Introducing Spatio-Temporal Momentum for after-work quants

Takeaway. Momentum need not live only in an asset’s own past—spatio-temporal models also use neighbours in the cross-section (or a graph) so “who moves with whom” informs the signal.

Spatio-Temporal Momentum (Wee Chung Gan / related authorship on arXiv as Tan, Roberts, Zohren line—2023) extends momentum thinking into time × space features for multi-asset settings.

Fit for learners: Momentum that mixes time (own past) and space (cross-sectional / network neighbours)—ML-friendly multi-asset features.

This Hub note is a plain-language intro. Not investment advice. Pair with slow momentum / fast reversion and TSMOM.

The problem

Classic TSMOM looks at each series alone; classic cross-sectional momentum ranks everyone but still uses simple past returns. Richer dependence—sectors, correlations, learned graphs—is easy to ignore in a first notebook.

The idea / core intuition

• Combine temporal momentum (own history) with spatial structure (other assets / neighbours).
• Use modern ML-friendly architectures or feature designs that share information across the panel (details in the PDF).
• Goal: better multi-asset signal design than isolated lookbacks—still subject to costs and overfitting.
• Read the paper for exact model family and experiments; this Hub note only sketches the intuition.

Why after-work Python learners care

Once you can build a returns panel, “space” is just other columns. Spatio-temporal momentum is motivation to move from one ticker to DataFrame panels and simple neighbour features.

Tiny evening exercise: pick 5 correlated toy series; for each day, compare each name’s 20-day return to the average of the other four; print a few rows—learning only.

One-evening reading checklist

1. Define “spatial” vs “temporal” in this momentum context.
2. Contrast with pure TSMOM and pure cross-sectional sorts.
3. Note one modelling choice from the paper’s abstract/intro.
4. Name one leakage risk when using contemporaneous neighbours.

Download PDF

Download PDF

Landing: arXiv abs

Important disclaimer

Not investment advice. This Hub note is educational only. Spatio-temporal momentum research is not a ready-made trading bot.

What this is not

• Not CEF / subsidy framing
• Not Quant or Macro course CTA
• Not a graph-NN product pitch

Next step

Single course CTA: SoloQuant Academy’s Introduction to Python Programming—get solid on tables and loops first; panel momentum code comes after.