Cascade Tracker
The five-domino sequence maps how stress propagates through the asset hierarchy. The order matters: when crypto decouples from tech (domino 1), watch whether tech-to-broad follows.
Breakdown Cascade
How to Read the Pairs

LOCKED

p-value < 0.05. The relationship is statistically intact. Spread deviations mean-revert within the estimated half-life. Business as usual.

STRETCHING

p-value 0.05–0.10. The relationship is under strain. Spread Z-scores may be elevated. Worth monitoring — could revert or break.

BREAKING

p-value ≥ 0.10. The statistical link has dissolved over the 60-day window. The market is repricing the structural assumption behind this pair.

Pair Health
Each card shows the Engle-Granger cointegration test result, spread dynamics, and mean-reversion characteristics. The research question frames why this pair matters. Full methodology →

Half-life tells you how many days the spread takes to revert halfway to its mean — a short half-life (under 15 days) means the relationship snaps back fast. The Hurst exponent below 0.5 confirms mean-reversion; above 0.5 signals trending (breakdown). Variance ratio near 1.0 means random walk; well below 1.0 confirms mean-reversion. The dashed long-run row beneath each card is the establishment layer: Johansen trace, Engle-Granger in both directions on one- and three-year windows, ADF of the log ratio and weekly-frequency EG — it answers whether a long-run relationship exists at all, while the 60-day row above answers how the spread is behaving right now. The two can honestly disagree.

0
Locked
0
Stretching
7
Breaking
← Back to Temperature

What the Cointegration Monitor measures

Two assets are cointegrated when a stable long-run relationship ties them together — the pair can drift apart day to day, but the spread between them mean-reverts. The monitor tracks Closelook's core cross-asset pairs — Bitcoin against gold (the two system-hedges), Bitcoin against the Nasdaq (risk-asset coupling), and the joint quadrant of both — and scores each pair on three questions: is the long-run relationship currently intact, how stretched is the spread against its own history, and in which direction is the tension pointing. A pair that leaves its cointegration regime is information in itself — it marks a structural change in how the two assets are being priced, not just a noisy week.

We read it alongside the Money Temperature composite: temperature says how hot each instrument runs on its own tape, cointegration says whether the relationships BETWEEN them still hold. Regime shifts show up here first when they are cross-asset in nature — the Cointegration 101 walks through the construction, the test statistics and the failure modes. Live pair states load in the dashboard above and refresh with each update cycle. closelook.net is an investment research diary — readings are our own measurements, not investment advice.

Read next

FAQ · from the current data · as of 2026-08-29

Quick answers

What does the Cointegration Monitor track?

Six cross-asset pairs under rolling 60-day Engle-Granger surveillance — including Bitcoin against gold and Bitcoin against the Nasdaq — scored as LOCKED (relationship intact), STRETCHING (under strain) or BREAKING (statistical link dissolved).

How many pairs are currently locked, stretching or breaking?

As of 2026-08-29, 0 of 6 pairs are LOCKED, 0 STRETCHING and 7 BREAKING.

Which pairs are currently breaking?

As of 2026-08-29, the pair(s) in BREAKING status — where the 60-day cointegration relationship has statistically dissolved — are SPY/GLD, QQQ/SPY, BTC-USD/QQQ, GLD/UUP, TLT/SPY, VEU/SPY, IGV/SMH.

What do LOCKED, STRETCHING and BREAKING mean?

LOCKED = Engle-Granger p-value under 0.05, the relationship is statistically intact and spread deviations mean-revert. STRETCHING = p-value 0.05–0.10, the relationship is under strain. BREAKING = p-value 0.10 or above, the statistical link has dissolved over the 60-day window. Each card also carries a separate LONG-RUN verdict — Johansen trace plus Engle-Granger in both regression directions on rolling one-year and three-year windows, an ADF test of the log price ratio, and a weekly-frequency Engle-Granger. ESTABLISHED means a long-run relationship exists (Johansen finds at least one cointegrating relation at 95% and the three-year test rejects at 5%); NOT ESTABLISHED means the pair co-moves without a statistical long-run anchor, and the 60-day row should be read as a spread-behaviour gauge rather than a cointegration claim.