Time‑Series Momentum Guide
- A 23‑page breakdown details how top hedge funds use time‑series momentum (TSM) to generate returns across asset classes. - The paper includes practical Python implementation ideas tailored to equities and derivatives strategies. - It serves as a blueprint for testing momentum persistence, transaction‑cost sensitivity, and factor‑timing robustness. (x.com)
Time-series momentum is a simple trading rule with a hedge-fund pedigree: if an asset has been rising over recent months, buy it; if it has been falling, sell or short it. A new explainer making the rounds packages that idea into a 23-page guide with code-oriented steps for testing it in stocks and derivatives. (docs.lhpedersen.com) The core academic reference is a 2012 paper by Tobias Moskowitz, Yao Hua Ooi, and Lasse Heje Pedersen. It examined 58 liquid futures and forward contracts across equity indexes, currencies, commodities, and sovereign bonds, using more than 25 years of data. (aqr.com) That paper found an asset’s own past 12-month excess return predicted its next-month return, with persistence lasting about one year before partially reversing over longer horizons. The authors also reported positive 12-month momentum profits for every contract in their sample, not just the average portfolio. (docs.lhpedersen.com) This is not the same as the better-known stock momentum trade that ranks winners against losers. Time-series momentum looks only at each asset’s own history, asking whether its recent trend is positive or negative. (aqr.com) That distinction matters in practice because the strategy is usually built across many markets at once, then scaled by volatility so one noisy asset does not dominate the book. AQR’s original dataset describes monthly long-short factors based on a 12-month signal and a 1-month holding period across the four major futures groups. (aqr.com) The newer guide’s appeal is less the signal itself than the implementation checklist around it. Quant developers typically need to decide the lookback window, rebalance frequency, volatility target, transaction-cost model, and whether they are trading cash equities, exchange-traded funds, or futures proxies. (github.com) Python has become the default language for that work because it can handle data cleaning, signal generation, backtests, and portfolio analytics in one stack. One public replication repository updated this week includes modules for returns, volatility, performance analysis, plots, and notebooks for futures, exchange-traded fund replication, and GARCH volatility modeling. (github.com) Researchers have also pushed the idea beyond raw asset prices and into “factor momentum,” which tests whether styles such as value, quality, or low volatility keep leading or lagging. A 2019 Journal of Portfolio Management article by Tarun Gupta and Bryan Kelly said a dataset of 65 characteristics showed robust momentum behavior among common equity factors as well. (aqr.com) That is why modern explainers focus so heavily on robustness checks. A strategy that looks strong before trading costs, weak after slippage, or dependent on one market regime is not the same thing as a durable signal. (aqr.com) The thread running through the research is straightforward: trend-following can be described with a few lines of code, but testing whether it survives real markets takes far more work. The latest guide is being shared as a map for that second part. (x.com)
