papers/lux-tokenomics/sections/05-game-theoretic-analysis.tex

\section{Game-Theoretic Analysis}
\label{sec:game}

\subsection{Validator Participation Game}

We model validator participation as a staking game $\Gamma = (\mathcal{N}, \mathcal{S}, \mathcal{U})$ where:
\begin{itemize}
    \item $\mathcal{N} = \{1, \ldots, n\}$ is the set of potential validators
    \item $\mathcal{S}_i = \{0, s_i^{\min}, \ldots, s_i^{\max}\}$ is the strategy space (stake amount)
    \item $\mathcal{U}_i(s_i, s_{-i})$ is the utility function for validator $i$
\end{itemize}

The utility for validator $i$ is:
\begin{equation}
    \mathcal{U}_i(s_i, s_{-i}) = \underbrace{r_{i,t}}_{\text{staking reward}} + \underbrace{p_i \cdot \text{tip}_t}_{\text{priority fees}} - \underbrace{c_i^{\text{hw}}}_{\text{hardware cost}} - \underbrace{s_i \cdot r_{\text{opp}}}_{\text{opportunity cost}}
\end{equation}
where $p_i$ is the probability of being selected as block proposer and $r_{\text{opp}}$ is the alternative rate of return.

\begin{theorem}[Nash Equilibrium Existence]
The staking game $\Gamma$ has a unique Nash equilibrium in pure strategies where each rational validator stakes $s_i^* = s_i^{\max}$ if $\mathcal{U}_i(s_i^{\max}, s_{-i}^*) > 0$ and $s_i^* = 0$ otherwise.
\end{theorem}

\begin{proof}
The utility function $\mathcal{U}_i$ is concave in $s_i$ because the reward function $r_{i,t}$ is concave in $s_i$ (proportional allocation with diminishing marginal returns) while costs are linear. For each validator, the best response is unique: stake the maximum amount if net utility is positive, zero otherwise. The threshold for participation is:
\begin{equation}
    s_i^{\min} \cdot \left(\frac{R_t \cdot u_i}{W} - r_{\text{opp}}\right) \geq c_i^{\text{hw}}
\end{equation}
This defines a partition of $\mathcal{N}$ into active and inactive validators that constitutes the unique Nash equilibrium.
\end{proof}

\subsection{Slashing as Deterrence}

Validators who deviate from the protocol face slashing penalties. We define three severity levels:
\begin{table}[H]
\centering
\caption{Slashing conditions and penalty structure.}
\label{tab:slashing}
\begin{tabular}{llrl}
\toprule
\textbf{Violation} & \textbf{Severity} & \textbf{Penalty (\%)} & \textbf{Cooldown} \\
\midrule
Extended downtime ($> 48$h) & Minor & 0.1 & 7 days \\
Double signing (equivocation) & Major & 5.0 & 30 days \\
Coordinated attack ($\geq 3$ validators) & Critical & 33.0 & Permanent \\
\bottomrule
\end{tabular}
\end{table}

\begin{lemma}[Slashing Incentive Compatibility]
Under the slashing penalties in Table~\ref{tab:slashing}, a rational validator's expected utility from honest behavior exceeds the expected utility from any Byzantine strategy, provided the validator controls less than $1/3$ of total stake.
\end{lemma}

\begin{proof}
Let $\pi_{\text{attack}}$ be the maximum profit from a successful Byzantine attack and $p_{\text{detect}}$ the detection probability. The expected utility from attacking is $\mathcal{U}_{\text{attack}} = \pi_{\text{attack}} \cdot (1 - p_{\text{detect}}) - \text{slash} \cdot s_i \cdot p_{\text{detect}}$. For double signing, $p_{\text{detect}} = 1$ (cryptographically provable), so $\mathcal{U}_{\text{attack}} = -0.05 \cdot s_i < 0$. For coordinated attacks requiring $> 1/3$ stake, the penalty of 33\% exceeds any extractable value under our fee structure.
\end{proof}

\subsection{Sybil Resistance}

The minimum stake requirement $s^{\min} = 2{,}000$ LUX and the proportional reward structure ensure that splitting stake across multiple validator identities provides no advantage:

\begin{proposition}[Sybil Neutrality]
For any validator with stake $s$, the total reward from running one validator with stake $s$ equals the total reward from running $k$ validators each with stake $s/k$, minus $k-1$ additional hardware costs.
\end{proposition}