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.
1,124 documents
This notebook builds a monthly model matrix for a US firm-characteristics study from a released panel whose characteristics have already been ranked across firms. It adds composites and interactions using values in the same firm-month row, without fitting…
This notebook constructs four model-based features for S&P 500 option research: next-day volatility forecasts from GJR-GARCH and stochastic volatility, plus the differences between each forecast and the option market’s implied volatility. It explains how the…
This document fits TabM, a weight-sharing neural ensemble, to option-derived features for predicting forward equity returns and risk-adjusted returns. Its members use one shared two-layer network, with separate activation scaling and output heads, then…
This document presents a structured state record for research agents, separate from their conversation transcript. The record stores the question and cutoff, gathered evidence, unresolved questions, tool-call history, gate outcomes, and synthesis status.…
This document studies position-level exits added to an existing crypto perpetual strategy: fixed stop losses, trailing stops, and time-based exits. It explains that these controls only close positions selected by the underlying strategy, and evaluates them…
This document explains how a forecasting agent’s search interface can constrain its evidence and make its activity auditable. A provider-neutral protocol returns typed results with source, publication date, relevance score, and content, while a cutoff filter…
This notebook examines linear-model regularization for predicting returns across currency pairs. It distinguishes collinearity among input features from dependence among the assets: currency pairs share underlying currencies, so their returns and rankings…
This configuration note explains the choice of the saga solver and a tighter convergence tolerance for L1 logistic regression on three-class labels. It contrasts saga with liblinear, citing runtime measurements on a Nasdaq microstructure panel and reported…
This module describes checks for whether prediction and backtest inputs cover the sessions and folds specified by a study. Relative checks compare candidate models with one another, so they can miss a shared failure: stale fold boundaries or an upstream join…
This notebook studies a supervised autoencoder for predicting forward returns from firm characteristics. It retains a narrow neural-network bottleneck but removes the conditional-factor interpretation and pricing restriction used by related models. The…
This notebook converts FX pair predictions into baseline trading results. At each decision time it ranks pairs, takes equal-sized long and short sleeves, and runs the selected prediction configurations and checkpoints through an existing backtest engine.…
This feasibility analysis checks whether a monthly long-short strategy based on US firm characteristics can be evaluated with the available panel. It verifies that the measures are encoded as ranks, describes how long firms remain observable, counts firms at…
This notebook compares two simulations of the same monthly ETF momentum strategy. Both use identical target weights, universe, dates, and total trading cost. One computes returns from lagged weights and asset returns; the other processes orders sequentially…
This notebook distinguishes predicting returns from claiming that momentum causes them. It treats a stock’s prior-year return, excluding the latest month, as the treatment and forward return as the outcome. Recent volatility, illiquidity rank, and volume…
This notebook fits PatchTST as one member of a declared sequence-model population for S&P 500 options. PatchTST divides a lookback window into fixed-size patches, embeds them as tokens, and uses attention to learn relationships among portions of the history.…
This notebook aggregates LightGBM results across multiple trading case studies and compares them with linear and TabM baselines. It examines the highest-IC configuration for each case study, reports heteroskedasticity-and-autocorrelation-consistent…
This utility describes a diagnostic workflow for data artifacts produced by quantitative research pipelines. It profiles numeric columns for nulls, non-finite values, zeros, distinct values, quantiles, and tail behavior, then flags columns that cross…
This notebook tests whether learned graph embeddings add predictive information to a tabular model for US stocks. It builds a network from absolute return correlations measured before the target period, constructs momentum, volatility, and trend features…
This notebook compares ways to measure volatility and develops heterogeneous autoregressive (HAR) volatility models alongside roughness analysis. It uses intraday returns to estimate session realized variance, distinguishes intraday movement from overnight…
This document describes an LSTM sequence model for an S&P 500 options study. Unlike cross-sectional models that rely on engineered features to summarize a symbol’s history, the LSTM receives chronological windows and learns which parts of the sequence…
The notebook builds prediction intervals around gradient boosting forecasts with split conformal prediction. It fits a model on an earlier training segment, measures absolute residuals on a later calibration segment, and uses a finite-sample order statistic…
This notebook compares four reference ETF allocations using historical crisis windows, hand-authored simultaneous asset shocks, Student-t Monte Carlo simulations, and market-regime statistics. It explains a maximum drawdown measure and uses daily…
This notebook compares CSV, Parquet, Feather, and HDF5 storage using a deterministic one-million-row OHLCV panel. It measures write time, materialized read time, file size, and reads that select only two columns. The timing policy uses a single write…
This notebook presents a pipeline for turning financial headlines into stock-level signals. It cleans and standardizes a news corpus, removes exact and prefix-matched duplicates within ticker-date groups, and uses sentence embeddings to represent headline…