EventCOT commercial percentile (GBPUSD) invalidation
Level77.21
TimeNext weekly observation
Why it matters
Crossing this published level invalidates the current market lean.
Thesis effectDaily Thesis
SourceCFTC Legacy Futures-Only
T01TodayState vector · Aug 22, 2026
What matters in the market today?
Start with the short version. Then inspect the anomaly, signal votes, historical analogs, and the level that would invalidate the read.
- Anomaly
- -2.96σ
- Analogs
- n=50
- Falsifier
- z=0.00
The market, in one view.
Start with posture, disagreement, history, and the exact level that breaks the read. The full evidence and calculation remain below.
- Market posture
- More defensiveThe largest unusual signal currently reads risk-off, while supportive votes remain visible.
- Most unusual signal
- COT commercial percentile (GBPUSD) · -2.96σPercentile 0.0 · n=52
- Evidence balance
- 1 support · 5 oppose3 no vote under the published threshold.
- Key risk
- Credit spread proxy · +2.54σThe strongest standardized vote against the current posture.
- Historical context
- 66% positive · n=50Median SPX return after 21 sessions: +0.86%. Context, not a forecast.
- Falsification condition
- COT commercial percentile (GBPUSD) ≥ 77.21Evaluated on the next weekly observation.
- Latest complete observation
- Aug 20, 2026Newest constituent: Aug 22, 2026.
- Anomaly
- Far from recent history
- Confluence
- How many signals agree
- Analog
- A similar past setup
- Falsifier
- The level that breaks the read
Since the prior thesis
What changed?
10 observable changes since the prior completed thesis.
- directionMarket lean changed
Risk-on became risk-off.
- anomalyMost unusual signal changed
HYG/LQD was replaced by COT commercial percentile (GBPUSD).
- reversal5s30s reversed
The component changed from contradictory to supportive.
- reversalRegime composite reversed
The component changed from supportive to contradictory.
- reversalBreadth reversed
The component changed from supportive to contradictory.
- reversalBroad dollar reversed
The component changed from supportive to contradictory.
- reversalVIX term structure slope reversed
The component changed from supportive to contradictory.
- deactivation2s10s lost its vote
The component moved below the published vote threshold or lost sufficient coverage.
- agreementEvidence balance changed
4 agree, 2 disagree, 3 no vote became 1 agree, 5 disagree, 3 no vote.
- coverageComplete observation window advanced
2026-08-10 became 2026-08-20.
Latest complete cross-asset observationAug 20, 2026
Newer constituent coverage through Aug 22, 2026. Published Aug 22, 2026. Captured Aug 22, 5:59 PM CDT.
COT commercial percentile (GBPUSD) printed 40.0 percentile, 2.96 standard deviations below its prior 52-observation mean, the largest absolute state-vector z-score.
- Direction
- Risk-off
- Percentile
- 0.0 · n=52
- Observation date
- Aug 18, 2026
02 · Confluence
1 components agree with the risk-off direction, 5 disagree, and 3 have no vote.
Transform each mapped component z-score into risk direction and count absolute component z-scores below 0.50 as no vote. Cross-asset COT has no single risk-direction mapping and receives no vote. Formula · Source
Agree
n=1- 5s30s-1.02σ5s30s Treasury curve · n=1 dimensions
Disagree
n=5- Credit spread proxy+2.54σHYG/LQD · n=1 dimensions
- Regime composite+1.67σRegime composite · n=1 dimensions
- Breadth+1.08σRSP/SPY breadth · n=1 dimensions
- Broad dollar+0.85σBroad dollar index · n=1 dimensions
- VIX term structure slope+0.59σVIX3M/VIX slope · n=1 dimensions
No vote
n=3- 2s10s-0.31σ2s10s Treasury curve · n=1 dimensions
- Net GEX·No standardized history · n=0 dimensions
- Max pain distance·No standardized history · n=0 dimensions
03 · Analog
Across n=50 analogs, SPX +21-session returns had a +0.86% median, 66.0% positive hit rate, and -3.19% worst decile; ES had a +0.79% median, 60.0% positive hit rate, and -3.41% worst decile.
Derived artifact · n=50 · Formula · Cosine distance over shared finite dimensions after each date is standardized against only its prior 252 sessions.
SPX · +21 sessionsn=50
2-6.19% to -4.84%
3-4.84% to -3.50%
6-3.50% to -2.15%
3-2.15% to -0.80%
8-0.80% to +0.55%
10+0.55% to +1.89%
11+1.89% to +3.24%
4+3.24% to +4.59%
2+4.59% to +5.94%
1+5.94% to +7.29%
ES · +21 sessionsn=50
3-6.12% to -4.81%
2-4.81% to -3.49%
7-3.49% to -2.18%
2-2.18% to -0.86%
9-0.86% to +0.45%
8+0.45% to +1.77%
12+1.77% to +3.08%
4+3.08% to +4.40%
2+4.40% to +5.71%
1+5.71% to +7.03%
| Series | Horizon | Positive | Median | Worst decile | Range | Sample |
|---|---|---|---|---|---|---|
| SPX | +1 session | 42.0% | -0.12% | -1.21% | -1.56% to +1.43% | n=50 |
| SPX | +5 sessions | 50.0% | -0.02% | -2.44% | -2.77% to +4.55% | n=50 |
| SPX | +21 sessions | 66.0% | +0.86% | -3.19% | -6.19% to +7.29% | n=50 |
| ES | +1 session | 40.0% | -0.15% | -1.26% | -1.59% to +1.41% | n=50 |
| ES | +5 sessions | 48.0% | -0.08% | -2.40% | -2.90% to +4.82% | n=50 |
| ES | +21 sessions | 60.0% | +0.79% | -3.41% | -6.12% to +7.03% | n=50 |
The read is invalid if COT commercial percentile (GBPUSD) records at or above 77.2 percentile, the z=0.00 level from the prior n=52 observations, on its next weekly print.
Next monitored set
What to watch next.
The thesis break level comes first. Scheduled releases follow only when their published timing can affect a component in the evidence system.
EventGross Domestic Product
LevelPublished schedule
TimeWed, Aug 26 · 8:30 a.m. ET
Why it matters
Growth revisions can alter the regime and the historical setup used for comparison.
Thesis effectRegime, rates, analog context
EventPersonal Income and Outlays
LevelPublished schedule
TimeWed, Aug 26 · 8:30 a.m. ET
Why it matters
The inflation and income release can move rates, the dollar, and cross-asset agreement.
Thesis effectRates, dollar, confluence
EventEmployment Situation
LevelPublished schedule
TimeFri, Sep 4 · 8:30 a.m. ET
Why it matters
Labor strength can move growth expectations, rates, and broad risk appetite.
Thesis effectRates, regime, breadth
Published timing is informational and may be revised by the publishing agency. Open the full sourced calendar.