Elevated top risk: internals are weakening beneath index highs; credit and jobs have not confirmed yet
SPY 759.97, 2.5% below its 52-week high · Type 1 · positioning blow-off (historically mostly V-shaped)
Key points this week
■
Top-risk score 50.3, elevated risk, -4.2 on last week (10-year percentile: 56%). In back-tests, of the 66 weeks with SPY within 3% of its 52-week high and a score in the same band, 18% were followed by a drawdown of 10% or more within 3 months (the top probability); the average for all weeks in that position is only 9%, so this band runs at 1.9× the average.
■
Cracks come from: Days the A/D line has lagged SPY's new high 36 days (score 98); % of stocks above 200-day line (top 1,000 by liquidity) 50.7% (score 77); New highs − new lows, 10-day average +1.3% (score 73).
■
Not yet confirming: Jobless claims off 2-year low +0.1% (score 7); Cross-asset divergence (SPY near high) 20% (score 20); Credit spread widening off 1-year low +7bp (score 28). If these turn higher, the score moves into a higher band.
■
Fuel 66 (Buffett indicator (market cap / GDP) 255% (score 99); Equity risk premium (1/CAPE − real rate) -0.17% (score 99); Shiller CAPE 40.5x (score 99)). Fuel sets how far a market falls once it breaks; in back-tests it had no timing power, so it is left out of the headline score.
■
Biggest moves this week: VIX/VIX3M (low = steep contango) -25 pts; Share of leading sectors broken down (SPY near high) -22 pts; High-beta vs low-vol, 6-month relative momentum -22 pts.
■
Type: Type 1 · positioning blow-off (historically mostly V-shaped).
Top-risk score
50.3
Elevated risk▼ -4.2 on the week
Top probability (≥10% fall within 3 months)
18%
Same-position average 9% · this band is 1.9× that
Historical frequency of a ≥10% SPY drawdown within 3 months · sample: 66 weeks in the same band with SPY within 3% of its 52-week high
Exhibit 1: Odds of a fall over the next 1 and 3 months: this week's band vs the historical average
Scenario
This week (elevated risk)
Same-position avg.
This week vs avg.
All-week avg. (ref.)
Fall ≥5% within 1 month
20%
11%
1.8× (+9 pts)
18%
Fall ≥5% within 3 months
33%
29%
1.2× (+5 pts)
34%
Fall ≥10% within 3 months
18%
9%
1.9× (+9 pts)
15%
Same-position average = all past weeks in which SPY was within 3% of its 52-week high; this week = the 66 of those weeks whose score was in the same band (elevated risk). Multiples and percentage points compare this week with the same-position average (computed on unrounded odds). Fall = SPY's maximum drawdown from that week's close within the window. Sample 2006-09-08 → 2026-07-31.
1. Score history
Exhibit 2: Top-risk score and SPY (past 3 years, weekly)
Note: headline score = crack score; fuel is background only. Shading shows the bands; red vertical lines are the historical tops used for calibration. Sources: Polygon, CBOE, FRED, CFTC, AAII, CNN, FINRA, multpl; WhyNext calculations.
Exhibit 6: Closest historical analogs (nearest on the six class scores, one per quarter)
Date
Distance
Score then
Next 13 wks return
Next 13 wks max drawdown
2025-01-10
22
60.4
-8.0%
-13.0%
2015-07-24
22
67.9
-0.2%
-7.4%
2023-09-22
27
63.5
+10.0%
-4.6%
Note: similar is not the same; small sample, for reference only. Sources: Polygon, CBOE, FRED, CFTC, AAII, CNN, FINRA, multpl; WhyNext calculations.
3. What would prove us wrong
■ Cracks are low. If leading sectors (IWM/XLF/KRE/SMH etc.) fall more than 10% while SPY makes new highs, or HYG/EEM/copper/BTC diverge, the score should go up at once — in January 2026 the Russell peaked only 6 days before the S&P 500.
■ Fuel is high, but that alone is not a timing signal: in back-tests, high fuel with few cracks mostly kept rising. If fuel stays high while SPY keeps making new highs and cracks do not rise, it is too early to bet against the market.
4. Track record and method
Exhibit 7: Full sample: score (4-week average) and SPY drawdown; red lines = Type 2 tops, blue lines = Type 1 tops
Exhibit 8: Historical outcomes by band (weeks with SPY within 3% of its 52-week high)
Band
Score
Weeks
Fall ≥5% in 1 month
Fall ≥5% in 3 months
Fall ≥10% in 3 months
Low risk
< 40
365
9%
24%
6%
Moderate risk
40–50
135
11%
32%
7%
Elevated risk
50–60
66
20%
33%
18%
High risk
≥ 60
28
25%
71%
46%
Average
594
11%
29%
9%
Sample 2006-09-08 → 2026-07-31; the ≥60 band rests on only a handful of independent episodes, so treat it as a rough scale, not a precise probability.
Predictive power: with SPY near its high, the score's AUC for a ≥10% drawdown over the next 13 weeks is 0.68 (0.5 = no predictive power); the base rate is 9%, against 24% for the top-scoring 20% of weeks. Method: 27 indicators in 6 classes are first aligned for publication lag, so no future data leaks in, then converted to rolling 10-year percentiles. The headline score uses only the "crack" indicators (breadth, credit turning, jobless claims turning, cross-asset divergence): back-tests show that "fuel" indicators such as euphoria, positioning and valuation, when high, have tended to precede further gains (AUC 0.36). Limits: the model can flag tops where the market weakens from within first (2007, 2022, July 2023, January 2026); it largely misses tops caused by a pure positioning squeeze or an outside shock (January 2018, February 2020).
Important disclosures This page is generated automatically each week by a deterministic quantitative model; no large language model is used in the scoring. All thresholds and probabilities come from historical back-tests since 2006. Independent market tops are few, and future results may differ. This page is for information only and is not investment advice or an offer to buy or sell. Markets carry risk; invest with care.