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
This stock-ranking strategy adds recent price-limit activity to a set of price, volume, technical, and financial features. The document explains that the limit-status field uses separate values for limit-down, no limit, and limit-up days; summing the last…
The article introduces crypto market microstructure analysis through granular data such as order books, liquidity, and order flow, then surveys machine-learning approaches that may help process large datasets. It describes recurrent neural networks such as…
This brief support exchange concerns an error encountered while using an SVR model and asks whether the model needs only training data and a feature list. The response identifies a more immediate issue: the upstream module output is missing two named…
This brief report summary proposes a two-stage equity-selection process. A fundamental model first creates a preliminary stock universe, and a deep learning model then screens that universe again. The stated aim is to improve excess returns while keeping…
This report describes a Chinese equity-selection process built from daily price and volume data. Genetic programming generates candidate factors: Rank IC is used to find factors with linear predictive relationships, while mutual information targets nonlinear…
This practitioner report reviews DeepAlpha, a neural model used to rank stocks from factor inputs, and compares it with other approaches. The author says a simple fully connected network appeared competitive with decision-tree and linear models, and…
This overview catalogs machine learning approaches that can be applied to quantitative investing, grouping them into tree models, linear models, unsupervised methods, deep learning, ensembles, and online learning. It links model families to tasks such as…
This report describes a monthly stock selection model that uses a random forest, an ensemble of decision trees trained with bagging, to estimate each stock’s probability of rising in the next period. Its workflow covers feature and label preparation,…
This Russian-language overview introduces LumiBot, a Python framework for building, backtesting, and running trading strategies through supported brokers. It describes a shared strategy lifecycle for hand-coded rules and AI-assisted agents, with historical…
The document reports a platform-specific issue in which an XGBoost classifier’s probability prediction function returns only endpoint values of zero and one. The user expected continuous class probabilities that could be used to rank securities, based on…
This introductory guide explains how machine learning can be applied to quantitative investing through a staged research workflow. Using a fruit-selection analogy, it describes gathering historical data, defining the prediction target, labeling observations,…
This discussion distinguishes machine learning for execution optimization from machine learning for directional trading. Execution begins with a known trade and seeks to reduce its cost; alpha generation must identify price direction and capture enough of…
This Chinese-language support post describes a feature-engineering question: how to count, over a recent window, the days when a five-day moving average exceeds a ten-day average. The proposed first feature assigns a positive or negative value according to…
The document raises a question about possible look-ahead bias in a stock-selection workflow. It points to a five-day return calculated from prices shifted forward in time and asks whether using that value as a feature in a test set leaks future information.…
The document presents a modified MACD indicator built by replacing standard exponential moving averages with a three-stage smoothing expression. It applies that transformation separately to closing prices over 12- and 26-period spans to form DIF, then…
This configuration specifies a Qlib time-series Transformer workflow for predicting Chinese CSI 300 stock returns from Alpha158 features. It defines a Chinese market data source, uses the CSI 300 as the instrument universe and SH000300 as the benchmark, and…
This research summary describes a unified framework for explaining and testing generalized linear models in multi-factor stock selection. It compares how these models can combine factors into a composite signal, then evaluates the signal through grouped…
This document describes a short-term screen for main-board stocks. It selects shares with turnover between 3% and 12%, a daily gain above 1% relative to the same market-board comparison, and an opening-auction move between -2% and 5%. It provides both a…
This article distinguishes overfitting from overtraining and explains why cross-validation does not, by itself, prevent a model from overfitting. Validation estimates predictive performance, including through out-of-sample data and averaged errors across…
The document describes an Expert Advisor for XAUUSD that combines a price dynamics filter, an online-learning neural network, and volatility-based risk management. Its market-state component derives velocity from price changes and an entropy-like measure…
The document explains a basic result in convex optimization: for a differentiable convex function, any point where the gradient is zero is a global minimizer. It argues this by applying the first-order inequality for convex functions, which bounds the…
This study examines how neural network hidden-layer configurations affect reinforcement-learning performance in stock portfolio optimization. Using the FinRL framework, agents receive open, high, low, close, and volume data and seek to improve portfolio…
The article describes using a deep neural network to estimate risk for an automated, beta-neutral long–short equity strategy. Financial statement data and historical prices serve as inputs; the model predicts return quantiles over horizons from five to 90…
This overview organizes machine learning algorithms in two ways: by learning setup and by functional family. It distinguishes supervised learning with labeled examples, unsupervised learning that finds structure in unlabeled data, and semi-supervised…