The most accurate description is not that the entire market has completed a deep correction. Technology and AI-heavy leaders have suffered a real drawdown while equal-weight equities and many other sectors remain supported. The SPX long-term trend, implied-volatility term structure and credit conditions are constructive, but NDX trend and Nasdaq internals have not completed a repair.
COR1M estimates option-implied average correlation among the 50 largest SPX constituents. It helps distinguish dispersion from synchronized selling, but cannot determine whether a correction is complete by itself. Low COR1M can coexist with a large drawdown in a concentrated technology portfolio, while spikes often occur along crisis-and-rebound paths.
A point-in-time month-end test from July 2016 through June 2026 finds only a 25.00% stable rate after pre-defined COR1M spikes. A new daily validation from July 2021 through July 2026 finds a 63-day return IC of only 0.055 for joint stress, versus a 0.714 AUC for severe drawdown. The pair belongs in a risk overlay, not as a standalone return factor.
Broad-index trends, the implied-volatility term structure and credit conditions remain supportive of risk assets. Nasdaq breadth and the NDX short-term trend have not confirmed a durable repair. Maintain core broad-market exposure and stage additions to technology/AI risk. A sharp COR1M spike is neither required nor a reason by itself to add risk.
1. Executive Summary
The key analytical error is to extrapolate a meaningful drawdown in a technology-heavy portfolio into a claim that the entire U.S. equity market has completed a cleansing correction. The data do not support that conclusion.
- The broad market has barely corrected. SPX is about 1.24% below its trailing 252-session closing high, equal-weight SPEQX only 0.41% below, and RUT 2.35% below. These are consolidation-scale moves, not broad capitulation. [Cboe SPX] [SPEQX] [RUT]
- The technology correction is real. NDX closed at 29,264.10, below its 20-day average of 29,691.37 and 50-day average of 29,431.17, and 4.55% below its trailing 252-session closing high. [Nasdaq]
- Systemic stress remains contained. VIX closed at 17.16 versus VIX3M at 19.64, a ratio of 0.874, leaving the constant-maturity implied-volatility curve upward sloping. HY and IG option-adjusted spreads were 2.69% and 0.77%, respectively. [Cboe VIX] [VIX3M] [FRED HY] [FRED IG]
- There has been no washout. Credit spreads sit near the tight end of the roughly three-year public history currently available, while NAAIM exposure remains elevated at 82.95. The regime is supportive, but it is not a classic high-conviction capitulation low. [NAAIM]
2. Price Trend and Market Breadth
| Index | 13 Jul close | 20D SMA | 50D SMA | 200D SMA | Below 252D closing high | Read |
|---|---|---|---|---|---|---|
| SPX | 7,515.34 | 7,469.52 | 7,440.71 | 6,969.43 | -1.24% | Constructive |
| SPEQX equal weight | 867.77 | 860.04 | 843.37 | 801.26 | -0.41% | Constructive |
| RUT | 2,953.17 | 2,982.29 | 2,908.57 | 2,641.29 | -2.35% | Neutral |
| NDX | 29,264.10 | 29,691.37 | 29,431.17 | 26,242.62 | -4.55% | Short-term weak |
Simple moving averages and drawdowns are mechanically calculated from official closing-price histories; they are not exchange-published signals. Highs refer to the highest close in the trailing 252 sessions, not intraday or all-time highs.
Broad participation remains supportive
About 62.93%/64.73%/66.73% of SPX constituents were above their 20-/50-/200-day averages. Equal weight remains close to its high, indicating that most large caps have not entered a synchronized bear phase. [StockCurry]
Nasdaq internals remain weak
Only about 50.54% of NDX constituents were above their 50-day average. Nasdaq registered 45 new 52-week highs against 119 new lows on 13 July, while the one-day advancer/decliner ratio was roughly 0.60. [Barchart] [Nasdaq screener]
Interpretation: a portfolio concentrated in semiconductors, AI or crypto-linked equities can suffer a meaningful drawdown while financials, healthcare, value and other equal-weight constituents offset the damage at the index level. Portfolio pain and a calm SPX are different exposure maps, not contradictory observations.
3. Why COR1M Is Insufficient
COR1M is Cboe's estimate of the one-month option-implied average correlation among the 50 largest SPX constituents. It asks whether heavyweight stocks are expected to move together—not whether a specific sector has fallen far enough.
