Skip to main content
Log in

Continuity of equilibria for two-person zero-sum games with noncompact action sets and unbounded payoffs

  • Feinberg: Probability
  • Published:
Annals of Operations Research Aims and scope Submit manuscript

Abstract

This paper extends Berge’s maximum theorem for possibly noncompact action sets and unbounded cost functions to minimax problems and studies applications of these extensions to two-player zero-sum games with possibly noncompact action sets and unbounded payoffs. For games with perfect information, also known under the name of turn-based games, this paper establishes continuity properties of value functions and solution multifunctions. For games with simultaneous moves, it provides results on the existence of lopsided values (the values in the asymmetric form) and solutions. This paper also establishes continuity properties of the lopsided values and solution multifunctions.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

References

  • Aubin, J.-P. (1998). Optima and equilibria: An introduction to nonlinear analysis. Berlin: Springer.

    Google Scholar 

  • Aubin, J.-P., & Ekeland, I. (1984). Applied nonlinear analysis. New York: Wiley.

    Google Scholar 

  • Berge, C. (1963). Topological spaces. New York: Macmillan.

    Google Scholar 

  • Fan, K. (1963). Minimax theorems. Proceedings of the National Academy of Sciences of the United States of America, 39(1), 42–47.

    Article  Google Scholar 

  • Feinberg, E. A. (2016). Optimality conditions for inventory control. In A. Gupta & A. Capponi (Eds.), Tutorials in operations research. Optimization challenges incomplex, networked, and risky systems (pp. 14–44). Cantonsville, MD: INFORMS.

    Google Scholar 

  • Feinberg, E. A., & Kasyanov, P. O. (2015). Continuity of minima: Local results. Set-Valued and Variational Analysis, 23(3), 485–499.

    Article  Google Scholar 

  • Feinberg, E. A., Kasyanov, P. O., & Voorneveld, M. (2014). Berge’s maximum theorem for noncompact image sets. Journal of Mathematical Analysis and Applications, 413(2), 1040–1046.

    Article  Google Scholar 

  • Feinberg, E. A., Kasyanov, P. O., & Zadoianchuk, N. V. (2012). Average cost Markov decision processes with weakly continuous transition probabilities. Mathematics of Operations Research, 37(4), 591–607.

    Article  Google Scholar 

  • Feinberg, E. A., Kasyanov, P. O., & Zadoianchuk, N. V. (2013). Berge’s theorem for noncompact image sets. Journal of Mathematical Analysis and Applications, 397(1), 255–259.

    Article  Google Scholar 

  • Feinberg, E. A., Kasyanov, P. O., & Zgurovsky, M. Z. (2016). Partially observable total-cost Markov decision processes with weakly continuous transition probabilities. Mathematics of Operations Research, 41(1), 656–681.

    Article  Google Scholar 

  • Feinberg, E. A., & Lewis, M. E. (2017). On the convergence of optimal actions for Markov decision processes and the optimality of \((s, S)\) policies for inventory control. Naval Research Logistics,. doi:10.1002/nav.21750.

    Article  Google Scholar 

  • Hu, Sh., & Papageorgiou, N. S. (1997). Handbook of multivalued analysis. Volume I: Theory. Dordrecht: Kluwer.

  • Jaśhkewicz, A., & Nowak, A. (2011). Stochastic games with unbounded payoffs: Applications to robust control in economics. Dynamic Games and Applications, 1(2), 253–279.

    Article  Google Scholar 

  • Jaśhkewicz, A., & Nowak, A. (2017). Zero-sum stochastic games. In T. Basa & G. Zaccour (Eds.), Handbook of dynamic game theory. Cham: Springer. doi:10.1007/978-3-319-27335-8_8-1.

    Chapter  Google Scholar 

  • Katehakis, M. N., Melamed, B., & Shi, J. (2016). Optimal replenishment rate for inventory systems with compound Poisson demands and lost-sales: A direct treatment of time average cost. Annals of Operations Research. doi:10.1007/s10479-015-1998-y.

  • Mertens, J.-F., Sorin, S., & Zamir, S. (2015). Repeated games. New York: Cambridge University Press.

    Book  Google Scholar 

  • Parthasarathy, K. R. (1967). Probability measures on metric spaces. New York: Academic Press.

    Book  Google Scholar 

  • Perchet, V., & Vigeral, G. (2015). A minmax theorem for concave–convex mappings with no regularity assumptions. Journal of Convex Analysis, 22(2), 537–540.

    Google Scholar 

  • Reed, M., & Simon, B. (1980). Methods of modern mathematical physics, I: Functional analysis. New York: Academic Press.

    Google Scholar 

  • Shi, J., Katehakis, M. N., & Melamed, B. (2013). Martingale methods for pricing inventory penalties under continuous replenishment and compound renewal demands. Annals of Operations Research, 208(1), 593–612.

