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 tutorial constructs a continuous futures series from individual ES contracts. It identifies the front month using trading volume, with a monotonic roll constraint, and compares volume-based switching with calendar-based timing. It explains why outright…
This notebook turns a real supply-chain knowledge graph and institutional 13F holdings into machine-learning features. It derives network measures such as centrality and dependency indicators, estimates supplier concentration and competitive exposure, and…
This notebook examines AAPL trade and quote records during the March 16, 2020 market crash to show how market microstructure changes under stress. It filters the tape to regular trading hours, distinguishes trade prints from national best bid and offer…
This tutorial builds a monthly ETF rotation simulator using a trailing risk-adjusted momentum ranking and a Treasury yield-curve regime filter. At each month-end, the rule ranks ten funds using their recent price history; risk-on months hold the top three…
This notebook demonstrates an operational cycle for an ETF prediction strategy: refresh market data, recompute financial features, retrain a Ridge model, save deployment artifacts, generate current cross-sectional predictions, replay the predictions through…
This notebook compares LightGBM with a minimal multilayer perceptron, TabM, and optionally TabPFN for cross-sectional ETF return prediction. Models are evaluated on shared walk-forward folds using rank information coefficient, with chronological validation…
This notebook develops a framework for translating market activity into modeled trading costs and strategy capacity. It introduces the square-root impact model, identifies volatility and average daily volume as key inputs, and distinguishes quantities…
This document compares four equity market data sources for studying trades, quotes, and limit order books: AlgoSeek TAQ, Databento market by order, NASDAQ ITCH, and IEX HIST. It explains each source’s granularity, coverage, cost or access conditions, storage…
This notebook isolates results for the LEAN engine from an audit comparing retained real-strategy runs with matching ML4T Backtest profiles. It identifies four supported workloads: ETF allocation, crypto perpetual funding, USD-quoted FX allocation, and a US…
This notebook introduces a workflow that connects standardized ETF data loading, a feature registry, and signal diagnostics. It shows how to discover indicator metadata, compute features using defaults or explicit parameters, and store configurations for…
This notebook compares a Random Forest with XGBoost, LightGBM, and CatBoost on a panel of anonymized firm characteristics and returns. The dataset supplies temporal training, validation, and test splits. Models are evaluated using Spearman rank information…
The notebook explains value at risk as a loss quantile and conditional value at risk, or expected shortfall, as the average loss beyond that threshold. It estimates one-day tail risk for a broad equity ETF using four approaches: empirical historical…
This notebook describes how to produce holdout predictions for a selected crypto perpetuals funding strategy. It resolves the configuration from validation results, carries its checkpoint choice forward, and builds a new training specification using data…
This notebook assesses a selected S&P 500 equity and options strategy by reconstructing its validation configuration from registered, full-coverage candidates. It traces the funnel from equal-weight baseline through allocation, risk controls, and cost…
This notebook explains Hierarchical Risk Parity as an alternative to mean-variance allocation when covariance estimates are noisy. HRP clusters assets using correlation distances, orders them according to the hierarchy, then recursively bisects the ordered…
This notebook presents a cross-case-study inventory of engineered financial features and the strongest registry information coefficient for each market study. It counts features and feature-name families, compares feature-space sizes, and summarizes…
The document explains how to identify market regimes by dividing returns into overlapping windows and treating each window as an empirical distribution. It measures the distance between windows through matched, sorted quantiles, so differences across the…
This notebook assesses a weekly S&P 500 options straddle strategy using registered backtests and a fixed selection process. It resolves the selected configuration from the registry after applying a liquid-universe restriction, then evaluates validation and…
This study builds a minute level measure of order flow pressure from market by order data for NVDA. It signs order additions, cancellations, and fills by side, weights events by size and distance from the midpoint, and smooths the resulting pressure series.…
This notebook presents a conditional autoencoder that preserves the structure of an instrumented latent factor model while replacing the linear map from firm characteristics to factor exposures with a neural network. Monthly returns are represented as…
This document explains how to create predictions for a reserved NASDAQ-100 microstructure holdout period using a configuration selected on validation data. It refits the chosen model using only history that ends before the holdout, with a label buffer wide…
This case study fits regularized linear models to returns from short at-the-money straddles held to expiry. Because premium income caps the gain while losses can grow without bound, a few severe losses dominate a squared-error regression objective. The…
This notebook uses real ETF data to assess whether a momentum signal remains useful under reasonable changes to its lookback, market regime, and implementation. It computes cross-sectional information coefficients between momentum and forward returns, then…
This notebook compares two locally run, open-weight embedding models on a retrieval task using passages from recent 10-K filings for a small set of large US companies. Both models receive the same two-sentence passages and a fixed set of queries spanning…