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.
21,023 documents
This competition entry modifies a Deep Alpha convolutional neural network by supplying five time slices for each of 98 factors, rather than using a single slice. The network structure and most parameters remain close to the original model because training…
This report examines changes in concentrated holdings among Chinese active equity funds using quarterly disclosures and a daily estimate of fund exposure to a core group of crowded stocks. It compares the core group’s holdings over time, separates changes in…
This MT5 indicator summarizes closed-deal history from the trading account and displays the results in a chart panel. It reads deals through the terminal’s history functions, aggregates them over a configurable lookback period, and refreshes on a timer. It…
The article describes an integrated order flow imbalance factor built from changes in bid and ask quantities across five limit order book levels. It explains how each level’s imbalance can be normalized by typical depth, then combined with principal…
VAMA is described as a simple moving average that takes each bar’s tick volume into account. The indicator has two user settings: the calculation period and the applied price. The description identifies the indicator’s basic inputs and its volume-aware…
This support exchange concerns an IndexError that occurs when a Chinese convertible bond data table is read during a backtest. The same table can be read outside the backtest, and comparable reads of other tables within the backtest reportedly succeed. That…
The report studies temporal changes in order-book depth and imbalance for Binance’s BTC/FDUSD market using minute-level data collected from July 1 to August 12, 2025. It describes an intraday depth peak around 11:00 UTC and a trough around 21:00 UTC,…
This guide explains advanced ways to configure strategy hyperoptimization in Freqtrade. It shows how to define a custom loss function, which receives trade results and backtest context and returns a score where lower values are preferred. The example…
This overview summarizes six ICML 2022 papers selected by G-Research staff. The topics include convex formulations for training two-layer ReLU networks, linear-time attention, branch-and-bound optimization of shallow decision trees, structured matrices for…
This note explains beta as a measure of an asset’s market risk exposure and introduces the Fama–French three-factor model as a way to capture size and value style exposures. It describes how to quantify style performance with long-short portfolios: small-cap…
This research proposes a stock selection factor that estimates similarity between companies using ten financial measures spanning profitability, solvency, operating efficiency, and growth. After cross-sectional normalization, it uses cosine similarity to…
The CheckMark pattern is presented as a way to classify the next candle as bullish or bearish. It compares the shape formed by the closing prices of the latest three bars with the shape of the corresponding three oscillator values. The pattern is identified…
This comparison explains how Lumibot and QuantConnect LEAN differ as algorithmic trading frameworks. Lumibot is presented as a Python-first library in which strategies use ordinary Python classes, broker and data adapters, and can combine deterministic rules…
DoubleEnsemble is an ensemble framework for financial prediction that addresses noisy training data and large feature sets. It combines two procedures: it uses each sample’s learning trajectory during training to identify influential examples and adjust…
The Relative Trend Index (RTI) is presented as a normalized indicator for gauging price position and trend strength. The described calculation estimates standard deviation over a short window, adds and subtracts that value from closing price to form upper…
This note explains R-squared as a measure of how much variation in a dependent series is explained by a fitted linear relationship. It presents an indicator that computes the squared correlation between a rolling sequence of prices and a time index,…
The document presents four functions for translating between chart pixel coordinates and market data: horizontal position to time, vertical position to price, and the inverse conversions. The time calculations use bar spacing, the chart’s first visible bar,…
This review of research on equity factor investing explains why low unconditional correlations between factors may not provide protection when macroeconomic conditions deteriorate. It describes selecting fast-moving state variables that reflect expectations…
Linda's Coil is a chart indicator for marking a current candlestick that is engulfed by a comparison candle. It has three settings: the lookback period, whether to accept only engulfment or also equality, and whether to compare with the immediately preceding…
This indicator uses the highs and lows of a candlestick series to identify directional price changes spanning a configurable number of bars. The stated default looks for a move across three candles, while the user can change that setting. It is presented as…
This document describes a MACD indicator built from Kaufman Adaptive Moving Averages (AMA). AMA adjusts its responsiveness to market conditions, aiming to reduce sensitivity to price noise while limiting lag in detecting trends. The indicator applies AMA…
The document describes a hybrid approach that combines ARMA-GARCH models with several neural network types to identify market patterns and predict the direction of shocks in defense equities and foreign exchange. ARMA-GARCH models capture linear time-series…
This document describes an Expert Advisor that exports MetaTrader 5 economic-calendar data to a CSV file for use by external research scripts, dashboards, bots, or spreadsheets. The terminal’s calendar can be accessed through MQL5, while the Python…
This note describes a daily liquidity signal built from intraday order book snapshots. It sums bid and ask volumes across the first five levels, then measures their normalized imbalance as the difference divided by the total. Separately, it calculates…