv\boldsymbol{\phi}\), where \(\boldsymbol{\phi}\) is the Shapley value. \(\sharp\)
Let the number \(q_{i}(v)\) be defined for every game \(v\) by
\[q_{i}(v)=\sum_{\{S:i\in S\neq N^{v}\}}(s-1)!(n-s-1)!\eta (v(S)),\]
where \(s=|S|\) and \(n=|N^{v}|\).
\begin{equation}{\label{pec95t1}}\tag{1}\mbox{}\end{equation}
Theorem \ref{pec95t1}. {(Pechersky and Sobolev \cite[p.62]{pec95}). Let \(\eta\) be a continuous nondecreasing function and
\[m=\inf_{t\in\mathbb{R}}\eta (t)<\eta (0)<\sup_{t\in\mathbb{R}}\eta (t)=M\]
(the numbers \(m\) and \(M\) can equal \(\infty\)). If a solution \(F\) satisfies Axioms PS1–PS6$(\eta )$ and \({\bf 0}\in F(v)\), then
\begin{equation}{\label{pec95eq1}}\tag{2}
q_{i}(v)=q_{j}(v)
\end{equation}
for all \(i,j\in N\). \(\sharp\)
The Axiom PS4 is not used in the proof of theorem. It follows, from (\ref{pec95eq1}) that if \({\bf 0}\in F(v)\) and \(\eta\) satisfies the
conditions of Theorem \ref{pec95t1}, then \(\boldsymbol{\phi}(\eta v)={\bf 0}\) (Pechersky and Sobolev \cite[p.65]{pec95}).
In the sequel, we shall define a family of solutions satisfying the proposed system of axioms. Let \(v\) be a game and \({\bf x}\in\mathbb{R}^{N}\). Let us take a convex function \(\zeta :\mathbb{R}\rightarrow\mathbb{R}\) and consider the functional
\[Q^{\zeta}({\bf x},v)=\sum_{\{S:S\neq N,\emptyset\}} (s-1)!(n-s-1)!\zeta (v(S)-{\bf x}(S)).\]
Let \(Y(v)\) be the set of preimputations in the game \(v\) given by
\[Y(v)=\{{\bf x}\in\mathbb{R}^{N}:{\bf x}(N)=v(N)\}.\]
Define now the map \(F^{\zeta}\) as follows
\[F^{\zeta}(v)=\arg\min\left\{Q^{\zeta}({\bf x},v):{\bf x}\in Y(v)\right\}.\]
(Pechersky and Sobolev \cite[p.65]{pec95}).
Proposition. (Pechersky and Sobolev \cite[p.65]{pec95}). The set \(F^{\zeta}(v)\) is nonempty and convex. If \(\zeta\) is not affine, then
the set \(F^{\zeta}(v)\) is compact. \(\sharp\)
It is clear that the map \(F^{\zeta}\) satisfies Axioms PS1, PS2, PS4, and PS5 (Pechersky and Sobolev \cite[p.66]{pec95}).
Proposition. (Pechersky and Sobolev \cite[p.66]{pec95}). If \(v_{1}(S)=v_{2}(N\setminus S)-v_{2}(N)\) for all \(S\subseteq N\), then \(F^{\zeta}(v_{1})=-F^{\zeta}(v_{2})\). \(\sharp\)
Proposition. (Pechersky and Sobolev \cite[p.66]{pec95}). If \({\bf x}\in F^{\zeta}(v)\), then
\begin{equation}{\label{pec95eq4}}\tag{3}
\sum_{\{S:i,j\not\in S\}}s!(n-s-2)!\left (\zeta^{\prime}_{+}(V(S\cup\{i\})-{\bf x}(S\cup\{i\}))-\zeta^{\prime}_{-}(v(S\cup\{j\})-{\bf x}(S\cup\{j\}))\right )\geq 0
\end{equation}
for all \(i,j\in N\). \(\sharp\)
If \(\zeta\) is continuously differentiable, then the inequality (\ref{pec95eq4}) transforms to the equality
\[\sum_{\{S:i,j\not\in S\}}s!(n-s-2)!\left (\zeta^{\prime}(V(S\cup\{i\})-
{\bf x}(S\cup\{i\}))-\zeta^{\prime}(v(S\cup\{j\})-{\bf x}(S\cup\{j\}))\right )=0.\]
This quality and inclusion \({\bf x}\in Y(v)\) are sufficient conditions for \({\bf x}\in F^{\zeta}(v)\) (Pechersky and Sobolev \cite[p.67]{pec95}).
\begin{equation}{\label{pec95p8}}\tag{4}\mbox{}\end{equation}
Proposition \ref{pec95p8}. (Pechersky and Sobolev \cite[p.67]{pec95}). If \(\zeta\) is continuously differentiable, then \(F^{\zeta}\) satisfies Axiom PS6$(\zeta^{\prime})$. \(\sharp\)
Let \(\mbox{epi }\zeta\) denote the epigraph of function \(\Psi\), that is, \(\mbox{epi }\zeta =\{(t,y)\in\mathbb{R}^{2}:y\geq\zeta (t)\}\).