ρ = [σ²index − Σw²iσ²i] / [2Σi<jwiwjσiσj]
COR1M closed at 5.51 on 13 July, an extremely low reading in Cboe's daily history. It had risen from 3.44 on 10 July but remained below its 30 June close of 5.79. That profile is consistent with high dispersion—technology weakness offset by strength elsewhere. [Cboe COR1M]
Low implied correlation can coexist with material risk
- If single-stock implied volatility rises while index volatility stays moderate, implied average correlation can fall.
- COR1M covers only the 50 largest SPX constituents; it does not represent small caps, crypto, futures or thematic stocks outside the basket.
- Offsetting sector moves suppress index volatility without reducing losses for investors concentrated in the losing sector.
- COR1M is more useful as a contemporaneous regime indicator than as a standalone one-to-three-month return forecast.
Best use: treat COR1M as one input in the “synchronized selling versus dispersion” bucket, with less weight than price trend, breadth, the volatility term structure and credit conditions.
4. Methodology: Turning “Is the Correction Complete?” into Testable Questions
Whether a correction is complete is not a one-indicator question. It contains three separate questions: is systemic stress still expanding, have price and participation repaired, and is the forward path sufficiently stable? Combining them into one verdict creates false precision.
| Layer | Research question | Primary inputs | Use | Do not use it to |
|---|---|---|---|---|
| Synchronized stress | Are constituents being sold together? | COR1M, VIX, VVIX | Separate dispersion from systemic de-risking | Call a bottom by itself |
| Price trend | Has the index re-established trend? | SPX 200D; NDX 20D/50D | Confirm long- and short-horizon direction | Treat moving averages as guaranteed support |
| Participation | Is the advance broadening? | SPEQX/SPX, member breadth, new highs/lows | Distinguish cap-weight leadership from broad repair | Treat equal weight as perfect breadth |
| Volatility term structure | Does near-term risk exceed longer-term risk? | VIX versus VIX3M | Identify acute stress inversion | Equate constant-maturity indices with VIX futures |
| Credit and liquidity | Has stress reached financing markets? | HY/IG OAS | Validate whether stress is systemic | Use the short public history as a core backtest factor |
| Positioning and sentiment | Has forced liquidation occurred? | NAAIM, put/call, surveys | Supplement crowding and capitulation analysis | Mechanically combine different frequencies and definitions |
Pre-specified historical-test rules
- Sample: July 2016 through June 2026, using the final trading day of 120 complete calendar months. The 13 July 2026 reading is a current snapshot, not a month-end observation.
- Point-in-time signals: the COR1M percentile uses only the preceding 756 trading days and excludes the signal day; moving averages and relative strength also use information available before the signal.
- COR1M spike: month-end COR1M is at or above the 80th percentile of its prior 756 sessions and has risen at least five points over 21 trading days.
- Stable advance: the next 63-trading-day SPX price return exceeds 3% and the path, including the entry close, has a peak-to-trough maximum drawdown no worse than -5%. This requires both direction and path quality.
- Three-state score: trend = SPX above its prior 200-day average; participation = SPEQX/SPX above its prior 126-day ratio mean; volatility term = VIX below VIX3M. At least two conditions define risk-on.
- Return convention: SPX close-to-close price returns, excluding dividends, transaction costs and execution lag; 21/63/126-day outcomes are measured from month-end observations.
5. Historical Test: A COR1M Spike Is Neither Necessary nor Sufficient
The sample does not support the claim that a sustained advance must wait for a large COR1M rise, nor does it support treating a spike as a sufficient buy signal. With only 12 spikes and 43 stable paths, the correct phrasing is “not supported by this historical sample,” not “impossible in the future.”
| Test | Count | Rate | Wilson 95% interval | Interpretation |
|---|---|---|---|---|
| Unconditional stable path | 43 / 117 | 36.75% | — | Historical base rate |
| Sufficiency: stable after spike | 3 / 12 | 25.00% | 8.9%–53.2% | Below base rate; wide interval |
| Necessity: spike before stable path | 3 / 43 | 6.98% | 2.4%–18.6% | Most stable advances lacked the spike |
COR1M quintiles: higher correlation maps to a rougher path, not monotonic direction
Each month is classified relative only to the prior 756 trading days, avoiding future information. The point-in-time quintiles are therefore not forced to have equal sample sizes. Q1 is the lowest-COR1M group and Q5 the highest.
