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
The document compares monthly Fama-French factor returns with values obtained by geometrically compounding the published daily series. It focuses on the market excess return and the SMB and HML factors, which are built from portfolios formed by size and…
The document explains how to align U.S. and Japanese market data when forecasting Japanese stock prices. Its central rule is to use only information available before the forecast target. Because the U.S. market closes while the next Japanese trading day has…
The document considers how duplicated or apparently similar securities can affect strategy research. Its responses distinguish true duplicates from instruments that merely share an issuer: Alphabet’s two share classes, for example, have different voting…
The discussion outlines mortgage valuation as a problem of projecting cash flows under changing interest rates, borrower behavior, and discounting. A proposed simulation approach generates possible rate paths, calculates amortizing payments, and discounts…
The document discusses combining forecasts from different trading methods, including sentiment measures, signal processing, and neural networks. One recommendation is to express forecasts in comparable units, such as expected returns or event probabilities,…
The document considers how to use parameters estimated from hourly Ornstein–Uhlenbeck (OU) data in a daily simulation. It offers two practical routes: aggregate the observed data to daily frequency and re-estimate the parameters, or simulate at hourly…
The document addresses how a sequential financial series becomes usable in a supervised machine-learning model. The key idea is to form examples by pairing information available at a decision time with the later outcome being predicted. Each such…
The document presents a question about implementing Tilley’s algorithm for pricing American options through simulated asset paths. The shown R approach generates risk-neutral paths, works backward through exercise dates, sorts paths by asset value, estimates…
The document addresses how to extend EURUSD history to dates before the euro existed, for quantitative analysis. One suggested proxy is the European Currency Unit (ECU), a pre-euro basket currency. Its constituent weights can be used to construct a…
The document discusses a three-class LSTM model for predicting financial trends from three time series. The question compares reported accuracy in a cited study with a lower result from an initial implementation and asks about model choices. The response…
The document discusses how to compare two correlated futures contracts whose prices and tick sizes differ when building a pairs strategy. One suggested approach sets both starting prices to one, then compounds each instrument’s percentage price changes to…
The document considers an equity backtest with fundamental and macroeconomic filters when some historical data for delisted stocks cannot be recovered. Its initial rule allows trades when a filter passes or its value is missing. The author worries that this…
A return regression can help estimate a strategy’s Sharpe ratio, but the regression alone does not determine it. The result depends on how the signal is converted into positions and, for portfolios, on the conditional covariance among assets. A tangency…
The document describes a workflow for forecasting financial time series with technical indicators and machine learning: calculate indicators, remove highly correlated features, lag inputs, train with a rolling or expanding window, and evaluate on held-out…
The document asks how to price American options when the underlying follows risk-neutral dynamics with volatility that varies with spot. The response identifies this as a special case of a local volatility model, where volatility may depend on time and spot,…
The discussion compares QuantLib pricing approaches for geometric and arithmetic Asian options. It reports that the finite-difference Black–Scholes Asian engine supports discrete arithmetic averaging, not continuous geometric averaging. A discrete geometric…
The document addresses an apparent mismatch between simulated discount factors in a Hull–White one-factor Monte Carlo model and the market discount curve at the initial time. Its central point is that the model’s no-arbitrage setup is intended to reproduce…
The discussion frames hedge selection as choosing weights on available instruments to reduce the combined portfolio’s PnL variance. A simple example shows that when matching instruments exist, the method should recover offsetting positions; with imperfect…
The discussion considers why a multi-period mean-variance portfolio plan can be time-inconsistent: a strategy that maximizes expected terminal wealth subject to a risk penalty today may no longer be optimal when reconsidered at a later date. It asks whether…
The document discusses how to feed time series with uneven observation density into a deep learning model, especially when each batch should retain historical context. It argues that the appropriate representation depends on the model and learning method.…
The document weighs alternatives to Value at Risk (VaR), especially Expected Shortfall and TailVaR, while discussing why VaR can give an incomplete picture of losses beyond a chosen threshold. One response notes growing attention to tail-sensitive measures…
The question considers estimating a portfolio Value-at-Risk from correlated asset returns assumed to follow a multivariate normal distribution. It proposes finding the portfolio's worst return on a probability ellipsoid determined by a chi-square cutoff,…
The document explains which stocks qualify for the annual Fama–French size and book-to-market portfolio sorts. Portfolios are formed at June end using market equity for size and book-to-market information based on prior fiscal-year book equity and December…
The document asks how to solve a three-dimensional backward Black–Scholes PDE with time- and state-dependent coefficients. Applying a Crank–Nicolson scheme across all three state variables at once creates a difficult coupled system. The response points to…