    Article  Google Scholar 

  • Yanovskaya, E. B. (1974). Infinite zero-sum two-person games. Journal of Soviet Mathematics, 3(5), 520–541.

    Article  Google Scholar 

  • Zgurovsky, M. Z., Mel’nik, V. S., & Kasyanov, P. O. (2011). Evolution inclusions and variation inequalities for earth data processing I. Berlin: Springer.

    Book  Google Scholar 

Download references

Acknowledgements

The authors thank William D. Sudderth for his valuable comments on von Neumann’s and Sion’s minimax theorems. The authors thank referees for their insightful suggestions.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Eugene A. Feinberg.

Additional information

Research of the first author was partially supported by NSF Grant CMMI-1636193.

Appendix: Properties of \(\texttt {A}\)-lower semi-continuous multifunctions

Appendix: Properties of \(\texttt {A}\)-lower semi-continuous multifunctions

This appendix describes some properties of \(\texttt {A}\)-lower semi-continuous multifunctions. Definition 4 and the definition of lower semi-continuous multifunctions imply that an \(\texttt {A}\)-lower semi-continuous multifunction is lower semi-continuous. The following example demonstrates that a lower semi-continuous multifunction may not be \(\texttt {A}\)-lower semi-continuous.

Example 5

Let \(\texttt {X}=\texttt {A}=[0,1],\) \(\texttt {B}=\mathbb {R},\) \(\mathtt {\Phi }_\texttt {A}(x)=\{x\}\cup \{\frac{1}{x}\}\) for \(x\in (0,1],\) \(\mathtt {\Phi }_\texttt {A}(0)=\{0\},\) and \(\mathtt {\Phi }_\texttt {B}(x,a)=\{a\}\) for all \((x,a)\in \mathrm{Gr}(\mathtt {\Phi }_\texttt {A}).\) Since each set \(\mathtt {\Phi }_\texttt {B}(x,a)\) is a singleton, where \((x,a)\in \mathrm{Gr}(\mathtt {\Phi }_\texttt {A}),\) and the graph of the multifunction \(\mathtt {\Phi }_\texttt {B}\) is closed, the multifunction \(\mathtt {\Phi }_\texttt {B}\) is lower semi-continuous. Let us consider the sequence \(\{x_n\}_{n=1,2,\ldots }=\{\frac{1}{n}\}_{n=1,2,\ldots }\) converging to \(x=0.\) Then \(b:=0\in \mathtt {\Phi }_\texttt {B}(0,0)\) and \(a^{(n)}=n\in \mathtt {\Phi }_\texttt {A}(x^{(n)}),\) \(n=1,2,\ldots \ .\) However, the sequence \(\{b_n\}_{n=1,2,\ldots }:=\{n\}_{n=1,2,\ldots }\) does not have a limit point. Thus, the multifunction \(\mathtt {\Phi }_\texttt {B}\) is not \(\texttt {A}\)-lower semi-continuous.

Let us provide sufficient conditions for \(\texttt {A}\)-lower semi-continuity.

Lemma 7

Let \(\mathtt {\Phi }_{\texttt {B}}:\mathrm{Gr}(\mathtt {\Phi }_{\texttt {A}})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) be a lower semi-continuous set-valued mapping. Then the following statements hold:

  1. (a)

    if \(\mathtt {\Phi }_{\texttt {A}}:\texttt {X}\mapsto S(\texttt {A})\) is upper semi-continuous and compact-valued at each \(x\in \texttt {X},\) then \(\mathtt {\Phi }_{\texttt {B}}:\mathrm{Gr}(\mathtt {\Phi }_{\texttt {A}})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is \(\texttt {A}\)-lower semi-continuous;

  2. (b)

    if \(\mathtt {\Phi }_{\texttt {B}}(x,a)\) does not depend on \(a\in \mathtt {\Phi }_\texttt {A}(x)\) for each \(x\in \texttt {X},\) that is, \(\mathtt {\Phi }_{\texttt {B}}(x,a_*)=\mathtt {\Phi }_{\texttt {B}}(x,a^*)\) for each \((x,a_*),(x,a^*)\in \mathrm{Gr}(\mathtt {\Phi }_\texttt {A}),\) then \(\mathtt {\Phi }_{\texttt {B}}:\mathrm{Gr}(\mathtt {\Phi }_{\texttt {A}})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is \(\texttt {A}\)-lower semi-continuous.

