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.
86 documents
This long-only system combines a volume-weighted moving average with a smoothed RSI variant. It seeks entries when the close is above the VWMA and the smoothed RSI is above its threshold; exits occur when price falls below the average and RSI drops below its…
This workflow describes automated cryptocurrency perpetual-futures trading that delegates per-asset decisions to a large language model. An hourly process gathers market indicators, funding rates, positions, account state, and historical trading performance,…
This machine-learning example predicts whether each crypto futures asset’s next closing price will rise or not, then uses those binary predictions as portfolio weights. It builds candidate features from futures data, including a weighted moving-average trend…
This tutorial template uses a Ridge Classifier to predict whether a stock's next closing price will be higher or lower than its current close. Its main example loads Amazon data from the Nasdaq-100 universe, uses log closing prices as features, and reserves…
This proposed strategy identifies direction from a short and a long simple moving average, with crossovers used to open positions. RSI is presented as a stand-in for machine-learning confidence: the code scales and rounds the indicator, then applies a…
This example demonstrates using a persistent object store to cache historical data so later backtests can avoid repeating a costly history request. On the first run, it retrieves daily SPY closing prices, feeds them into price and exponential moving average…
This EUR/USD strategy outline combines linear regressions over short, medium, and long lookback windows. The visible code calculates each regression’s current fitted value, slope, R-squared, correlation, and projected value at a future horizon. Inputs…
This document describes an automated cryptocurrency trading system organized around two workflows: an hourly analysis and decision cycle, and a separate frequent monitor for take-profit and stop-loss conditions. The analysis cycle gathers multi-timeframe…
This strategy combines a fixed neural network with price and technical indicator inputs, RSI-based adaptive stops, and a Super Trend filter. The network processes changes in volume, Bollinger Band measures, RSI, and MACD histogram to produce an output used…
This script trains a random forest classifier to label short-term Bitcoin price changes as rising, falling, or relatively flat. It samples the latest price hourly, calculates consecutive percentage changes over a rolling window, and assigns directional…
This daily GLD strategy applies a fixed linear score to five features derived from price and volume, including short moving averages and interaction terms. The feature weights and normalization constants are embedded in the script, and the strategy holds a…
This automated strategy uses the Kronos financial time-series model to forecast a future price distribution. It compares the forecast median with the current price to decide whether the projected opportunity is large enough to trade, while the lower and…
This document outlines an automated system that uses a large language model to produce entry, hold, or close decisions for multiple cryptocurrency perpetual futures. Its pipeline collects market and account data, calculates indicators across short and long…
This algorithm trains a separate feedforward neural network for each of three exchange-traded funds using recent daily opening prices. The model uses one input layer with a hidden ReLU layer and a single output, and is trained to predict the next opening…
This strategy uses ALMA as its main trend measure and combines it with EMA levels, RSI, ADX, Bollinger Bands, and ATR. The described long setup requires price above EMA50 and ALMA9, RSI above 30, ADX above 30, price below the Bollinger upper band, and a…
This example trains a random forest classifier each trading day to predict whether a Chinese rubber futures contract will close higher or lower on the next trading day. Its features are three price-derived series labeled as SMA, WMA, and momentum measures,…
This template builds an asset-by-asset classifier ensemble for forecasting whether the next close will be higher than the current close. Its input features are a moving-average based trend measure, the stochastic oscillator, normalized true range, and a…
This research strategy compares four logistic regression approaches on the same candlestick stream: a fixed model, per-bar online updates, periodic retraining on a recent rolling window, and updates triggered by worsening prediction loss. It uses price and…
This strategy pairs an ATR-based SuperTrend with volatility regime classification to generate trend-following entries. It estimates high, medium, and low volatility bands from a rolling ATR range, then conditions direction changes on the assigned regime:…
This example demonstrates a basic text-driven trading rule using custom economic news headlines and NLTK tokenization. It downloads a headline dataset, tokenizes each day’s text, and checks for user-selected positive and negative words. For the SPY equity…
This automated crypto futures system ranks liquid USDT perpetual contracts using a composite score built from four EMA relationships across multiple timeframes. It selects the strongest positive and negative signals, then combines those scores with recent…
The document describes an automated cryptocurrency perpetual futures system organized around data collection, AI decisions, order execution, and position monitoring. It combines short-interval candles for entry timing with longer-interval trend context,…
This educational FMZ strategy expands recent completed OHLCV candles into fixed tabular features and asks TabFM to classify the next candle as up, down, or flat. Price features are expressed relative to the prior close, while volume is scaled against the…
The document outlines an automated process for turning a natural-language crypto factor idea into a calculated signal and evaluation report. A language model identifies the factor’s direction and data needs, generates a JavaScript function, and the workflow…