advanced practice · coding problem
Interpret an ACF plot
Daily returns show ACF lag-1 ≈ 0.02, all within confidence band. Prices show ACF lag-1 ≈ 0.98. Interpret each — which series is closer to a random walk in returns?
The mechanism
Stationarity: distribution doesn't drift. Autocorrelation: how today relates to yesterday. Random-walk returns have ~0 autocorr; momentum must prove structure exists.
Expected evidence
Return ACF ≈0 → little linear memory (efficient-market-ish). Price ACF high → non-stationary levels; difference of prices (returns) matters.
Try this drill in the practice workspace →
Hints when you need them
- ACF on levels confounds trend with memory.
- Returns stationarity is the usual assumption.
Watch for these mistakes
- Tests on non-stationary levels
- Trend mistaken for momentum
- Ignoring regime shifts