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.
195 documents
This article describes techniques for reducing overfitting in feed-forward neural networks used to forecast market direction. It outlines L1 and L2 regularization, which penalize large model weights, and dropout, which randomly disables units during…
The article explains why backtests are needed to assess trading rules and why simulated performance is only a guide to live results. A useful simulation should reflect the intended market and broker conditions, use data at an appropriate level of detail, and…
This article frames alpha as an opportunity created when traders transact at disadvantageous prices, and emphasizes understanding why they do so. Reasons include limited information or behavioral biases, binding risk or mandate constraints, and non-profit…
This essay argues that traders should begin with practical problems and seek reading when they encounter a knowledge gap, rather than treating book consumption as a substitute for research or trading. It recommends focusing on developing an edge and managing…
This tutorial introduces dplyr workflows for manipulating daily stock-price observations. It explains how to filter rows for one or several tickers, reorder observations by date or trading volume, and select specific columns. It also demonstrates chaining…
The article introduces rsims, an R package for fast portfolio backtests that emphasizes translating target weights into trades while accounting for costs and constraints. It describes a threshold rule: trade toward a target only when the current weight moves…
The article outlines three practical sources of trading hypotheses. Traders can learn from other market participants who appear to have profitable approaches, while adapting ideas to smaller niches or constraints that may not suit large asset managers. It…
The article explains options as expiring bets whose fair value is the probability-weighted payoff. It illustrates the idea with a soccer match modeled as separate Poisson goal processes for the home and away teams. Expected goals imply probabilities for home…
Tesla’s addition to the S&P 500 creates predictable index-tracking demand, but the article argues that this flow may already be reflected in prices by the time a trade seems obvious. It reviews research on index additions: earlier work found most excess…
The document argues that binary rules, such as taking a position based only on whether price is above a moving average, discard information and conceal how signal strength relates to future returns. For a crypto trend example, it replaces the on/off…
The document explores whether currency prices can form stationary spreads suitable for mean-reversion analysis. It estimates a two-currency spread using ordinary least squares, then tests the residual with an augmented Dickey-Fuller procedure. It also…
A no-trade region places a buffer around a strategy’s target position. The portfolio is left alone while its current holding remains inside the buffer, and a trade is made only after it moves beyond the boundary. With minimum commissions, the example rule…
Rolling estimates such as 30-day volatility share most of their underlying observations from one day to the next. A naive comparison of adjacent estimates can therefore appear highly persistent even when much of that relationship is mechanically caused by…
This article curates books, papers, and course materials that the author found useful for learning algorithmic and quantitative trading. The recommendations are grouped into practical trading, foundational statistics and time series, machine learning,…
This course page presents a framework for systematic trading centered on identifying a plausible edge before building and evaluating a strategy. It argues that a strong backtest alone does not establish that a strategy is sound, and recommends formulating a…
This article uses simulated cryptocurrency price paths to explore how often a leveraged trend strategy might need rebalancing to manage drawdowns. The author builds a geometric Brownian motion simulator with autocorrelated returns and random jumps, using a…
The document explains how UVXY’s daily leverage target and maturity maintenance lead to recurring portfolio rebalancing, and uses spreadsheet models to examine two trades: shorting UVXY with periodic rebalancing, and shorting a basket of UVXY and an inverse…
The document outlines an experiment for studying how training-window length and predicted class-probability thresholds affect a financial prediction strategy. It constructs directional labels from returns and uses lagged returns and volatility measures as…
The document offers practical guidelines for trading equity options, emphasizing that the many contracts available on one underlying tend to have thinner liquidity and wider spreads than the underlying stock. It recommends using options when the trading…
The document presents statistical arbitrage as a broader portfolio problem than trading matched pairs. It ranks assets by expected cheapness or expensiveness, then builds long and short positions intended to capture relative value convergence while…
This article advises traders with small accounts to begin with comparatively simple, forgiving strategies that support consistent process-building and skill development. It cautions that niche, high-capacity-constrained opportunities may offer attractive…
This essay argues that systematic traders should begin with market observation and an explanation of why a possible edge exists, rather than searching broadly across indicators and parameters for a profitable backtest. Repeated experimentation can produce…
This analysis revisits whether SPY’s returns accrue mainly overnight or during regular trading hours. It calculates intraday returns from each session’s open to close and overnight returns from the prior close to the next open, then compares their cumulative…
This tutorial combines a Kalman filter written in R with a simple pairs trading system in Zorro. The filter updates a hedge ratio as new prices arrive, estimates the spread prediction error, and calculates its uncertainty. The trading logic uses that…