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.
5,018 documents
The document considers whether a market-timing model can identify both favorable regimes and conditions where its forecasts should be distrusted. It suggests that variables with weak direct forecasting power may still help distinguish regimes, using VIX as…
The document considers how to classify mutual funds into low, medium, and high volatility groups without choosing fixed volatility thresholds in advance. The answer proposes unsupervised clustering with k-means: provide the funds’ volatility measurements and…
The document discusses using Empirical Mode Decomposition (EMD) of EUR/USD prices as input to a machine learning model. It addresses a discrepancy between promising test-period results and poor predictions on the latest observation. The proposed explanation…
The document asks how to move from evaluating a machine-learning model on historical test data to producing forecasts from newly received broker data. The author describes training several model types on price and fundamental inputs, currently using a random…
The document asks how to compare two models that produce probability distributions for future returns. One proposed method is to add the probabilities assigned to observed outcomes, but the response instead evaluates the joint likelihood of the realized…
The document raises research design questions for comparing two sentiment datasets as predictors of volatility. It considers whether to include both GARCH and HAR models to broaden the analysis, and how to use sentiment measured continuously when the HAR…
The document discusses why Python is widely used for quantitative finance, while arguing that no single language dominates every part of financial technology. Its practical advantages include familiar syntax, a large developer pool, extensive libraries for…
The question describes a balanced financial time-series classification task in which several machine-learning algorithms perform well under time-ordered cross-validation but fail on a later test set. The response proposes that the underlying relationship may…
The document explores why a significant deterministic time trend can coexist with a negative, significant coefficient on the lagged level in a regression of first differences. The response considers a series made up of a linear trend plus a stationary error:…
The note addresses whether reinforcement learning is meaningful for trading models that assume orders do not affect order-book dynamics. Its answer is that actions need not change the state transition process: a decision process can still specify available…
The discussion asks how to infer a stochastic process from a long observed series and questions whether a histogram can identify its dynamics. The response recommends beginning with discrete-time time-series analysis rather than trying to recover an…
The document considers how to model hedge fund returns for simulation, noting that return moments alone do not specify a distribution. One proposed approach represents returns as exposures to a broad set of tradable ETF factors. The cited research uses…
The document outlines how spline methods can fit functions defined over separate regions while enforcing continuity and differentiability at their boundaries. It first describes univariate splines, where each interval has a polynomial segment and neighboring…
The document asks whether M&A research can support more quantitative methods than the small-sample OLS and fixed-effects studies the questioner has encountered. It points to two possible directions: forecasting M&A activity with a predictor model, and…
The answer considers whether banks and asset managers are likely to develop trading algorithms in-house or rely on specialist firms. It argues that either path may be difficult within traditional institutions, citing gaps in management’s quantitative…
The document contrasts quantitative investing with quantitative trading, emphasizing the larger capital base and capacity requirements of asset management. Strategies that exploit small arbitrage opportunities may not scale to the size of investment…
The document describes an attempt to find directional patterns in second-by-second market depth data using dimensionality-reduction and classification methods. A UMAP visualization appeared to separate observations labeled by subsequent price direction, but…
The document weighs R and Python for quantitative finance tasks, including time-series data handling, statistical modeling, mathematical computation, visualization, derivatives pricing, system integration, and deep learning. Its answers describe different…
The discussion addresses volatility forecasting for financial stocks observed at five-minute intervals and notes that standard ARCH and GARCH approaches may not work as well at this frequency. One response reports that, in a thesis and related literature,…
The document describes an effort to reproduce parameter estimates from continuous-time stochastic models of US inflation. It lists reported estimates for log-normal models with constant, local, and mean-reverting Vasicek drift, using historical CPI data and…
The document explains an apparent asymmetry in simulated currency prices generated from a GARCH model. The model is fitted to log returns, which may have a symmetric innovation distribution, but the simulation is transformed back into price levels by…
The document discusses a machine-learning return forecast evaluated with walk-forward validation while varying training-window length, prediction horizon, step size, and embargo. The reported out-of-sample results deteriorate as the step or embargo…
The discussion considers whether an economics researcher who prices options and calibrates models should learn C++ to speed computationally intensive work, and how to connect it with higher-level languages. One respondent recommends a derivatives-pricing…
The document describes how to distinguish order events in full-depth equity market data, particularly when preparing intraday records for analysis or machine learning. It maps message labels to their likely roles: adds on the bid or ask represent limit…