Bounded Budget Connection (BBC) Games

• BBC Games are network formation models where agents establish connections under fixed budget constraints, influencing both individual decisions and overall network structure.
• They involve rigorous analysis of Nash equilibria, NP-hard computational challenges, and diverse bidding mechanisms including Richman, Poorman, and Robin Hood approaches.
• Applications span decentralized network design, online resource allocation, and auction-based game theory, providing insights into connectivity, efficiency, and strategic behavior.

Bounded Budget Connection (BBC) Games are a class of network formation and control games in which agents (nodes or players) form connections subject to explicit budget constraints, with individual and global objectives determined by the resulting network structure. BBC models are core to understanding decentralized network formation in economics, computer science, and combinatorial optimization, especially under adversarial, resource-sharing, and auction-like dynamics.

1. Formal Definitions and Core Model Structure

The archetypal BBC game, introduced in (0806.1727), is specified by the tuple $(V, w, c, l, b)$, where $V$ is the set of $n$ nodes (players), $w(u,v)$ is a preference weight quantifying how much $u$ wants to be close to $v$, $c(u,v)$ is the cost for $u$ to build a link to $v$, $l(u,v)$ is the traversal length, and $b(u)$ is $u$'s budget to expend on links. Each node $u$ chooses its strategy $S_u \subseteq \{(u, v)\}$ with $\sum_{(u,v)\in S_u} c(u,v)\leq b(u)$. The global network is $G(S)=(V,E)$ with $E=\cup_u S_u$.

Objective functions typically take the form:

• Sum-objective: $\mathrm{cost}_u(S)=\sum_{v\in V} w(u, v)\, d(u,v)$, minimizing total (possibly weighted) shortest-path distances.
• Max-objective (BBC-Max): $\mathrm{cost}_u(S)=\max_{v\in V} w(u,v)\, d(u,v)$, minimizing maximum weighted distance.

A pure Nash equilibrium (PNE), or stable graph, is a strategy profile where no single node can decrease its cost by unilateral deviation.

2. Key Results: Existence, Complexity, and Structure of Equilibria

Existence and NP-hardness

In general (non-uniform) BBC games, establishing whether a PNE exists is NP-hard—even if only link costs, lengths, or preference weights are non-uniform (0806.1727). Counterexamples show that cycles of improving responses with no stable point are possible. Fractional BBC games (relaxing discrete links to allow fractional capacities) admit PNE via convexity and fixed-point arguments.

Variant	PNE Existence	Complexity
Non-uniform	Not always	NP-hard
Fractional BBC	Always	Poly-time
Uniform (n,k)	Always	Poly-time

Uniform BBC Games

When all nodes have equal budgets and all costs/lengths are set to $1$ ($n,k$-uniform), equilibrium existence and fairness are guaranteed. Stable graphs (costs within a multiplicative and additive constant for any pair of nodes) always exist. Explicit constructions ("Forest of Willows") span the space from minimum to maximum social cost. Regular graphs (identical "buying patterns") are generally not stable unless $k$ is linear in $n$.

Diameter and Price of Anarchy/Stability

BBC equilibria display sharp differences depending on budget and objective:

• Tree budgets ($\sum b_i = n-1$) yield spanning trees: diameter $\Theta(n)$ (MAX), $\Theta(\log n)$ (SUM).
• All budgets $1$ ensure constant diameter.
• Larger budgets can increase equilibrium diameter in MAX versions: $\Omega(\sqrt{\log n})$ ("Braess's paradox" for BBC games) (Ehsani et al., 2011).
• SUM equilibrium graphs have $2^{O(\sqrt{\log n})}$ diameter at worst.
• For SUM versions with per-node budget $k$, equilibria are either $k$-connected or diameter $<4$.
• Price of stability is $O(1)$; price of anarchy: $2^{\Omega(n/k)}$ (BBC), exponential bounds in various budget/distance-constrained settings.

Instance	MAX Diameter	SUM Diameter
Tree ($\sum b_i=n-1$)	$\Theta(n)$	$\Theta(\log n)$
All budgets $1$	$O(1)$	$O(1)$
All budgets $>1$	$\Omega(\sqrt{\log n})$	$2^{O(\sqrt{\log n})}$

3. Directed Bidding and Auction-Based Connection Games

BBC games have been generalized to bidding games in reachability and infinite-duration settings (Avni et al., 2017, Avni et al., 2023, Avni et al., 2018, Almagor et al., 23 Dec 2024).

Richman Bidding

Players bid for control of a token in a graph; the winner pays the bid to the loser and moves the token. The central property is the existence of a threshold budget $Th(v)$ at every vertex, which determines the minimal initial allocation required to enforce reachability/parity/mean-payoff objectives: $Th(v) = \frac{1}{2}(Th(v^+) + Th(v^-))$ where $v^+, v^-$ are neighbors with maximum/minimum thresholds. These thresholds correspond to win probabilities in random-turn simple stochastic games.

Poorman Bidding (First-Price)

Bids are paid to the bank, not opponents; losing player retains budget. Threshold budgets exist and are computable for reachability objectives (periodic for quantized budgets, closed formulas for special graphs). For infinite-duration quantitative objectives (mean-payoff), the guaranteed value for a player depends on initial budget ratios and connects exactly to biased random-turn games, i.e., the mean-payoff corresponds to a process where player $i$ acts with probability $r_i$.

