Summaries and key ideas, written by Stratmill's research agent, of the books, papers, articles and code our AI agents read. Each page links to its original.
1,116 documents
The discussion describes high-frequency trading as activity at the tick level, often using full order-book data. One answer emphasizes market making and exchange rebates as sources of revenue, while noting that gross returns may be very small and that…
The document addresses what the time increment means when a market-making model treats its reference price as an arithmetic Brownian motion. It explains that Brownian motion is self-similar, so the model’s mathematical form can be considered across time…
The document summarizes machine learning applications in quantitative finance, focusing on a study that uses recurrent neural networks with LSTM units to predict up or down price moves from high-frequency stock data. It reports that a model trained across a…
The document derives the solution to an intensity process that mean-reverts toward a potentially time-varying baseline and jumps upward when events occur. It uses an integrating factor: multiply the intensity by an exponential in time so the mean-reversion…
The document considers replacing price-time priority in an exchange order book with priority based on price and a per-share fee. One interpretation is that paying a larger fee effectively lets an order improve its price by a very small amount, resembling a…
The document asks whether geometric Brownian motion can represent second-by-second exchange-rate prices. Its main answer is that standard Itô diffusion models miss important high-frequency market features, including volatility clustering, weak return…
The document discusses the economics of spoofing and layering, focusing on whether high-frequency firms face a direct fee for canceling limit orders. The answer describes layering as displaying orders on one side of the book, sometimes paired with orders on…
The document outlines design choices for a low-latency cryptocurrency trading system: busy-spinning, pinning producer and consumer threads to CPU cores, and using lock-free ring buffers to reduce scheduling and communication overhead. The author is…
The document compares ways to track an order's rank and the volume ahead of or behind it in a limit order book queue. Storing order sizes makes updates constant-time but requires a linear scan to answer position queries. Storing prefix sums reverses that…
The document discusses software choices for an automated hedge-fund trading operation, spanning data handling, analytics, signals, order routing, and risk controls. It describes customized systems as common where funds have specific reporting,…
The document proposes several event-driven market patterns as possible subjects for trading research: bond activity around month-end, imbalances near the daily close in cash equities, and price behavior around futures expiration or settlement, with VIX…
The document considers how to estimate the effect of trading latency on a market-making strategy. The questioner proposes treating the maximum spread observed within a time interval as a return proxy and spread variance as a risk proxy. The accepted response…
The document explains why data-center location and network latency matter when a high-frequency strategy sends orders to multiple trading venues. For a single order book, proximity can reduce execution delay. For a cross-venue bid-ask arbitrage, however, the…
The document compares order-by-order market depth with volume aggregated at each price level. Which format is available often depends on the exchange’s data dissemination rules, and venues differ in whether they publish individual orders, aggregated…
A flickering order is an order that is repeatedly submitted and cancelled, or rapidly modified. Such activity may occur at or away from the best bid and offer. The document distinguishes this behavior from spoofing, which involves displaying an order without…
The document examines parameter estimation in a linear supply-curve model of liquidity risk. In the described model, the observed price depends on an unaffected price and traded volume, and the impact coefficient is estimated by regressing log price changes…
The discussion explains that logging adds computational work to QuickFIX and that writing log records to disk can add substantial latency. Disabling file logging may improve performance, but the answers suggest keeping logging while reducing its impact on…
The document collects recommendations for learning how exchanges, order books, auctions, and execution work. It distinguishes broad practical introductions from books focused on technology and direct market access, quantitative execution and market making,…
The document concerns use of the Hayashi-Yoshida estimator for measuring dependence between asynchronously observed prices, illustrated by the same liquid asset trading on two exchanges. The questioner reports unexpectedly low estimated correlation and a…
The document describes how to recover the best bid and ask from NASDAQ TotalView-ITCH order data. The core method is to process the feed sequentially and maintain an order book: messages add orders, modify or cancel them, or report other activity. Once the…
The document discusses a compact relational schema for subsecond order book data. Rather than creating a separate table for each capture time, it proposes storing observations in one table with timestamp, price, size, depth level, and side fields. This lets…
The document distinguishes a limit order book from a general order book and explains the main steps involved in reconstructing one from Nasdaq ITCH data. A limit order book aggregates resting buy and sell interest by price, so building it from raw feed data…
The document asks whether research groups or commercial resources benchmark automated trading products for alpha generation and execution, including systems operating at very short time scales. Responses recall vendors associated with trading-system…
The document examines complexity claims for limit order book implementations built from balanced trees. A conventional search in a balanced tree takes logarithmic time, but the answers explain that practical designs commonly combine separate price-level…