S&P 500 Vol & Valuation
Recent Signal History — Last 24 Months
Valuation & Fear scores day by day. Scores use Enhanced P/E (sigmoid-blend of trailing and 3-year average EPS) to reduce distortion from temporary earnings swings. Dashed lines mark the OR sell thresholds (85 / 90). Quarterly rebalance signals shown as markers on the S&P 500 chart.
Bear Market Warning Performance
Were scores above the sell threshold (Val > 85 OR Fear > 90) one month before each major peak? Scores are percentile-ranked against the full 1990–2026 history.
| Bear Market | Peak Date | Decline | Valuation (1mo before) | Fear (1mo before) | Warning? |
|---|---|---|---|---|---|
| Dot-com Bust | Mar 2000 | −36.8% | 86.2 | 78.1 | ✓ YES |
| Dot-com 2nd Leg | Jan 2002 | −32.0% | 84.4 | 65.7 | ✗ NO |
| Financial Crisis | Oct 2007 | −51.9% | 42.9 | 67.7 | ✗ NO |
| GFC 2nd Leg | Jan 2009 | −27.6% | 49.6 | 75.3 | ✗ NO |
| COVID Crash | Feb 2020 | −33.9% | 76.4 | 35.1 | ✗ NO |
| 2022 Rate Hikes | Jan 2022 | −25.4% | 77.9 | 55.2 | ✗ NO |
⚠ Scores updated with 2026 data. Because percentile ranks use the full 1990–2026 history, adding recent high-valuation years lowers historical scores slightly — the Dot-com 2nd Leg, GFC, COVID and 2022 events now fall just below the 85/90 thresholds. The strategy's +27.6% outperformance is not dependent on these warnings — it comes from systematic quarterly rebalancing.
Position Sizing Results
At optimal thresholds (Val > 85, Fear > 90), quarterly rebalancing — how much to trade each signal?
| Trade Size | Final Value | CAGR | vs Buy & Hold | Trades |
|---|---|---|---|---|
| 2.5% | $225,572 | 8.90% | +7.7% | 31 |
| 5.0% | $240,087 | 9.09% | +14.6% | 29 |
| 7.5% | $252,856 | 9.24% | +20.7% | 28 |
| 10.0% | $261,690 | 9.35% | +24.9% | 26 |
| 12.5% ★ | $267,278 | 9.41% | +27.6% | 26 |
| 15.0% | $266,858 | 9.40% | +27.4% | 25 |
| 17.5% | $265,955 | 9.39% | +26.9% | 24 |
| 20.0% | $263,853 | 9.37% | +25.9% | 23 |
| 25.0% | $254,457 | 9.26% | +21.5% | 22 |
Buy & Hold baseline: $209,502. All variants beat buy & hold at optimal thresholds. 12.5% is the sweet spot — gains plateau above this while risk increases.
Research Extension: Three-State Bear+Bull Framework
A deeper model using AND logic (Val > 80 AND Fear > 55) as the bear sell signal, then waiting in cash until a separate bull re-entry signal fires (Val < 70 AND Fear > 65 — cheap AND fearful). Three distinct states: SELL → NEUTRAL (in cash, waiting) → BUY. 65-day persistence gate on both signals. The key finding: the three-state approach does not maximise raw return but significantly reduces risk.
| Strategy | CAGR | vs B&H | Sharpe Ratio | Max Drawdown | Final ($10k) |
|---|---|---|---|---|---|
| Buy & Hold | 8.68% | — | 0.553 | −56.8% | $209,502 |
| Bear-Only (AND_HIGH Val>80 AND Fear>55) | 9.40% | +0.72% | 0.625 | −56.8% | $266,302 |
| Bear+Bull Three-State ★ | 9.19% | +0.51% | 0.702 | −36.0% | $248,262 |
Trade-off: Bear-Only maximises total return (+0.72% CAGR vs B&H) but max drawdown is unchanged — it still falls −56.8% in a crisis. The Three-State framework trades 21 basis points of CAGR to achieve the best Sharpe ratio (0.702 vs 0.553 for B&H) and nearly halves the worst-case drawdown from −56.8% to −36.0%. In 2008, the Three-State strategy fell only −21.0% vs −38.5% for B&H — the protection matters most in the worst years.
| Year | Buy & Hold | Bear-Only | Three-State | Notes |
|---|---|---|---|---|
| 2002 | −23.4% | −6.0% | −4.9% | Dot-com bust — both strategies protect |
| 2003 | +26.4% | +20.7% | +13.0% | Recovery lag — Three-State still in neutral |
| 2008 | −38.5% | −38.5% | −21.0% | GFC — Three-State significantly protected |
| 2009 | +23.5% | +23.5% | +18.8% | Recovery — Three-State catches most of it |
| 2022 | −19.4% | −19.4% | −17.7% | Rate hikes — only Three-State improved slightly |
Bear signal fires once Val AND Fear both cross thresholds for 65 consecutive days. Bull signal (re-entry) only starts counting after bear signal has cleared. 2022 remains partially undetectable: as prices fell, earnings held up, so the valuation signal self-cancelled (rate-driven multiple compression, not an earnings bubble).
