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 response explains how Bayesian modeling could relate strategy performance to changing factors, while correcting the idea that historical observations themselves constitute a prior. It distinguishes observed data from prior information about unknown…
The document poses a research question about whether a stock’s price relative to its 52-week high can help explain subsequent returns. The proposed setup uses the price-to-high ratio as a predictor and a forward return, such as the following week’s return,…
The document explains why least squares and maximum likelihood can produce the same parameter estimates in a linear regression with normally distributed errors. Writing the conditional likelihood for each observation and taking its logarithm gives a constant…
The document recommends Python tools for work commonly taught through time series software. It presents pandas as the core library for representing and managing dated observations, with Series and DataFrame structures. The capabilities highlighted include…
The document asks how to include yesterday’s forecast error in a regression intended to predict future equity returns. Its replies connect lagged residuals to moving-average terms: an MA model uses past errors, while combining lagged observations and lagged…
The document considers estimating a separate slope coefficient for each intraday time interval when past returns are used to predict subsequent returns. With few days of observations and a weak relationship, independent ordinary least squares estimates may…
The document examines a method for estimating how long one time series appears to follow another. The author runs Granger causality tests across candidate lag lengths and selects the lag with the largest F-statistic. That approach gives a plausible short lag…
The document asks whether an American Statistical Association statement on p-value misuse will affect finance. The response frames the issue as part of a wider critique of statistical significance, pointing to Deirdre McCloskey’s work on the costs of…
The document raises a conceptual question about combining seasonal ARIMA and GARCH models for time-series forecasting. It distinguishes the familiar SARIMA prediction of future series values from a GARCH forecast, which the author understands as producing…
The document considers how to validate a low-frequency crypto strategy with roughly 25 trades per year, focusing on the limits of statistical evidence from a small sample and the dependence introduced by adding correlated assets. One reply suggests extending…
The document discusses language choices for a pricing library that should support parallel pricing, control maintenance costs, and remain accessible to newer quantitative researchers. It argues that there is no single best language and suggests Python as a…
The document frames the Jane Street market prediction task as choosing whether to take each opportunity using anonymized features, trade weights, and realized outcomes. Its objective combines aggregate profit with a clipped Sharpe-like factor, so a…
The document surveys ways to combine fundamental, market, analyst, and alternative data in quantitative investment analysis despite their different observation frequencies. One option is to resample inputs to a common frequency; another is to train separate…
The document collects recommendations for learning C++ with quantitative finance applications in mind, particularly for a Python developer who wants to build pricing libraries. One reply suggests choosing a book suited to the learner’s current level: an…
The document proposes clustering stocks using financial statement ratios such as price-to-book, price-to-earnings, and return on equity, potentially alongside market measures such as beta and annual price change. The suggested workflow is to form groups from…
The document outlines a practitioner framework for evaluating alternative data signals when simple relationships may already be widely known. It recommends writing a specific hypothesis and pass/fail criteria before inspecting results, with the decision…
The discussion considers how to estimate covariance when historical relationships may shift with observable conditions. Although framed around selecting a portfolio of athletes, its examples—player matchups and racing-track characteristics—illustrate why…
This discussion asks how to generalize covariance and correlation when modeling dependence in heavy-tailed settings and extreme events. One response proposes mutual information, which quantifies how much observing one variable reduces uncertainty about…
This exchange clarifies how to interpret labels in the triple barrier method. Each labeled instance corresponds to a specific trading circumstance and receives a single outcome according to which event closes it: an upper barrier win, a lower barrier loss,…
The document raises a model-development question about choosing observation dates for behavioral credit scoring. Such models use account activity, arrears, advances, and other customer information to estimate whether an existing borrower will default during…
The document examines a proposed return model combining Fama-French factors with industry random intercepts. The modeler reports a strong relationship between residuals and fitted values and suspects that correlations among industry effects or a nearly…
The document considers how to adapt a return-prediction neural network from time bars to dollar bars when its features and prediction schedule depend on regular clock times. Dollar bars do not arrive at the same time or count across instruments, so the…
The document asks whether two state-space formulations used with Kalman filters are equivalent when the latent state has a nonzero mean. One formulation expresses the transition in deviations from a mean; the other uses a direct autoregressive transition and…
The document discusses how to evaluate a market forecasting model when each observation at time T is paired with an outcome measured n days later. Randomly assigning calendar years to training and testing sets can create boundary problems because future…