Research preview

The market’s view of a possible October rate hike has collapsed to a 24 % probability. This sharp shift illustrates how new data and model choices can dramatically alter expectations. In this article we connect that market move to core ideas from high‑dimensional machine learning, showing how bias, variance, regularisation and debiasing affect macro‑forecasting and trading decisions. From Raw Signals to a Probability Estimate A typical macro‑forecast model starts with a large set of predictors: inflation reports, labour market indicators, commodity prices, forward curves and sentiment measures. When the number of variables grows comparable to the number of observations, ordinary least‑squares estimates become unstable....