| Horizon / outcome | Q1 low | Q2 | Q3 | Q4 | Q5 high |
|---|---|---|---|---|---|
| 63D mean return | 2.25% | 4.13% | 3.27% | 2.90% | 6.24% |
| 63D mean max drawdown | -6.14% | -5.21% | -7.32% | -8.06% | -10.83% |
| 63D observations | 39 | 19 | 23 | 18 | 18 |
| 126D mean return | 5.98% | 6.80% | 4.94% | 8.35% | 9.90% |
| 126D mean max drawdown | -8.77% | -8.87% | -13.49% | -9.72% | -13.22% |
Key interpretation: the highest-COR group does not have the lowest subsequent mean return, but it experiences materially deeper path drawdowns. It often captures crisis periods and the rebound that follows, not a mechanical “spike first, then stable rally” sequence. COR1M says more about path risk than monotonic return direction.
The combined state adds some information, but it is not a trading holy grail
| Model | Sample | Stable-path AUC | Risk-on stable rate | Risk-off stable rate | Risk-on / off 63D mean |
|---|---|---|---|---|---|
| COR1M point-in-time percentile only | 117 | 0.547 | — | — | — |
| Three-state: trend + participation + vol term | 117 | 0.619 | 39.81% (103) | 14.29% (14) | 3.31% / 4.66% |
| Three-state, non-overlapping 63D sample | 39 | 0.612 | 48.48% (33) | 16.67% (6) | 3.77% / 1.42% |
| Add HY credit (short sample) | 30 | 0.467 | 36.36% (22) | 37.50% (8) | 4.83% / 6.44% |
An AUC near 0.5 is close to random ranking; 1.0 is perfect. The three-state framework has better in-sample discrimination of stable paths than COR1M alone, but only 14 risk-off months and no consistent improvement in mean return. The defensible conclusion is that the multi-state framework adds limited path information, not evidence of a robust tradable edge.
Robustness checks and failed extensions
- Using only the top 20% COR level, only a 10-point 21-day jump, or an absolute COR of at least 40 still produces sufficiency rates of 23.8%–28.6% and necessity rates of 11.6%–14.0%.
- Relaxing “stable” to a positive 63-day return with drawdown no worse than -10% yields 58.3% sufficiency after a spike, still below the 70.1% unconditional rate; necessity is only 8.5%.
- Sampling every third month to create non-overlapping 63-day windows gives AUCs of 0.612 for the three-state score and 0.545 for COR alone. Direction is unchanged, but there are only 39 observations.
- The currently available public FRED/ICE BofA HY file supplies only about three years of history and does not improve the model. Cboe RUT data begin in late 2020, and substituting RUT/SPX for the breadth proxy also fails to improve it.
Reproducible artifacts include official raw CSV files, monthly observations, source SHA-256 hashes, the full event table and results JSON generated by one dependency-free script. Monthly 63/126-day windows overlap; the non-overlap check is small. Results exclude dividends, costs, taxes and execution delay and make no significance or causal claim.
5A. Five-Year Daily Validation: A Risk Factor, Not Stable Alpha
A separate daily test validates the month-end conclusion over 1,255 common trading days from 14 July 2021 through 14 July 2026. Each indicator is ranked only against its own prior 756 closes, excluding the current day. Signals form after the close and returns begin at the next session's close, avoiding same-close execution look-ahead.
co-stress(t) = min[pVIX(t), pCOR(t)]
forward return = SPX(t+1+h) / SPX(t+1) - 1
What COR1M adds beyond VIX: stress breadth, not a safer bottom
| 63D state | Overlapping days | Mean return | Positive rate | Mean path MDD | Stable rate |
|---|---|---|---|---|---|
| VIX above 80th percentile only | 76 | +6.64% | 81.58% | -7.19% | 31.58% |
| VIX and COR1M both above 80th | 85 | +4.42% | 64.71% | -10.26% | 20.00% |
| Both-high minus VIX-only | — | -2.22pp | -16.87pp | -3.06pp | -11.58pp |
The point-in-time VIX and COR1M percentiles correlate at 0.735, so a simple average is substantially redundant. More importantly, the 63-day return relationship changes sign: co-stress IC is -0.184 in July 2021–December 2023 and +0.443 in January 2024–July 2026. Across all 63 non-overlapping starting phases, the 10th–90th percentile IC range spans -0.226 to +0.250.
