What is cross-sectional momentum?

Rank a universe by trailing 3-to-12-month return, buy the recent winners and (in the academic construction) short the losers. It profits from relative continuation and was documented by Jegadeesh and Titman in 1993.

What is time-series momentum?

Each asset is judged on its own trailing return alone: hold it long if it has been rising, go flat or short if falling. Moskowitz, Ooi and Pedersen documented it across 58 futures markets in 2012, and the trend-following record runs back more than a century.

Which type of momentum is better?

They carry different risks rather than different quality. Cross-sectional momentum is a relative bet that fails hardest in crashes at regime turns; time-series momentum takes directional exposure and embeds a market-timing component, which is where its crisis performance and its net-long-tilt debate both come from.

Can a strategy use both kinds of momentum?

Yes, and Shishin does: the regime gate is time-series momentum applied to the whole market, the daily ranked board is cross-sectional selection within the day, and the breakout trigger itself is an absolute condition each stock must meet on its own.

What lookback period does momentum use?

The classic academic windows are 3 to 12 months of trailing return with roughly monthly rebalancing; trend-following commonly uses 12 months. At horizons under about a month the effect flips toward short-term reversal.

Time-series vs cross-sectional momentum: two trades

This article summarises published academic research. Historical effects are not a guarantee of future returns, and nothing here is investment advice.

“Momentum” is one word for two different trades. Cross-sectional momentum buys the stocks that beat their peers and shorts the ones that lagged. Time-series momentum asks a simpler question of each asset alone: has it been going up? The two are cousins, not twins, and knowing which one a strategy actually runs tells you what risk it is carrying.

The two definitions

Cross-sectional (or relative) momentum is the 1993 Jegadeesh and Titman construction: rank the universe by trailing 3-to-12-month return, go long the top decile, short the bottom, and hold for a few months. In their sample the winners-minus-losers spread earned on the order of one percent per month before costs, and the finding has survived three decades of replication attempts, which is rare enough to be remarkable.

Time-series (or absolute) momentum is the 2012 Moskowitz, Ooi and Pedersen construction: for each instrument, look only at its own trailing 12-month return, hold it long if that return is positive, short (or flat) if negative. Across 58 futures markets spanning equities, bonds, currencies and commodities, that rule was profitable in essentially every asset class, and the broader trend-following literature has pushed the evidence back more than a century.

Why the distinction matters

The cross-sectional trade is a relative bet. It can profit in a falling market as long as its longs fall less than its shorts, and it carries little direct market exposure by construction. Its characteristic failure is the momentum crash, when the relative ordering inverts violently at a regime turn.

The time-series trade is a directional bet with a built-in market-timing component: when most assets have been falling, it de-risks toward flat or short across the board. That timing component is where much of its celebrated crisis performance comes from, and it is also the heart of the academic debate about it. Later work argued that a meaningful share of time-series momentum’s excess return traces to its long-run net-long tilt and to the persistence of the market factor itself, rather than to per-asset magic. The honest reading is that both effects are real, related through the autocorrelation of returns, and neither is a free lunch.

A breakout system is both at once

A rules-based breakout board does not have to choose sides, and ours does not. The regime gate is time-series momentum applied to the market as a whole: it reads breadth and trend, deploys an engine only when the environment has been improving, and stands the book down to cash when it has not. The daily board is cross-sectional: on the days an engine does run, it scans ~5,800 US names and ranks the morning’s setups against each other, so capital goes to the relative leaders. And the entry trigger itself, a close above a defined pivot, is an absolute-momentum condition at the single-stock level: a stock qualifies by its own behaviour, not merely by beating peers (the craft version of this is the breakout setup).

Stacking the two is deliberate. The time-series layer decides WHETHER to be exposed, which is where the drawdown control lives; the cross-sectional layer decides WHERE, which is where the selection edge lives. Most of the academic fights about which momentum is “better” dissolve once you stop treating them as rivals and start treating them as different controls on the same underlying phenomenon: returns trend, in the cross-section and in time.

Sources & further reading

Jegadeesh, N. & Titman, S. (1993). “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency.” Journal of Finance, 48(1), 65 to 91.
Moskowitz, T. J., Ooi, Y. H. & Pedersen, L. H. (2012). “Time Series Momentum.” Journal of Financial Economics, 104(2), 228 to 250.
Hurst, B., Ooi, Y. H. & Pedersen, L. H. (2017). “A Century of Evidence on Trend-Following Investing.” Journal of Portfolio Management, 44(1), 15 to 29.
Goyal, A. & Jegadeesh, N. (2018). “Cross-Sectional and Time-Series Tests of Return Predictability: What Is the Difference?” Review of Financial Studies, 31(5), 1784 to 1824.

Shishin Research