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 technical note explains how an MQL5 script can inspect an ONNX model’s inputs and outputs before running inference. It retrieves each input and output count, name, and type information, then reports tensor data types and dimensions. The example uses a…
The document describes a stock-price forecasting study that compares a Kalman filter, XGBoost, and a hybrid that uses XGBoost to select features before passing them to a Kalman filter. Inputs include price and volume data, accounting measures, and quarterly…
This brief answer addresses how to keep multiple StockRanker models available for reuse. It recommends following a separate guide’s approach to load the models, then placing them together in a dictionary for storage. The intended benefit is faster future…
This article introduces dense neural networks and backpropagation, then outlines an A-share stock selection experiment. A DNN uses connected layers of neurons; training sends predictions forward and propagates errors backward to update weights. The example…
The document outlines a pairs selection framework that first reduces security-return features with principal component analysis, then clusters the compact representations using OPTICS or DBSCAN. OPTICS can identify clusters without a fixed cluster count;…
The article investigates why XGBoost training took far longer in some Python process setups than with joblib’s loky backend. Its experiments compare direct execution and several multiprocessing approaches, then identify the OpenMP thread limit as the key…
This BigQuant backtest handler describes a daily portfolio process driven by a date-filtered prediction ranking. It allocates roughly one holding-period fraction of portfolio value to daily purchases during an initial staging phase; afterward, it can use up…
The article explains why investment teams need to understand how machine-learning models produce decisions, especially when those decisions affect clients, risk controls, or deployment choices. It contrasts inherently interpretable approaches, such as linear…
This overview explains why feature selection matters: removing uninformative inputs can reduce dimensionality and training time, limit overfitting, and improve generalization. It frames selection as choosing a subset that optimizes a model-related criterion,…
The document raises a modeling question about predicting whether a future price move is positive using a label that takes only two values. The label compares a later close with an earlier open and assigns one class for an increase and the other for a…
This article presents a time-series momentum strategy that switches between a short lookback signal and a longer one according to market volatility. The short signal uses one month of returns and the slow signal uses twelve months; each takes a long or short…
The document explains how to build trading agents within a strategy framework. It describes a workflow in which research agents gather market evidence and a dedicated trading agent can submit orders, while deterministic Python can enforce fixed execution and…
The study estimates a fund’s factor exposure from its net asset value returns and factor return series, aiming to provide more timely exposure tracking than holdings-based estimates from delayed periodic reports. It trains an LSTM on features describing the…
This README catalogs resources for researching and implementing systematic trading, including software libraries, strategies from published papers, books, videos, blogs, and courses. It groups tools by functions such as backtesting, live trading, analytics,…
This forum exchange addresses a user who sees predictions appear quickly from XGBoost and random forest modules, without a visible training process, loss, or mean squared error. The reply suggests that the dataset may be small and recommends using dedicated…
This overview describes a wizard-style strategy builder intended to help newcomers assemble quantitative strategies through guided interface steps, while also supporting faster prototyping for experienced developers. Users can select templates and building…
This brief forum exchange concerns exporting hyperparameter-search results to a spreadsheet when displaying a large result set makes the web page slow or unusable. The platform response says direct Excel export is not supported and advises users to reduce…
This Chinese-language overview surveys Microsoft Research Asia’s published work on applying artificial intelligence to quantitative investing since 2017. It says the research spans stock selection, risk models, algorithmic trading, data augmentation,…
This user proposal for a quantitative trading platform focuses on making research and strategy evaluation more credible. Suggestions include allowing users to schedule live strategy runs, screening factors using their information coefficients over prior…
This notebook outlines an end-to-end equity alpha research workflow for CSI 300 constituents. It loads daily constituent data, constructs an Alpha158 dataset, divides observations into training, validation, and test periods, applies missing-label removal and…
The document introduces DDG-DA, a method for adapting forecasting models when streaming data changes over time. Rather than waiting to detect a shift and then fitting to recent observations, it first predicts how the data distribution may evolve, generates…
This assignment describes two Chinese equity ranking approaches focused on micro-cap stocks. Both screen for positive trailing earnings, the smallest market-cap segment, and exclude special-treatment, suspended, and recently listed shares. The XGBoost model…
The document introduces cosine similarity as a way to compare two vectors by the angle between them. It also describes the equivalent calculation as their dot product divided by the product of their magnitudes. Cosine distance is presented as one minus…
This short platform troubleshooting note addresses a backtest failure that appeared after parameters were changed. Its proposed fix is to remove the inline comments from entries in the feature list. The post points readers to an example of the adjusted…