Paper: arXiv 2309.04259
Authors: Paul Bilokon, Burak Gunduz
Abstract
This work aims to bridge the existing knowledge gap in the optimisation of latency-critical code, specifically focusing on high-frequency trading (HFT) systems. The research culminates in three main contributions: the creation of a Low-Latency Programming Repository, the optimisation of a market-neutral statistical arbitrage pairs trading strategy, and the implementation of the Disruptor pattern in C++. The repository serves as a practical guide and is enriched with rigorous statistical benchmarking, while the trading strategy optimisation led to substantial improvements in speed and profitability. The Disruptor pattern showcased significant performance enhancement over traditional queuing methods. Evaluation metrics include speed, cache utilisation, and statistical significance, among others. Techniques like Cache Warming and Constexpr showed the most significant gains in latency reduction. Future directions involve expanding the repository, testing the optimised trading algorithm in a live trading environment, and integrating the Disruptor pattern with the trading algorithm for comprehensive system benchmarking. The work is oriented towards academics and industry practitioners seeking to improve performance in latency-sensitive applications.
Complexity vs Empirical Score
- Math Complexity: 4.5/10
- Empirical Rigor: 6.0/10 (capped from 7.0: no out-of-sample test; see penalty flags)
- Quadrant: Street Traders — empirical and applied, lower math complexity (effective empirical rigor capped from 7.0)
Why this score: The paper focuses on practical C++ optimization techniques and includes backtesting with statistical significance metrics, but lacks advanced mathematical derivations beyond basic statistical arbitrage concepts.
Research Flowchart
flowchart TD A["Research Goal<br>Bridge knowledge gap in<br>optimizing latency-critical C++ code<br>for HFT systems"] --> B["Key Methodology<br>Created LLPR & Implemented<br>Disruptor pattern; Optimized<br>Statistical Arbitrage Strategy"] B --> C["Data & Inputs<br>Market data streams<br>Algorithmic trading logic<br>Traditional queuing models"] C --> D["Computational Process<br>Rigorous statistical benchmarking<br>Cache optimization &<br>Latency analysis (speed, cache util)"] D --> E["Key Findings & Outcomes<br>• Disruptor > Traditional queues<br>• Significant latency reduction<br>• Validated profitable strategy<br>• Future: Live deployment"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes three contributions: a Low‑Latency Programming Repository, an optimised market‑neutral statistical arbitrage pairs‑trading strategy, and a C++ implementation of the Disruptor pattern.
Source sentence
the creation of a Low-Latency Programming Repository, the optimisation of a market-neutral statistical arbitrage pairs trading strategy, and the implementation of the Disruptor pattern in C++.
Assumes 3
- The price spread between correlated assets is mean‑reverting, enabling a market‑neutral strategy. model
Source sentence
By leveraging the mean-reverting nature of price spreads between correlated assets, it provides a market-neutral approach to investment
- Cointegrated asset pairs exhibit a stationary linear combination, allowing mean‑reversion trading signals. statistical
Source sentence
If Yt and Xt are both non‑stationary but their linear combination, say Zt = Yt −γXt, is stationary, then Yt and Xt are cointegrated.
- Historical daily closing prices from Yahoo! Finance accurately represent the price dynamics of the assets used for cointegration testing. data
Source sentence
The data for this analysis was obtained from Yahoo! Finance containing the relevant adjusted closing prices.
Data
- Adjusted closing prices for Goldman Sachs (GS) and Morgan Stanley (MS) market-prices public
Source sentence
Multiple US equities were tested for cointegration over five year period with daily closing price as the intervals. Goldman Sachs Group Inc (US:GS) and Morgan Stanley (US:MS) showed strong cointegration.
Costs
Unclear
Finds 5
- Reduced branching technique cut average execution time from 7.35 ns to 4.68 ns per iteration, a ~36% speed increase. performance
Source sentence
The conventional branching approach resulted in an average execution time of 7.35 nanoseconds per iteration. In contrast, the reduced branching technique yielded an average time of 4.68 nanoseconds per iteration, marking an approximate speed increase of 36%.
- Prefetching improved function speed by ~23.5%, reducing execution time from 8,235,924 ns to 6,301,400 ns. performance
Source sentence
prefetching has improved the execution speed of the WithPrefetch function by approximately 23.5% (8235924 ns for NoPrefetch vs. 6301400 ns for WithPrefetch).
- Cache Warming and Constexpr each delivered roughly 90% latency gains in repository benchmarks. performance
Source sentence
Cache Warming and Constexpr showing the most significant gains at approximately 90% each.
- Optimised pairs‑trading algorithm reduced latency by 87.38%, from a mean of 517,559 ns to 65,588 ns. performance
Source sentence
optimised version of the trading algorithm significantly reduced latency by 87.38%, reducing the mean time from 517,559ns to 65,588ns.
- Individual techniques contributed: Inlining 21.41%, SIMD 31.28%, Buffer Data Structure 48.58% latency reduction. performance
Source sentence
Techniques like Inlining, SIMD, and Buffer Data Structure individually contributed to latency reduction by 21.41%, 31.28%, and 48.58% respectively.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 26 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.