Track Record
Has this read ever been right? Below, the composite is reconstructed back to 1997 from deep-history proxy metrics (Shiller CAPE & real-EPS growth, Nasdaq/S&P stretch, AAII sentiment, FINRA margin debt, FRED rates & fundamentals) and run through the exact same point-in-time engine the live score uses.
- Scores before 2023-06-01 are reconstructed from proxies; after that date the live metrics take over. The splice is a genuine methodology change and is marked on the chart.
- Every proxy is an index-level series — none is rebuilt from today's S&P constituents, so there is no survivorship bias.
- This is a valuation-aware standing-risk monitor, not a crash timer: it flagged the big bubbles and stayed low in genuinely cheap markets, but reads elevated through historically expensive eras too.
Event history
Reconstructed composite at each event, and the S&P 500 drawdown over the following 12 months. Crashes should read elevated before the fall; calm periods should read low with shallow drawdowns.
| Event | Date | Reconstructed score | Percentile of own history | Next-12mo S&P drawdown |
|---|---|---|---|---|
| Dot-com peak | 2000-03-01 | 69 Elevated | 79.5th · Elevated | -10.1% |
| Global Financial Crisis | 2007-10-01 | 56 Elevated | 53.8th · Moderate | -28.5% |
| COVID crash | 2020-02-01 | 61 Elevated | 78.1th · Elevated | -30.6% |
| 2022 bear market | 2022-01-01 | 55 Elevated | 45.8th · Moderate | -24.9% |
| Calm — mid-2004 | 2004-06-01 | 40 Elevated | 7.8th · Low | -5.2% |
| Calm — mid-2013 | 2013-06-01 | 53 Elevated | 44.9th · Moderate | -3.5% |
| Calm — mid-2017 | 2017-06-01 | 56 Elevated | 52.8th · Moderate | -0.8% |
All scores reconstructed. Drawdown = peak-to-trough S&P 500 close over the 12 months after each date.
Historical read-through: every month, not 7 dates
The event table above checks 7 hand-picked dates — easy to cherry-pick. This buckets every month of reconstructed history by the composite's percentile-of-own-history (fixed 33rd/67th-percentile cutpoints, ranked as-of each month with no lookahead — Low/Moderate/Elevated) and reports the actual forward-12mo S&P 500 return for each bucket.
| Tier | Months (n) | Mean fwd 12mo return | Median fwd 12mo return |
|---|---|---|---|
| Low | 102 | 6.69% | 10.46% |
| Moderate | 148 | 5.67% | 11.27% |
| Elevated | 59 | 14.85% | 11.92% |
On this reconstructed history, "Elevated" months have shown a higher mean/median forward return than "Low" months — the opposite of what a working risk signal would predict. Higher scores have not preceded worse returns here, so the score should not currently be read as predictive at this 12-month horizon.
Reconstructed composite, full history
live · reconstructed (pre-2023-06-01)