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
This study examines high-frequency trading characteristics for Bitcoin and Ether in the post-COVID period. It analyzes returns, the average number of trades per time unit, and traded volume using multifractal detrended cross-correlation analysis, a method…
This study asks whether financial price series contain detectable structure beyond a random walk. It argues that conventional time-series methods may have difficulty distinguishing small departures from randomness, and instead applies modern multivariate…
This report examines how to detect changes between more stable and more unstable market conditions in real time. It focuses on Bayesian Online Changepoint Detection (BOCPD), a method for identifying likely structural breaks as observations arrive, and…
MacroHFT is a reinforcement-learning approach for minute-level cryptocurrency trading. It addresses two stated challenges in existing methods: agents may overfit and fail to adapt their policies to financial context, while decisions from a single agent can…
This study examines whether Bitcoin trading developed statistical features associated with more mature financial markets. It analyzes one minute price changes beginning in 2012 across selected subperiods, focusing on return distributions, volatility…
The paper develops general compound Hawkes processes and studies their properties in limit order books. It establishes a law of large numbers and functional central limit theorems for several specific versions of these processes, providing a mathematical…
The study tests the Informer forecasting architecture as a basis for automated strategies on high-frequency Bitcoin data. It trains variants with RMSE, Generalized Mean Absolute Directional Loss (GMADL), and quantile loss, then turns predicted future returns…
The paper studies asymmetry between Bitcoin returns and volatility, measured using squared returns, across daily and higher-frequency data. It reports that return–volatility cross-correlations are mostly insignificant at the daily level. At higher…
This paper develops generalized versions of order flow imbalance indicators to explain short-term stock price changes. Its construction accounts for changes in non-minimum quotation units in actual trades, extending conventional Order Flow Imbalance (OFI)…
The paper studies order flow imbalance (OFI), an indicator based on the asymmetry between bid-side and offer-side events that is associated with short-term price direction. It uses Hawkes processes to estimate OFI while modeling lagged dependence between bid…
This paper describes a stock pairs-trading system that searches across an N-stock universe instead of monitoring only a preselected pair. It represents stocks as nodes in a directed graph, with edge weights based on instantaneous price differences and…
This study maps changing relationships among cryptocurrencies using high-frequency returns from 2020 to 2025. It creates weighted, directed networks from statistically significant Granger causal links between asset log returns, treating link direction and…
This research examines how aggregated trade flow relates to price changes across intraday time scales. It reports that price impact follows a nonlinear pattern that is broadly similar across instruments, while the specific curves differ between trade-volume…
This study uses computational experiments with a modified Mike–Farmer order-driven model to examine how order flow shapes autocorrelation in ultra-high-frequency stock returns. It measures return dependence with the Hurst index and varies three order-flow…
This paper introduces an Extended State-Dependent Hawkes Process for modeling limit order book activity. The model allows states to disappear, a feature intended to represent high-frequency market behavior. The authors use Karush–Kuhn–Tucker conditions to…
This paper models interaction between a slow institutional investor and a high-frequency trader whose combined orders affect an asset’s price. The faster trader observes a price-predicting signal more often but faces periodic end-of-day inventory…
The paper proposes a quantum approach to high-frequency statistical arbitrage across stocks. It combines variable-time condition-number estimation with quantum linear regression, and also develops supporting methods for estimating condition numbers and…
The document describes an empirical market invariant that links an instrument’s volatility with traded volume, average spread, and volume available in the order book. It uses this relationship to estimate what the authors call instantaneous volatility,…
The document introduces a realized local volatility surface as a way to estimate conditional volatility during abrupt movements in an underlying asset. It frames the estimate as a generalized Wiener measure derived from historical prices and links its use to…
This document compares two explanations of price impact: a stationary Kyle model, where order flow conveys information about fundamentals, and a propagator model, where trades affect prices through a decaying impact kernel. The comparison finds that the…
This overview surveys how orders arriving in a limit order book contribute to price formation. It considers theoretical and empirical approaches to modeling order flow, and critically reviews recurring patterns observed in market data. The discussion…
The paper analyzes high-frequency intraday Bitcoin data from 2019 to 2022, separating the sample into two periods with an abrupt change in volatility. It characterizes returns as anomalous diffusion: subdiffusive over short intervals and weakly…
The paper models prices as moving in response to order flow. Market-order price impact and the arrival rates of market and limit orders depend on a liquidity process, with the two order types mutually exciting one another so that liquidity tends to revert. A…
This study examines whether access speed affects the profitability of high-frequency order book imbalance (OBI) trading. Because historical data generally do not reveal traders’ identities or locations, the authors use an agent-based, discrete-event…