Cross-sectional momentum
Rank markets and hold the strongest · 8 min
- Market A
- Market B
- Market C
- Market D
- Market E
Ranked by return over the window, from best to worst.
All eight markets in a portfolio fall 20% in a shared crash. A long-only rule that holds the two best-ranked markets…
Cross-sectional momentum compares markets with each other. Every month (or week), rank all markets by their return over the lookback and hold the top few; a long-short version also shorts the bottom few. It asks "which is strongest?", not "is it going up?".
Because it is relative, a long-only version still falls in a shared selloff. Rotations are frequent, so turnover and costs matter. And when many traders hold the same winners, the reversal after a crash can be violent, as last year's losers rebound fastest: the momentum crash. Comparing against an equal-weight portfolio of the same markets, not against cash, is the fair test.
- Top-ranked markets
- All eight, equal weight
- Crisis days (shared selloff)
Long only: a cross-sectional rule still holds markets in a selloff, because it ranks them against each other, not against zero.
Generated markets, not real data. They show how a rule depends on the market it meets; they do not predict how real markets behave.
Ranking five markets
- 6-month returns
- A +12%, B −4%, C +20%, D +3%, E −10%
- Ranked
- C, A, D, B, E
- Long-only top 2
- 50% C, 50% A
- Long-short top/bottom 1
- long C, short E
- Next month C, D lead
- sell A, buy D
Why can momentum portfolios suffer sharp losses right after a market crash?
3-month returns: P +8%, Q +15%, R −2%, S +11%, T +1%. Which two does a top-2 rule hold?
Last month the rule held C and A. This month the ranking is C, D, A, B, E. What trades does a top-2 rule make?
Missed questions come back for review after one day, then three, then a week.
A market closed at 50 twelve months ago and at 55 today. What is its 12-month return (in %)? A time-series momentum rule holds the sign of that return.
Hint: Return = today ÷ then − 1.
Three years of a fresh generated world are on screen; the next two years do not exist yet. Tune the rule on what you can see, commit to two predictions, then generate the future.
- Your rule so far
Over the next two years, after costs, this rule will…
Will the deepest drawdown after the split be worse than before it?
Generated markets, not real data. They show how a rule depends on the market it meets; they do not predict how real markets behave.