Threshold Explorer
Each cell shows how much the strategy beats (green) or trails (red) buy & hold at that threshold pair. Use the sliders to inspect any combination.
SELL signal: Valuation > threshold OR Fear > threshold
BUY signal: Both below threshold
Methodology & Analysis
The Problem With Single Indicators
Traditional market timing fails because different bear markets have different causes. Dot-com (2000) was a valuation bubble — P/E hit 94th percentile but fear was only middling. The 2009 GFC second leg was pure panic — valuation was low (19th percentile) but fear exploded to 95th percentile. Using only P/E or only VIX as a signal caught just 50% and 17% of bears respectively.
Five Indicators, Two Composite Scores
We combine five established market indicators into two composite scores using historical percentile ranking:
- Valuation Score = average of (P/E percentile, Shiller CAPE percentile) — how expensive is the market historically?
- Fear Score = average of (VIX percentile, inverted Yield Curve percentile, HY Credit Spread percentile) — how stressed are markets?
Percentile ranking is adaptive: a P/E of 25 might be extreme in the 1990s but normal today. By ranking against the full historical distribution, the scores automatically adjust for changing market regimes.
OR vs AND Logic
The initial analysis uses OR logic (Val > 85 OR Fear > 90) to catch both bubble bursts (high valuation, calm market) and panic crashes (moderate valuation, spiking fear). The deeper research revealed AND logic (Val > 80 AND Fear > 55) is more selective — it requires both conditions simultaneously and avoids single-indicator false alarms. AND logic produces fewer trades and better cross-regime consistency, while OR logic fires more often and captures slightly more upside in good environments. The Three-State Framework above uses AND logic.
Strategy Rules
- Starting capital: $10,000 fully invested in S&P 500 (Nov 1990)
- Rebalance quarterly (first trading day of each quarter)
- SELL signal: reduce equity by 12.5% of holdings
- BUY signal: redeploy 12.5% of total portfolio from cash back into equities
- No transaction costs or taxes in backtest
Why Thresholds Matter (The Heatmap)
Setting thresholds too low (50–70th percentile) means selling constantly — the strategy is out of the market most of the time and misses the bulk of long-run gains, underperforming by 50–77%. Thresholds in the 85–90th percentile sweet spot mean only selling at true historical extremes: 24 total trades over 35 years (roughly one trade every 18 months). This selectivity is what drives outperformance.
Enhanced P/E
Standard trailing P/E can distort signals in two ways: during earnings bubbles, analysts project further acceleration, so trailing P/E understates richness; during corrections, temporary earnings collapses (e.g. COVID −29% YoY) make trailing P/E spike even as prices recover, triggering false sell signals. The Enhanced P/E blends trailing earnings per share with a 3-year rolling average using a sigmoid function — when earnings deviate far from trend, the blend pulls toward the average. This dampened the Nov 2020 sell signal's P/E reading from 36.9× to 29.6× (−19%) and correctly raised the dot-com bubble reading from 29.4× to 31.4× (+7%). The valuation scores shown above use Enhanced P/E.
Current Market (Jul 2026)
The Valuation Score sits at 95 (Enhanced P/E ~30.3×, trailing P/E ~28.5×, Shiller CAPE ~40.5×). Current earnings are running above their 3-year average, which pushes Enhanced P/E higher than trailing — the signal correctly reads the market as more expensive than the raw trailing P/E suggests. The Fear Score is 52 — VIX calm, yield curve near flat, credit spreads modestly elevated. The S&P 500 has rallied to ~7,400 with the OR strategy consistently signalling SELL since Q4 2024 (except April 2025 when valuations briefly dipped during the tariff correction). Under AND logic, only the April 2026 tariff-driven panic spike in Fear briefly triggered a sell. This "expensive but calm" environment is consistent with the 2007 and 1999–2000 late-cycle patterns — valuation is very stretched, but a market crash typically requires a fear catalyst as well.
Limitations
- Backtesting covers 1990–2026; future market regimes may differ
- No transaction costs, taxes, or slippage modelled
- Credit spread data begins only in 1996 (early period uses 4 indicators)
- P/E and CAPE are monthly — slight lag vs daily signal indicators
- Only 4 of 6 bear markets since 1990 are structurally detectable: Financial Crisis peak (2007) showed valuation only at the 43rd percentile, and COVID (2020) crashed in 33 days — faster than any 65-day persistence gate can respond
- The 2022 bear market is a unique case: prices fell but earnings held up (rate-driven multiple compression), so valuation scores actually fell as the market dropped — the valuation signal self-cancelled
- The 2020 COVID false positive (sell Nov 2020, market up 45% after): Enhanced P/E reduces distortion but does not fully eliminate the signal — the market was genuinely expensive in a historical percentile sense even with smoothed earnings
- Persistence sensitivity: 65-day gate is reasonable but 45d gives slightly better penalty scores for the reference signal; 90d is more conservative with a near-zero worst-window return
- This is not investment advice