Home  ›  Signals  ›  Z32620 LC Variance-Ratio Predictability Signal

· Analytic

Z32620 LC Variance-Ratio Predictability Signal

Rolling Lo-MacKinlay variance-ratio statistic on CME Live Cattle, signed by recent 20-day direction — Grade C+ market-microstructure predictability overlay.

At a glance

As of Oct 30, 2026. Updated nightly.

OBJECTIVE

The Lo-MacKinlay variance-ratio (VR) test is a classical measure of market predictability. VR(2) compares the ratio of 2-period return variance to twice the 1-period return variance. In a random-walk market, VR equals 1; in a momentum-rich market, VR exceeds 1; in a mean-reverting market, VR falls below 1. The departure from 1 quantifies serial predictability in returns. This signal computes a rolling 252-day Lo-MacKinlay VR(2) statistic on CME Live Cattle, subtracts 1, z-scores, and multiplies by the sign of the 20-day cumulative return. The result captures DIRECTIONAL predictability — when VR deviates significantly from 1 AND the market has been trending, the signal amplifies in the direction of the trend. This operationalises the empirical finding from Truth LC_T051 (Chow-Denning martingale-difference-sequence test rejected for LC) as an actionable directional signal. Z-score on 504-day rolling window. Tanh-bounded to [-1, +1]. State thresholds at plus or minus 0.2 (moderate) and plus or minus 0.5 (extreme). Daily release cadence. The memorable finding: Grade C+ microstructure predictability overlay. Daily Spearman IC against LC fwd60d is minus 0.038 at p-value 3.6e-4 across 9,049 observations — marginal. Non-overlap independence test at p-value 0.11 — marginal (close to the 0.1 significance threshold). Phase -1 (Negative Momentum-Predictability) delivers +1.60% fwd60d (+0.58 pp edge, hit 61%). Phase +1 (Positive Momentum-Predictability) delivers +0.67% fwd60d (-0.35 pp edge, hit 56%). Phase 0 delivers +0.51% (-0.51 pp below baseline). Current reading: EXTREME_POS at composite +0.661 — positive VR-directional-predictability regime. The empirical pattern shows Phase +1 modestly bullish vs Phase -1 strongly bullish; the regime is constructive but not high-conviction. Read on for historical anchors and how to pair with the LC fundamental stack.

LIVE CHART

STATE-CONDITIONAL EDGE

THE TAKEAWAY

THE REGIME MAP

Computed from 9,049 trading days with observable LC fwd60d. Baseline LC fwd60d (unconditional): plus 1.02 percent mean. Grade C+ market-microstructure overlay.

The historical pattern

THE TRACK RECORD

Scroll inside the table for the full history. Forward 60-day LC is measured from each run’s start date; a reader cannot trade off a start-date they only identify in hindsight, so this is a transparency log, not an alpha claim. Asterisked (*) returns are partial. The run is less than 60 trading days old. The full 60-day comparable fills in as more data arrives.

The signal computes a rolling 252-day Lo-MacKinlay variance-ratio VR(2) statistic on CME Live Cattle, subtracts 1 (to center on random-walk expectation), z-scores, and multiplies by the sign of the 20-day cumulative return (to make it directional). Operationalises Truth LC_T051/T046/T052 microstructure findings. Z-score on 504-day rolling window with 60-day minimum sample. Tanh-bounded to [-1, +1]. State thresholds plus or minus 0.2 (moderate) and plus or minus 0.5 (extreme). Daily release.

Not investment advice · The analysis on this page is informational and educational. Signal readings, historical case figures, and empirical statistics are derived from publicly available data and are provided without warranty. Hypothetical performance results have inherent limitations, including the benefit of hindsight. Readers who intend to trade should consult a licensed investment professional and see the research brief for the full disclaimer.

Last refreshed Nov 01, 2026 02:13 PM ET