Current Read at the 14 July 2026 Close
| Indicator | Close / point-in-time percentile | Interpretation |
|---|---|---|
| SPX | 7,543.59; -0.87% from 63D high | Above 20D 7,475 and 50D 7,447 averages |
| VIX | 16.50; 51.72% | Near neutral |
| COR1M | 4.60; 1.46% | Extreme dispersion |
| Co-stress | 1.46% | No synchronized systemic selloff signal |
One-Month Playbook: Stay Constructive, Do Not Chase
The window runs from 15 July through 14 August 2026. All exposure percentages are relative to the investor's pre-existing target U.S. equity allocation, not total account value. Scenario weights are planning assumptions, not model probabilities.
Keep 80%–90%
- Hold core exposure
- Do not sell merely because COR1M has not spiked
- Do not fill the allocation at SPX 7,610 in one trade
Scale near 7,460
- Near a 2% pullback and the 20/50D averages
- Require 7,380 to hold
- Add 10pp only if VIX and COR1M are not both above 80%
De-risk on price break plus co-stress
- Stop adding below 7,380
- If VIX is roughly ≥19.5 and COR1M ≥22.6, cut to 50%–60% of target
- Do not bottom-fish immediately below 7,230
Breakout rule: after two closes above 7,610, with VIX below roughly 19.5 and COR1M below 22.6, add 10 percentage points but cap exposure at 90% of target. Re-entry after stress: require SPX to reclaim its dynamic 20-day average and co-stress to fall below its 60th percentile, currently approximated by VIX below 17.2 and COR1M below 15.9.
| Event (China Standard Time) | Discipline |
|---|---|
| 30 Jul, 02:00 — FOMC | Do not max short-dated options or leverage before the decision |
| 7 Aug, 20:30 — Payrolls | If labor weakens again, assess growth risk rather than only betting on easing |
| 12 Aug, 20:30 — CPI | Test whether June disinflation persists |
Across 686 overlapping historical days within 2% of the 63-day high and with both stress percentiles below 80%, the next 21 days averaged +0.74%, were positive 63.70% of the time, had a -3.12% mean path MDD, and breached -5% in 14.14% of observations. That supports mild constructive exposure, not high leverage. The macro backdrop is mixed: June CPI was -0.4% month-on-month and 3.5% year-on-year, core CPI 2.6% year-on-year; June payrolls rose 57,000 with 4.2% unemployment; the 10-year Treasury yield closed at 4.58% on 14 July.
The pre-specified primary test is co-stress versus forward 63-day SPX price return. Other horizons, state tables and event studies are diagnostic. Overlapping regressions use Newey-West lag h-1 and all non-overlapping phases are reported. Results exclude dividends, costs, tax and slippage and do not incorporate the reader's holdings or risk tolerance; this is a conditional risk framework, not personalized investment advice.
6. Volatility, Tail Risk and Credit (Original Snapshot: 13 Jul 2026)
| Indicator | Latest | What it measures | Current read |
|---|---|---|---|
| VIX | 17.16 | 30-day SPX implied volatility | Normal |
| VIX3M | 19.64 | 3-month SPX implied volatility | Above VIX |
| VIX/VIX3M | 0.874 | >1 means the near constant-maturity point exceeds 3M | No inversion |
| VIX9D | 15.13 | Short-horizon event risk | Below VIX |
| VVIX | 95.28 | Implied volatility of VIX | Neutral |
| SKEW | 145.69 | Relative pricing of tail insurance | Near 3Y median |
| HY OAS | 2.69% | High-yield credit risk premium | Risk appetite strong |
| IG OAS | 0.77% | Investment-grade credit premium | Low stress |
HY and IG spreads sit near the 8th and 13th percentiles, respectively, of the roughly three-year public history currently available. Tight spreads support risk assets but also indicate little distress premium or cushion. This is a benign regime, not a crisis low. Near-term event risk did rise on 13 July: VIX gained 2.13 points and VIX9D 3.98 points. Because the short end remained below three-month and longer constant-maturity measures, the move had not yet developed into a sustained stress inversion.
Volatility: Cboe historical data. Credit: ICE BofA series via FRED, latest observation 10 July. The downloaded public credit history runs from 14 July 2023 through 10 July 2026.
