Abstract: This research develops a general stochastic approach to modeling delays in financial markets. Utilizing the framework of "Large Platonic Markets"—which accommodate infinitely many assets and restricted information settings—the paper investigates two primary types of latency: information delays and order execution delays. The study demonstrates that these delays can be modeled as random and inhomogeneous processes over time. The central finding is the "inheritance" of the No Asymptotic $L^p$-Free Lunch ($NAFL_p$) condition, proving that a market that is fundamentally free of arbitrage remains so even when complex delays are introduced. Finally, the paper provides a methodology for verifying the absence of arbitrage in environments with multiple brokers operating at varying trading speeds.
Introduction
In classical financial theory, markets are often treated as frictionless environments where information is instantaneous and execution is immediate. However, modern high-frequency trading (HFT) environments are defined by microseconds of latency. To mitigate the advantages of HFTs and promote market stability, some exchanges have implemented intentional "speed bumps"—artificial delays in order execution.
While these mechanisms are intuitive, their mathematical implications for market fairness and arbitrage are complex. "Large Platonic Markets with Delays" provides a rigorous theoretical foundation for these phenomena, exploring whether the introduction of time lags creates new opportunities for risk-free profit (arbitrage) or if the fundamental "no-arbitrage" properties of a market are preserved under such constraints.
Modeling Market Latency
The paper builds upon the concept of Large Platonic Markets, a framework designed to handle an infinite or uncountable number of assets while allowing for "restricted information," where traders may only observe a subset of the total market data. Within this framework, the authors define two distinct stochastic delay mechanisms:
- Information Delay ($\delta$-delay): This represents a scenario where the information available to a trader at time $t$ is actually a snapshot from an earlier time $\delta(t)$. This models the "latency" in data feeds.
- Order Execution Delay ($\pi$-delay): This represents the lag between the moment a trader submits an order and the moment the transaction is actually recorded at the exchange.
Unlike previous research that often relied on constant time lags, this model allows delays to be inhomogeneous and random. This means the delay can fluctuate based on the specific asset, the time of day, or random market events, mirroring the volatility of real-world digital infrastructure.
The Inheritance of Fairness
The core contribution of this work is the proof of inheritance. The authors utilize advanced probability theory—specifically filtration theory and stopping times—to show that if a standard market satisfies the "No Asymptotic $L^p$-Free Lunch" ($NAFL_p$) condition, a version of that market with added information or execution delays will also satisfy it.
Essentially, the research proves that delays do not create new arbitrage opportunities out of thin air. If a market is fair in its "instant" form, it remains fair when slowed down. This provides a formal theoretical justification for the use of speed bumps by exchanges: they can curb the speed advantage of certain participants without undermining the fundamental integrity of the price discovery process.
Key Insights and Theoretical Implications
The Duality of Delays
The authors demonstrate that order execution delays can often be mathematically mapped to a specific form of information delay. This unification simplifies the theoretical landscape, allowing researchers to treat different types of market latency under a single mathematical umbrella.
The "Insider" Trap
While adding a delay preserves a fair market, the authors highlight that the converse is not necessarily true. If a market is only arbitrage-free because of an existing delay, removing that delay—thereby giving a participant "instant" access—creates an "insider trading" scenario. In such cases, the previously hidden arbitrage opportunities become exploitable, potentially destabilizing the market.
Multi-Broker Strategies
The paper suggests a practical approach to verifying market fairness in fragmented environments. By analyzing a hypothetical "fastest" broker, the authors show that if this fastest entity cannot find an arbitrage opportunity, then a market where traders choose between various slower brokers is also guaranteed to be arbitrage-free.
Conclusion and Future Directions
This research confirms that the fundamental theorem of asset pricing (FTAP) holds even in the presence of complex, random delays. The results are robust, holding true even when delays vary by asset and event, making the theory applicable to modern fragmented markets where different assets exhibit different liquidity and latency profiles.
However, the model does have limitations. It relies on asset prices being càdlàg (right-continuous with left limits), which may not capture every nuance of sub-microsecond tick data. Additionally, the current framework focuses on "frictionless" arbitrage, ignoring transaction costs and spreads. Future research could integrate these delays with transaction cost models to further refine our understanding of how latency interacts with market liquidity and trading costs.
Summary updated