Remark 26

Let \(\mathtt {\Phi }:\texttt {X}\mapsto S(\texttt {B}),\) where \(\mathtt {\Phi }(x)\) can be interpreted as the set of actions for Player II, when this set does not depend on the actions of Player I, as this takes place for games with simultaneous moves. Then we can define the sets

$$\begin{aligned} \mathtt {\Phi }_\texttt {B}(x,a):= \mathtt {\Phi }(x),\qquad \qquad (x,a)\in \mathrm{Gr}(\mathtt {\Phi }_\texttt {A}). \end{aligned}$$
(52)

The definition of a lower semi-continuous multifunction implies that, if the multifunction \(\mathtt {\Phi }:\texttt {X}\mapsto S(\texttt {B})\) is lower semi-continuous, then the multifunction \(\mathtt {\Phi }_{\texttt {B}}:\mathrm{Gr}(\mathtt {\Phi }_{\texttt {A}})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is lower semi-continuous too. Lemma 7 implies that the lower semi-continuity of \(\mathtt {\Phi }_{\texttt {B}}:\mathrm{Gr}(\mathtt {\Phi }_{\texttt {A}})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is equivalent to its \(\texttt {A}\)-lower semi-continuity in the following two cases: (a) for two-person zero-sum games with perfect information, when the decision sets \(\{\mathtt {\Phi }_{\texttt {A}}(x)\}_{x\in \texttt {X}}\) for the first player are compact and the dependence of \(\mathtt {\Phi }_{\texttt {A}}(x)\) by the state variable x is upper semi-continuous, and (b) for two-person zero-sum games with simultaneous moves.

Proof of Lemma 7

(a) Let \(\{x^{(n)}\}_{n=1,2,\ldots }\) be a sequence with values in \(\texttt {X}\) that converges and its limit x belongs to \(\texttt {X}.\) Let also \(a^{(n)}\in \mathtt {\Phi }_\texttt {A}(x^{(n)}),\) for each \(n=1,2,\ldots ,\) and \(b\in \mathtt {\Phi }_\texttt {B}(x,a)\) for some \(a\in \mathtt {\Phi }_\texttt {A}(x).\) Let us prove that b is a limit point for a sequence \(\{b^{(n)}\}_{n=1,2,\ldots }\) with \(b^{(n)}\in \mathtt {\Phi }_\texttt {B}(x^{(n)},a^{(n)})\) for each \(n=1,2,\ldots \ .\) Indeed, Lemma 4, being applied to \(\mathbb {X}:=\texttt {X},\) \(\mathbb {Y}:=\texttt {A},\) and \(\varPhi :=\mathtt {\Phi }_\texttt {A},\) implies that the sequence \(\{a^{(n)}\}_{n=1,2,\ldots }\) has a limit point \(a\in \mathtt {\Phi }_\texttt {A}(x).\) Therefore, b is a limit point of a sequence \(\{b^{(n)}\}_{n=1,2,\ldots }\) with \(b^{(n)}\in \mathtt {\Phi }_\texttt {B}(x^{(n)},a^{(n)})\) for each \(n=1,2,\ldots ,\) since \(\mathtt {\Phi }_{\texttt {B}}:\mathrm{Gr}(\mathtt {\Phi }_{\texttt {A}})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is a lower semi-continuous set-valued mapping.

(b) Since \(\mathtt {\Phi }_{\texttt {B}}:\mathrm{Gr}(\mathtt {\Phi }_{\texttt {A}})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is a lower semi-continuous set-valued mapping and \(\mathtt {\Phi }(x,a)\) does not depend on \(a\in \mathtt {\Phi }_\texttt {A}(x)\) for each \(x\in \texttt {X},\) the following statement holds: if a sequence \(\{x^{(n)}\}_{n=1,2,\ldots }\) with values in \(\texttt {X}\) converges and its limit x belongs to \(\texttt {X},\) \(a^{(n)}\in \mathtt {\Phi }_\texttt {A}(x^{(n)})\) for each \(n=1,2,\ldots ,\) and \(b\in \mathtt {\Phi }_\texttt {B}(x,a)\) for some \(a\in \mathtt {\Phi }_\texttt {A}(x),\) then b is a limit point of a sequence \(\{b^{(n)}\}_{n=1,2,\ldots }\) with \(b^{(n)}\in \mathtt {\Phi }_\texttt {B}(x^{(n)},a^{(n)})\) for each \(n=1,2,\ldots ,\) that is, \(\mathtt {\Phi }_{\texttt {B}}:\mathrm{Gr}(\mathtt {\Phi }_{\texttt {A}})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is \(\texttt {A}\)-lower semi-continuous set-valued mapping. \(\square \)

The following two statements, which are not used in this paper, provide additional properties of \(\texttt {A}\)-lower semi-continuous set-valued mappings for the case, when \(\texttt {B}\) is a vector space. Let \(\texttt {B}\) be a vector space and \(\mathtt {\Phi }_\texttt {B},\mathtt {\Psi }_\texttt {B}:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) be set-valued mappings. Let us define for each \((x,a)\in \mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\)

$$\begin{aligned} \mathtt {\Phi }_\texttt {B}(x,a)+\mathtt {\Psi }_\texttt {B}(x,a):=\{b_0+b_1\,:\,b_1\in \mathtt {\Phi }_\texttt {B}(x,a),\, b_2\in \mathtt {\Psi }_\texttt {B}(x,a) \}. \end{aligned}$$

Lemma 8

Let \(\texttt {B}\) be a vector space and \(\mathtt {\Phi }_\texttt {B},\mathtt {\Psi }_\texttt {B}:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) be \(\texttt {A}\)-lower semi-continuous set-valued mappings. Then the set-valued mapping \( \mathtt {\Phi }_\texttt {B}+\mathtt {\Psi }_\texttt {B}:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is \(\texttt {A}\)-lower semi-continuous.

