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.
16,761 documents
This Chinese-language discussion describes problems encountered when reusing a saved XGBoost model in a separate strategy and moving a feature-based strategy from research into simulated trading. Keeping visible strategy settings unchanged did not reproduce…
This study describes active risk parity as a portfolio built around a risk-parity allocation, with controlled deviations based on investment views. In its two-asset example, Chinese equities and corporate bonds are rebalanced using rolling covariance…
This indicator toolbox searches for regular and hidden divergences between price and seven technical series: MACD, RSI, MFI, OBV, CCI, a stochastic momentum measure, and PVT. It identifies recent local highs or lows over a lookback window, compares the price…
The README introduces a Python framework for backtesting strategies on OHLC or OHLCV price data. Its example defines a moving-average crossover system: it buys when the shorter simple moving average crosses above the longer one and sells when the reverse…
This meetup compilation surveys practical topics in quantitative investing, including factor design, machine learning, strategy validation, portfolio sizing, live execution, and sector rotation. It recommends building a factor library by collecting and…
The document describes a library that packages deep learning models for quantitative research, including feedforward networks, LSTMs, Transformers, TabNet, and TRA. Its quick-start workflow covers selecting a model, configuring an optimizer and loss…
This A-share strategy narrows a stock universe in three stages: rank industry or concept indices by short- and medium-term momentum, select stocks with a current limit-up and a recent prior limit-up, then confirm strength after the open before buying.…
This Lumibot example demonstrates a simple futures holding strategy. It configures a US futures market, checks for the first trading iteration, then creates and submits a buy-to-open order for one futures contract with a specified symbol and expiration date.…
This note describes a reproduction of a sell-side factor based on overnight stock returns, intended to capture information advantages held by informed traders. It says the research defines the factor in stages, from a traditional overnight-return measure to…
This Chinese-language research summary examines how to use volatile price and volume factors in industry rotation strategies. It proposes using external market variables to forecast a trend factor’s returns or information coefficient, then timing exposure…
The document describes a four-hour EURUSD crossover system using exponential moving averages with periods 2 and 35. A cross above triggers a buy and closes a sell; a cross below triggers a sell and closes a buy. It also specifies stop-loss, take-profit, and…
This guide explains how historical data drives a backtest and how venue book configuration affects simulated matching. It ranks inputs by detail: market-by-order depth, market-by-price depth, top-of-book quotes, trade ticks, and interval bars. More detailed…
This Chinese-language post describes an A-share stock screen combining price range, convertible-bond, and early-session price conditions. The stated rules require an amplitude above a threshold, a nonempty outstanding convertible-bond name, and a 9:25 price…
This stock-selection proposal combines a positive MACD reading with upward-moving averages and positive returns. Its final description extends the return condition to a positive five-day average, while the sample formulas and Python example use related but…
This discussion distinguishes two approaches to Chinese A-share limit-up trading: a daily model that forecasts which stocks may hit the limit the next day, and an intraday model that uses current market information to identify candidates for the remainder of…
CloseProfit v2 is described as an expert advisor that closes positions when an aggregate profit or loss reaches a configured threshold. It also removes pending orders when that trigger is reached. A scope setting determines whether the advisor operates…
The document explains how FreqAI trains trading agents through reinforcement learning. An agent processes historical candles and chooses among actions such as entering or exiting long and short positions. A custom reward function scores its decisions, while…
The post describes a short routine for closing all positions in a minute-based backtest. Its stated setup is to remove the existing strategy, run the routine in the minute backtest environment, and bind the original capital account. The routine records a…
The study revisits Bitcoin strategies that buy when price reaches a recent high (MAX) or low (MIN), using daily data from late 2015 through August 2024. It examines lookback windows from 10 to 50 days, compares in-sample results with an out-of-sample period…
This indicator transforms a stochastic oscillator through two exponential smoothing stages and then maps the result into a bounded signal value. The supplied settings use a 130-period stochastic, a smoothing period of 10, and thresholds at 0.85 and -0.8. A…
This Backtrader example demonstrates execution modeling with volume-aware order fillers. Its strategy calculates a buy size as a configurable percentage of the current bar’s reported volume, then closes the position on a later eligible opportunity. A…
This brief support exchange addresses whether a strategy’s data is standardized anew each day during simulated trading. The response says that when the strategy uses parameters bound for live trading, simulated trading standardizes that day’s data. A…
This Backtrader example combines a moving-average crossover entry signal with three buy limit orders placed at progressively lower prices after an upward crossover. The orders can be linked using one-cancels-other behavior, so execution of a linked order…
This overview explains XGBoost regression as an ensemble of decision trees built sequentially to reduce a loss. Each new tree uses gradient information from prior predictions, while complexity controls and regularization are intended to limit overfitting.…