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.
219 documents
This documentation describes interactive tear sheets for examining pairs research. The cointegration view presents individual asset stationarity test results and normalized prices, then reports Engle–Granger analysis for both portfolio orientations. It…
This tutorial presents a workflow for evaluating a high-frequency grid market-making approach on Binance Futures. It covers selecting trading pairs, obtaining historical depth and trade data, converting that data into the backtester’s format, modeling…
This implementation uses a Kalman filter to update the intercept and hedge ratio between two assets as each new observation arrives. It treats one asset as the response and estimates the coefficient for the other, producing a residual spread and its…
This document explains a method for choosing entry boundaries in a cointegration-based pairs trade. The strategy fades a spread when it crosses a preset upper or lower threshold, then closes when it returns to its mean. With position weights set by the…
The code implements an Ornstein–Uhlenbeck model for mean-reverting portfolios and pairs of assets. It fits the model’s long-run mean, reversion speed, and noise variance to historical prices, then uses an optimal double-stopping framework to calculate entry…
This document contains a dated daily price series identified as RB, with fields for opening, high, low, last, and settlement prices. The visible records begin in 1994 with missing values across the price fields, while later entries show populated prices…
This code describes a data formatter for a momentum model. It defines target returns, normalized returns over several horizons, MACD features, and optional change-point, calendar, and ticker identity inputs. It also assigns columns roles such as target,…
This exchange model describes how a backtest can simulate partially filled limit orders. It supports limit orders with several time-in-force rules and uses a queue model to track an order’s position at its price level. When trades occur at that price, the…
This document describes a utility for creating synthetic order latency observations from market feed data. It keeps events that contain both exchange and local timestamps, resamples them at a configurable interval, and uses each interval’s last timestamps to…
The code describes a grid market-making approach that repeatedly places buy and sell limit orders around a forecast mid-price. The forecast is simply the current best bid and ask midpoint, with no alpha adjustment in this implementation. A relative…
This introduction reviews stochastic control models for allocating wealth between a mean-reverting spread and a risk-free asset. It describes the spread with an Ornstein–Uhlenbeck process and outlines work by Jurek and Yang, which considers investors with…
The module describes a first-stage screening method for pairs trading strategies built around copulas. It compares every two-asset combination in a supplied price panel and ranks the pairs using Spearman rank correlation, Kendall rank correlation, or the…
This module fits an Ornstein–Uhlenbeck (OU) process to either one price series or a spread formed from two asset price series. It estimates the long-run level, reversion speed, and volatility by maximizing a likelihood for discretely sampled observations,…
This documentation describes a data-import utility for obtaining commonly used market data to benchmark quantitative algorithms against real-world prices. It can retrieve ticker collections for major US equity universes, including S&P 500 and Dow…
This module implements analytic methods for selecting entry and exit thresholds for a mean-reverting spread, following a published optimal-threshold framework. It transforms thresholds into dimensionless units using the model’s mean-reversion speed, long-run…
The document describes a bivariate Nelsen 14 copula, a model for dependence between two variables after expressing them on uniform scales. It provides analytical forms for the copula cumulative distribution, density, and conditional probability, along with a…
This reference explains several ways to measure dependence among asset returns and distance between correlation structures. Distance correlation can detect nonlinear dependence and is zero exactly when variables are independent, unlike Pearson correlation,…
This module implements a relative-value trading rule built around a two-state Markov regime-switching model. It fits the model to a univariate time series, identifies the current high-mean or low-mean regime, and uses the estimated regime mean and standard…
This document describes a backtest exchange model for limit orders that treats every execution as a full fill. It supports good-till-canceled and post-only orders, tracks orders by price level, and uses a queue model to estimate whether trades at an order’s…
This introduction defines codependence as a relationship in which information about one random variable helps determine another, while emphasizing that dependence does not establish causality. It presents Pearson correlation as a familiar measure, then…
This document describes a market-depth implementation that stores bid and ask quantities by price tick in hash maps and tracks the current best bid and ask separately. It supports both aggregated level-two depth and level-three order records, with operations…
This introduction explains how copulas separate dependence between variables from their individual marginal distributions, then presents vine copulas as a way to model dependence across many variables. Rather than impose one rigid high-dimensional copula, a…
The document outlines a pairs selection framework that first reduces security-return features with principal component analysis, then clusters the compact representations using OPTICS or DBSCAN. OPTICS can identify clusters without a fixed cluster count;…
This document explains a Rust reader for supplying chronological market data to a backtest. Data can come from NumPy files or in-memory inputs, and a cache tracks active readers so loaded datasets can be reused and removed when no longer needed. The reader…