Proposition. (Pechersky and Sobolev \cite[p.67]{pec95}). The map \(F^{\zeta}\) satisfies Axiom PS3 if and only if the point \((0,\zeta (0))\) is extremal in \(\mbox{epi }\zeta\). \(\sharp\)
Proposition. (Pechersky and Sobolev \cite[p.69]{pec95}). Let \(\zeta\) be a convex, continuously differentiable function and \({\bf x},{\bf y}\in F^{\zeta}(v)\). Then for every \(S\), \(\zeta^{\prime}(v(S)-{\bf x}(S))=\zeta^{\prime}(v(S)-y(S))\). \(\sharp\)
\begin{equation}{\label{pec95t2}}\tag{5}\mbox{}\end{equation}
Theorem \ref{pec95t2}. {\em (Pechersky and Sobolev \cite[p.69]{pec95}). Let \(\eta\) be a continuous non-decreasing function and the point \(0\) be its growth point \((\)i.e. in every neighborhood of \(0\) there is a point \(t\) such that \(\eta (t)\neq\eta (0))\). Let \(\zeta\) be some primitive of \(\eta\) \((\)i.e. \(\zeta^{\prime}=\eta )\). Then the map \(F^{\zeta}\) satisfies Axioms PS1–PS6$(\phi )$. If, moreover, the function \(\eta\) satisfies the inequality \(\inf_{t}\eta (t)<\eta (0)<\sup_{t}\eta (t)\), then \(F^{\zeta}\) is the unique map satisfying the system of axioms. \(\sharp\)
If the function \(\eta\) is strictly increasing, then \(F^{\zeta}(v)\) is a singleton (Pechersky and Sobolev \cite[p.70]{pec95}).
Example. (Pechersky and Sobolev \cite[p.70]{pec95}) Let \(\epsilon\in\mathbb{R}\) and
\[\eta (t)=\left\{\begin{array}{ll}
0 & t\leq\epsilon\\ t-\epsilon & t>\epsilon\end{array}\right .\]
Consider the map \(F^{\zeta}\), where \(\zeta^{\prime}=\eta\). It is clear that if the \(\epsilon\)-core \(C_{\epsilon}(v)\) of game \(v\) is nonempty, then \(F^{\zeta}(v)=C_{\epsilon}(v)\). On the other hand, it can be shown that if \(v(N)\) is small enough with respect to other \(v(S)\), then \(F^{\zeta}(v)\) consist of the unique point which coincides with the Shapley value of the game \(v\). By Theorem \ref{pec95t2} and Proposition \ref{pec95p8} \(F^{\zeta}\) satisfies the proposed system of axioms only if \(\epsilon\leq 0\). Moreover if \(\epsilon <0\), then \(F^{\zeta}\) is the unique map satisfying the system. \(\sharp\)
We now consider the following question: Is there a selector of the presented solution? This studies the possibility of constructing the mapping \(f\) such that \(f(v)\in F(v)\) with \(|f(v)|=1\). Surely it would be interesting to find selectors satisfying the Axioms PS1-PS6($\phi$), but it is impossible since \(F^{\zeta}\) is the unique map satisfying the axioms. Therefore we will claim selectors to satisfy the following analogues of PS6($\phi$).
- Axiom PS6′($\phi$). There is such an opertaion \(\oplus\) on the set of games, that \(v_{1}\oplus v_{2}\in v_{1}\oplus_{\phi}v_{2}\) and \(f(v_{1})=f(v_{2})={\bf 0}\) implies \(f(v_{1}\oplus v_{2})={\bf 0}\).
Since \(f\) is single-valued, the Axioms PS1-PS5 can be simplified. For example, PS2 follows in this case from PS1 and PS4 (Pechersky and Sobolev \cite[p.71]{pec95}).
\begin{equation}{\label{pec95t3}}\tag{6}\mbox{}\end{equation}
Theorem \ref{pec95t3}. (Pechersky and Sobolev \cite[p.71]{pec95}). Let \(\eta\) be a continuous nondecreasing function on \(\mathbb{R}\) for which \({\bf 0}\) is a growth point. Then the map \(F^{\zeta}\), where \(\zeta^{\prime}=\eta\), admits a selector satisfying Axioms PS1–PS6’$(\eta )$. \(\sharp\)
We now present the set of axioms defining the prekernel. This set of axioms is in some sense analogous to the axiom system PS1–PS6($\phi$). We replace only PS2 and PS6($\phi$) by the following axioms
- Axiom PS2*. If \(v(S\cup\{i\})=v(S\cup\{j\})\) for every \(S\) with \(i,j\not\in S\) and \({\bf x}\in F(v)\), then \(x_{i}=x_{j}\).
- Axiom PS6*. If \({\bf 0}\in F(v_{1})\cap F(v_{2})\) and \(v_{3}(S)=\max\{v_{1}(S),v_{2}(S)\}\), then \({\bf 0}\in F(v_{3})\).
Let us introduce the following axiom
- Axiom PS7. Let \(a\in\mathbb{R}\), \(v_{1}(N)=v_{2}(N)\) and \(v_{1}(S)=v_{2}(S)+a\) for \(S\neq N,\emptyset\). Then \(F(v_{1})=F(v_{2})\).
The prekernel}$K(v)$ of the game \(v\) is given by
\[K(v)=\left\{{\bf x}\in\mathbb{R}^{N}:\max_{\{S:i\in S,j\not\in S\}}(v(S)-{\bf x}(S))=\max_{\{S:j\in S,i\not\in S\}}(v(S)-{\bf x}(S))\mbox{ for }i,j\in N, {\bf x}(N)=v(N)\right\}.\]
Proposition. (Pechersky and Sobolev \cite[p.74]{pec95}). The map \(K\) satisfies Axioms PS1, PS2*, PS3-PS5, PS6* and PS7. \(\sharp\)
Theorem. (Pechersky and Sobolev \cite[p.75]{pec95}). Let \(F\) satisfies Axioms PS1, PS2*, PS3-PS5, PS6* and PS7. Then \(F(v)\subseteq K(v)\) for any game \(v\). \(\sharp\)
The above theorem shows that the map \(K\) is the maximal solution satisfying the proposed system of axioms. Unfortunately, it is not unique. For example, the prenucleolus also satisfies this axiom’s system (Pechersky and Sobolev \cite[p.77]{pec95}).


