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.
5,018 documents
The document describes a trading system that predicts buy or sell labels and sizes positions using predicted probabilities. Despite similar out-of-sample F1 scores in training and backtest periods, the strategy loses money after commissions and slippage. The…
This explanation describes how to map market observations into a multilayer perceptron for time-series prediction. Each observation forms an input vector, and every hidden neuron computes a weighted combination of its input features, adds a bias, and applies…
This note considers how much historical data to use when forecasting financial time series, such as predicting the next daily market close. It cautions against choosing a lookback window by searching for the interval whose forecast most closely matches a…
The document addresses calculating rolling slope features for a time series stored in a pandas data frame. The original attempt fits a line to slices of volume and volume-price-trend observations, then writes results at positions beyond the current frame,…
This document is aimed at a project supervisor seeking a broad map of machine learning methods relevant to an algorithmic momentum strategy. It names families such as neural networks, reinforcement learning, regression, clustering, Markov models,…
The document asks how to estimate the degrees of freedom for a multivariate Student t distribution used to simulate asset returns. The questioner already has a sample mean vector and covariance matrix and notes that, when the degrees of freedom exceed two,…
The document considers whether a linear regression can predict a stock-related label more effectively by combining current observations with observations from earlier periods. Concatenating these feature vectors can capture temporal information, but it also…
The discussion explains why Black–Scholes–Merton is primarily an option-pricing framework, not a recipe for generating entry and exit signals in spot markets. Its geometric Brownian motion assumptions imply lognormal stock prices, but the answer notes that…
The note explains how to test whether past values of one series help predict another series over time. It recommends Granger causality, starting with an autoregressive benchmark that explains the target variable using its own lagged values. The researcher…
The document asks which market series can illustrate alternating high- and low-volatility regimes for a regime-switching model. The questioner considers oil, equity indexes, and individual stocks, but observes that a steadily rising stock series with…
The document asks for an R library that can model Markov-switching EGARCH volatility, with controls for the volatility term structure, long-term volatility, calibration, and forecasting. The answer recommends the MSGARCH package and points to its vignette…
The document asks why rough Heston calibration or neural-network approximation may use the Hurst parameter, correlation, volatility of volatility, and forward variance, while leaving out mean reversion, initial variance, and long-run mean variance as…
This exchange addresses a question about a Riemann-sum approximation to a rough-volatility prediction formula. The question concerns why a term involving the cosine of the Hurst parameter, divided by pi and multiplied by a power of the time increment, is…
The document compares data-driven prediction with theory-based mathematical modeling and presents a broader three-part taxonomy: statistical, equilibrium-based, and empirical approaches. Statistical methods seek predictive accuracy with limited prior…
The document explains how fractional differencing applies the lag operator to a price series and produces a weighted sum of current and past observations. Expanding the operator $(1-B)^d$ with the generalized binomial theorem gives the weights: each power of…
The document asks whether a skewed target must be transformed to a normal distribution and which machine learning methods rely on normality. It names Linear Discriminant Analysis and Quadratic Discriminant Analysis as methods derived under a multivariate…
The document describes an expectation-maximization approach for estimating a conditional Gaussian model when observations of either variable may be missing. It frames the conditional relationship between two variables as a form of a bivariate Gaussian and…
This document poses a factor-modeling question: how to put returns measured on different scales onto comparable scales before predicting target returns. It lists several candidate methods: subtracting the mean and dividing by standard deviation, dividing by…
The document describes a feature-selection problem in quantitative finance. The author splits a time series chronologically into training, test, and out-of-sample periods, then uses Spearman correlations and a p-value threshold to select features. The…
The document outlines a workflow for modeling dependence among several variables with different marginal distributions. It maps observations through empirical distribution functions to pseudo-uniform values, fits a t-copula to those values, draws new joint…
The document concerns a comparison of GARCH, GJR-GARCH, and GARCH-MIDAS models for FTSE MIB returns, with the Climate Policy Uncertainty Index as the lower-frequency explanatory variable. The researcher reports that conventional GARCH and GJR models perform…
The document discusses how to make price-derived features comparable across stocks with different price levels and across periods when an instrument’s price changes substantially. One answer recommends log returns as a common scale, followed by…
The discussion asks whether econometric time-series models and neural networks can be compared as competing approaches to forecasting, including by measures such as error and accuracy. Its central explanation is that the categories overlap: familiar…
The document sketches a high-frequency trading system split between real-time execution and analytical work. The real-time side consumes market data, maintains local limit order books, runs strategies, and creates, updates, or cancels orders. The analytical…