Proof of Lemma 8

Let \(\{x^{(n)}\}_{n=1,2,\ldots }\) be a sequence with values in \(\texttt {X}\) that converges and its limit x belongs to \(\texttt {X}.\) Assume that \(a^{(n)}\in \mathtt {\Phi }_\texttt {A}(x^{(n)})\) for each \(n=1,2,\ldots ,\) and \(b\in \mathtt {\Phi }_\texttt {B}(x,a)\) for some \(a\in \mathtt {\Phi }_\texttt {A}(x).\) Let us prove that b is a limit point of a sequence \(\{b^{(n)}\}_{n=1,2,\ldots }\) with \(b^{(n)}\in \mathtt {\Phi }_\texttt {B}(x^{(n)},a^{(n)}),\) \(n=1,2,\ldots .\) Indeed, since \(\mathtt {\Phi }_\texttt {B}(x,a)=\mathtt {\Phi }_\texttt {B}(x,a)+\mathtt {\Psi }_\texttt {B}(x,a),\) there exist \(b_1\in \mathtt {\Phi }_\texttt {B}(x,a)\) and \(b_2\in \mathtt {\Psi }_\texttt {B}(x,a)\) such that \(b=b_0+b_1.\) The \(\texttt {A}\)-lower semi-continuity of \(\mathtt {\Phi }_\texttt {B}:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) and \(\mathtt {\Psi }_\texttt {B}:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) imply that \(b_i,\) \(i=0,1,\) is a limit point of a sequence \(\{b_i^{(n)}\}_{n=1,2,\ldots }\) with \(b_1^{(n)}\in \mathtt {\Phi }_\texttt {B}(x^{(n)},a^{(n)})\) and \(b_2^{(n)}\in \mathtt {\Psi }_\texttt {B}(x^{(n)},a^{(n)})\) Therefore, \(b=b_0+b_1\) is a limit point of a sequence \(\{b^{(n)}\}_{n=1,2,\ldots }\) with \(b^{(n)}:=b_0^{(n)}+b_1^{(n)}\in \mathtt {\Phi }_\texttt {B}(x^{(n)},a^{(n)})+\mathtt {\Psi }_\texttt {B}(x^{(n)},a^{(n)}),\) \(n=1,2,\ldots \ .\) Thus, the set-valued mapping \(\mathtt {\Phi }_\texttt {B}^0+\mathtt {\Phi }_\texttt {B}^1:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is \(\texttt {A}\)-lower semi-continuous. \(\square \)

Corollary 8

Let \(\texttt {B}\) be a vector space, \(\mathtt {\Phi }:\texttt {X}\mapsto S(\texttt {B})\) be a lower semi-continuous set-valued mapping, \(\mathtt {\Phi }_{\texttt {A}}:\texttt {X}\mapsto \mathbb {K}(\texttt {A})\) be an upper semi-continuous set-valued mapping, and \(\mathtt {\Psi }_\texttt {B}:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) be a lower semi-continuous set-valued mapping. Let us consider the set-valued mapping \(\mathtt {\Phi }_\texttt {B}:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) defined in (52). Then the set-valued mapping \(\mathtt {\Phi }_\texttt {B}+\mathtt {\Psi }_\texttt {B}:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) is \(\texttt {A}\)-lower semi-continuous.

Proof

According to Lemma 7, the set-valued mappings \(\mathtt {\Phi }_\texttt {B}, \mathtt {\Psi }_\texttt {B}:\mathrm{Gr}(\mathtt {\Phi }_\texttt {A})\subset \texttt {X}\times \texttt {A}\mapsto S(\texttt {B})\) are \(\texttt {A}\)-lower semi-continuous. Therefore, Lemma 8 implies that their sum is \(\texttt {A}\)-lower semi-continuous. \(\square \)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Feinberg, E.A., Kasyanov, P.O. & Zgurovsky, M.Z. Continuity of equilibria for two-person zero-sum games with noncompact action sets and unbounded payoffs. Ann Oper Res 317, 537–568 (2022). https://doi.org/10.1007/s10479-017-2677-y

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10479-017-2677-y

Keywords

Mathematics Subject Classification

Navigation