Robin Hood Bidding

Introduces linear redistribution: before each bidding step, the richer player gives a fraction $\lambda$ of the difference to the poorer (“wealth regulation”). The threshold function persists, but threshold determinacy fails—games may be undetermined at threshold. Threshold computation reduces to a Mixed-Integer Linear Program (MILP), accommodating the non-uniqueness of solutions and complicated average property: $$Th(v) = \max\left( \min\left( WR^{-1}\left( \frac{Th(v^+) + Th(v^-)}{2} \right), 1 \right), 0 \right)$$ where $WR^{-1}(x)=\frac{x-\lambda}{1-2\lambda}$.

4. Flow-Based and Fractional BBC Models

Fractional BBC games allow players to purchase fractions of links, forming capacitated flow networks (0812.0598, 0806.1727). Each node allocates weights on outgoing edges subject to cost constraints, and utility is determined by the cost of a minimum-cost unit flow to a destination. Equilibria correspond to personalized equilibria in associated matrix games: each player selects a distribution over strategies and arranges fractional matchings (flows) with opponents to maximize utility.

Computationally, while rational equilibria always exist, computing even approximate equilibria for fractional BBC is PPAD-hard. Two-player personalized equilibria admit polynomial-time solutions, but the general multi-player case is intractable.

5. Budget Games and Ordered Resource Allocation

Budget games (Drees et al., 2014) generalize BBC by introducing multi-resource constraints and task selection: each player chooses tasks with demands over shared resources with fixed budgets. If demand exceeds budget, proportional sharing applies. In ordered budget games, arrival order determines allocation, modeling settings with prioritization. Existence of Nash and strong equilibria is guaranteed in ordered games, but not in the strategic simultaneous case, and convergence to equilibrium may be exponential.

Social welfare optimization in budget games is NP-hard in general but admits constant-factor approximation in matroid settings. Price of anarchy is at most $2$ ($2 - \varepsilon$ is achievable).

6. Controlled Random Processes and Online Strategies

Online BBC games involve decision-making under streaming edge arrivals with immediate, irrevocable choices and global budget constraints (Frieze et al., 2022). Precise trade-offs between time (number of revealed edges) and budget (number of permitted purchases) have been established for key properties:

• $k$-vertex-connectivity can be achieved with $c_kn$ edges (explicit $c_k<k$, $c_k\sim k/2$)
• Hamiltonicity achievable at hitting time by purchasing at most $Cn$ edges ($C>1$ absolute; cannot take $C\to 1$)
• Perfect matching achievable at classical thresholds with minimal budget
• Spanning trees and small trees attainable under sharp budget/time thresholds ($b\gg(n/t)^{k-2}$)
• Construction of cycles of fixed length requires $b\gg\max\{n^{k+2}/t^{k+1}, n/\sqrt{t}\}$

This framework provides optimal and near-optimal benchmarks for BBC algorithms in practical online systems.

Property	Time Threshold	Budget Threshold	Remarks
Spanning tree	$\tau_C$	$n-1$	Forest growing
Min degree $k$	$\tau_k$	$o_k n$, $o_k\sim k/2$	k-NN graph, explicit constant
Hamiltonicity	$\tau_2$	$Cn$ ($C>1$)	Can't have $C\to 1$
Perfect matching	$\sim n \log n/2$	$\sim n/2$	Greedy + expansion
$k$-vertex tree	see text	$(n/t)^{k-2}$	Inductive construction

7. Bounded-Distance Network Games as BBC Variants

Bounded-Distance Network Creation Games (MaxBD, SumBD) (Bilò et al., 2011) invert BBC constraints: instead of limiting links by budget, players must maintain eccentricity or average distance below a bound $R$ or $D$, minimizing edge purchases. Nash equilibrium exists, but best response is NP-hard. Efficiency (price of anarchy) is unbounded for non-uniform distance bounds, but with uniform bounds PoA is sharply characterized:

• MaxBD: PoA $=O(n^{1/(\lfloor\log_3 R\rfloor+1)})$ for $R\ge3$, PoA $\Omega(\sqrt{n})$ for $R=2$
• SumBD: PoA $=O(n^{1/(\lfloor\log_3(D/4)\rfloor+2)})$ for $D\ge3$, $O(\sqrt{n\log n})$ for $2\le D<3$, PoA $=O(1)$ for $D=n^{\omega(1/\sqrt{\log n})}$

These models offer complementary insights to edge-budget BBC games, reflecting centrality and connectivity requirements.

---

Bounded Budget Connection Games form a central framework for the rigorous paper of decentralized network formation, strategic resource allocation, and auction-like control on graphs. The theory encompasses combinatorial and flow-based strategies, structural and economic properties (including paradoxes and fairness mechanisms), computational hardness and algorithmic advances (MILPs, dynamic programming, personalized equilibrium), and online/streaming protocols with precise optimality bounds. These results underline both the complexity and canon of possible behaviors in resource-constrained graph formation, informing the design and analysis of real-world social, peer-to-peer, and overlay networks, and strategic systems under budget and centrality constraints.