7. Positioning and Sentiment: No Capitulation Entry
The NAAIM Exposure Index was 82.95 on 8 July, down from 98.59 on 24 June but still elevated. NAAIM explicitly describes the series as a record of risk adjustments made by active managers over the prior two weeks, not as a predictive indicator. [NAAIM]
Cboe's latest daily page places the equity put/call ratio in a low-to-neutral range. Daily put/call readings are noisy because of 0DTE activity, index hedging and block trades, and should not be used as standalone timing signals. [Cboe Daily Statistics]
What remains unproven: completion of the technology correction, a durable turn in Nasdaq breadth, and a full reset in crowded positioning.
8. Scenarios and Transmission
Rotation persists; tech repairs gradually
- SPX holds its 50-day average
- Credit remains contained
- NDX chops around its 20-/50-day averages
- Posture: retain core exposure; stage tech additions
Breadth and technology align
- NDX reclaims 29,431 and 29,691
- Nasdaq new highs exceed new lows
- RUT regains its 20-day average
- Posture: tolerate more concentrated risk
Rotation becomes a broad correction
- SPX breaks 7,441 and RUT 2,909
- VIX/VIX3M rises above 1
- HY OAS widens materially through ~3.2%
- Outcome: correlation and mechanical deleveraging rise
Downside mechanism
Rates or earnings shock → crowded technology positions weaken → Nasdaq new lows broaden → the implied-volatility curve inverts → credit widens → portfolio correlations rise and systematic deleveraging accelerates. COR1M would rise as part of this transmission; it would not be the cause of the initial shock. Current evidence is near the first two steps, not the full chain.
9. Reusable Confirmation Framework
Before increasing risk, require at least three
- NDX closes above 29,431 on a sustained basis and preferably above 29,691.
- Nasdaq new highs exceed new lows for two to three sessions.
- RUT holds 2,909 and reclaims 2,982.
- Equal-weight SPX does not weaken versus cap weight.
- VIX/VIX3M remains below 0.95.
- HY OAS remains below roughly 3.2% without rapid widening.
Re-underwrite risk if signals cluster
- SPX and RUT both break their 50-day averages.
- VIX/VIX3M rises above 1 and stays there.
- VVIX accelerates above roughly 115.
- HY OAS moves toward 4%.
- Less than 50% of SPX members remain above the 50-day average.
- COR1M and VIX jump together, signaling a shift from dispersion to synchronized selling.
These are monitoring thresholds, not causal forecasts or guaranteed support levels. The moving-average levels change daily. Roughly 3.2% is near the median HY OAS in the currently available public sample, while 4% is a stress-review zone—not a universal law.
Action discipline
- Broad market: trend remains intact; classify as “hold/monitor,” not “wait for inevitable panic.”
- Technology and AI: classify as “wait for proof/stage exposure,” not “add because COR1M is low.”
- Concentrated portfolios: benchmark risk against the portfolio's actual technology, semiconductor and crypto exposures; SPX alone is insufficient.
10. Sources, Reproduction and Limitations
Source posture: this is a public-data market-structure report. The evidence supports a tactical “monitor, stage, wait for confirmation” posture, but not precise sizing. It does not include institutional real-time positioning, dealer gamma, proprietary CTA models or a look-through of the reader's portfolio.
- Cboe: SPX, RUT, SPEQX, VIX, VIX3M, and COR1M.
- Nasdaq: NDX history and market screener snapshot.
- FRED / ICE BofA: U.S. HY OAS and U.S. Corporate IG OAS.
- Supplemental breadth and positioning: StockCurry, Barchart Momentum, NAAIM Exposure Index, and Cboe Put/Call.
- COR1M methodology: Cboe Implied Correlation Index White Paper v1.0.5.
Reproduction: run node backtest.mjs in the research directory to rebuild results/backtest.json, monthly_observations.csv and the summary from the saved official-source files. The output records SHA-256 hashes so inputs can be audited.
Principal limitations: breadth uses third-party closing statistics; credit data lag equities and the currently available public history is only about three years; technical levels move with each close; backtest windows overlap and class sizes are unbalanced; public positioning series have different frequencies and definitions. The report excludes valuation, earnings-revision breadth, total-return indices and the reader's actual holdings. Historical percentiles are sample-dependent, not fixed causal thresholds.