Limit Order Book Simulation

• Limit Order Book Simulation is a computational model representing electronic market dynamics through aggregated buy and sell orders with defined state spaces.
• It employs methodologies such as state-dependent point processes, deterministic clearing operators, and price-priority matching to enforce market rules.
• The framework provides practical insights into liquidity dynamics, order flow impacts, and regulatory implications for financial markets.

A limit order book (LOB) simulation is a computational approach for modeling the high-frequency dynamics of order placement, cancellation, and matching in electronic financial markets. The LOB aggregates the outstanding buy and sell limit orders at various price levels, representing the market’s microstructure. Simulation frameworks enable researchers to analyze liquidity dynamics, price formation, and the impact of different order flow patterns or trading rules. Approaches vary from fully probabilistic point-process models and fluid limits to agent-based and deep learning-driven methods, with applications in market mircostructure analysis, trading strategy evaluation, and regulatory design.

1. Mathematical Foundations of Limit Order Book Simulation

At the core of flexible LOB simulation frameworks is a decomposition of market activity into order flow (modeled as a multivariate spatial point process) and deterministic market clearing (implemented via a mass-transport operator or matching-rule-based operator). Let the state space be $E = \mathbb{N}^d \times \mathbb{N}^d$, where each $X = (X^+, X^-)$ specifies the number of buy ($X^+_i$) and sell ($X^-_i$) limit orders at price level $i = 1,\ldots,d$. The admissible set of LOB states, ensuring no order crossing, is

$\mathcal{L} = \{ X\in E : a(X) > b(X) \},$

with $a(X)$ and $b(X)$ denoting the ask (lowest index with $X^-_i > 0$) and bid (highest with $X^+_i > 0$), respectively.

Order flow is represented as a state-dependent marked point process. Four event types are permitted at each price $i$ and size $z>0$:

• Limit buy: $X \mapsto X + (z e_i, 0)$
• Limit sell: $X \mapsto X + (0, z e_i)$
• Cancel buy: $X \mapsto X - (z e_i, 0)$, valid if $z \le X^+_i$
• Cancel sell: $X \mapsto X - (0, z e_i)$, valid if $z \le X^-_i$

The infinitesimal generator for order flow alone, acting on test functions $f: E \to \mathbb{R}$, is

$$L_o f(X) = \sum_{i=1}^d \sum_{z=1}^\infty \Big[ \lambda_+(i, X, z) (f(X^+ + z e_i, X^-) - f(X)) + \lambda_-(i, X, z) (f(X^+, X^- + z e_i) - f(X)) + C_+(i, X, z) (f(X^+ - z e_i, X^-) - f(X)) + C_-(i, X, z) (f(X^+, X^- - z e_i) - f(X)) \Big],$$

where $\lambda_{\pm}$ and $C_{\pm}$ are state-dependent intensities for the four event types. Sufficient growth controls on intensities (e.g., power-law decay in $z$ with $\alpha > 1$) ensure the generator is well-defined (Cont et al., 2023).

After each event, a deterministic clearing operator $C: E \to \mathcal{L}$ enforces the no-crossing property via price-priority matching (maximum-volume, price-priority match-and-prune). The explicit calculation utilizes the imbalance function $g_X(k) = S_X(k) - B_X(k)$, where $S_X(k)$ and $B_X(k)$ are cumulative sell and buy quantities at/inside threshold $k$. The projection identifies new post-clearing bid/ask levels and truncates mass as necessary.

Combining these yields the full LOB generator acting on functions $f: \mathcal{L} \to \mathbb{R}$,

$$Lf(X) = \sum_{i, z} \lambda_+(i, X, z)[f(C(X^+ + z e_i, X^-)) - f(X)] + \ldots$$

where each term reflects an order flow