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 document introduces a trading model that combines GARCH forecasts of next-period volatility with quantile regression to estimate upper and lower value-at-risk boundaries. It explains how ARCH models use past squared returns and how GARCH adds lagged…
This research summary describes an equity selection process built from daily price and volume data. Genetic programming searches for candidate factors, using RankIC to favor linear relationships or mutual information to seek nonlinear ones. To focus on…
This forum excerpt concerns constructing training labels for a Chinese stock dataset and asks why label extraction produces an empty training set. The displayed expression derives a future return from a later close divided by a later open, assigns a…
This document describes an AI-assisted value strategy organized as three roles. A research agent reviews company reports and current prices to identify a small group of businesses with steady profits, durable competitive advantages, and prices judged…
This article examines how the length of an AI model’s training window can affect an equity strategy. It compares manually rolled long windows spanning several years with shorter windows ranging from a month to a few years, and recommends evaluating each…
This podcast recap profiles FC Genesis co-founder Berk Ozdogan and the firm’s approach to algorithmic trading in digital assets. It describes a progression from traditional financial markets into crypto, where the firm uses signal-based strategies and…
This tutorial explains how an AlphaModel fits a model from a prepared AlphaDataset and produces prediction scores for a chosen data segment. It describes a shared interface for fitting, prediction, and model details, then distinguishes training, validation,…
PROphet is an Expert Advisor built around two independent linear perceptrons that classify candlestick-shift inputs. One model distinguishes buy signals from flat or sell conditions; the other distinguishes sell signals from flat or buy conditions. The…
This brief BigQuant support exchange addresses a simulation error reported as having no data remaining after missing rows are dropped. The user had already tried extending the feature-extraction lookback by several different periods, without resolving the…
The post asks why a daily stock-ranking strategy sells positions in backtests but appears only to buy them in simulation. Its code includes a trailing stop based on a percentage below the highest observed price since purchase, then manages capital and ranks…
This brief BigQuant support exchange explains an error raised when a time-series minimum function receives a window argument that is not an integer. The example first derives a value from the position of the highest high within a fixed lookback, then passes…
This tutorial introduces gplearn, a Python library that uses genetic programming for regression and symbolic regression. Candidate expressions are iteratively selected, crossed, and mutated to find mathematical relationships in data. The resulting formulas…
The document describes an indicator for testing how price changes scale with the number of bars in a window. It calculates average price changes across sliding windows, dividing by the window length raised to an exponent F. It searches values from 0.1 to 1.0…
This article presents dynamic time warping (DTW) as a way to find price series with shapes similar to a manually selected reference pattern. Unlike point-by-point comparison, DTW can align sequences whose patterns unfold at different speeds or have…
The post raises a feature-selection issue in two model implementations: both select dataframe columns by taking everything after the first two columns and before the last. This works when the dataframe is ordered as datetime, symbol, features, label, because…
This overview describes quantitative investing as the use of rules or factors developed by analyzing historical market data. It presents broad stock coverage and repeated trading as a way to aggregate small predictive edges, and says quant strategies may…
The post describes a discrepancy observed after adding a rolling-training module to a strategy on a Chinese quantitative trading platform. Without rolling training, the author says simulated live trades and later tracking backtests matched. With rolling…
The document explains how Freqtrade’s Hyperopt process searches strategy parameter combinations by repeatedly backtesting historical data. It begins with random combinations and then uses an Optuna sampler to explore parameter spaces while minimizing a…
The document gives brief examples of using large-scale and unconventional data to inform investment research. Satellite images of retail and office parking areas are described as a way to estimate business activity, while satellite monitoring of shipping is…
This tutorial explains bias, variance, irreducible error, and how model complexity relates to underfitting and overfitting. It then surveys feature-selection approaches: evaluating feature subsets, using forward, backward, or bidirectional greedy search,…
The document lists fitness measures for a genetic programming module that evolves stock factors. It describes ICIR as the mean information coefficient divided by its standard deviation, with the information coefficient measuring the relationship between…
This tutorial describes generating synthetic multivariate financial time series with SDV’s PAR synthesizer, a conditional autoregressive GAN approach. It fits sequences using asset identifiers and stable context fields, with returns and volume as modeled…
This brief Chinese-language support thread concerns a rolling-training error reported as “NaTType does not support.” A response attributes the failure to a module whose output lacks a date column, and says the rolling-training process needs that field. The…
This study summary describes a low-risk portfolio approach for South Korean equities. It forecasts KOSPI 200 constituent volatility with Gaussian process regression, support vector regression, an artificial neural network, and GARCH, then ranks stocks into…