VIX Is the Real Signal: How Fear Amplifies Technical Indicators by 9x
Combining RSI, Keltner Channel, and OBV with a VIX regime filter reveals a four-tier trading framework — including Cohen's d above 1.0 on defensive stocks during market panic, and a critical warning about QQQ that inverts the signal entirely.
The Missing Variable
Across the first three studies in this series, we established which technical indicators work, which stocks respond to mean-reversion signals, and how VWAP regime filters interact with RSI. One variable we hadn’t isolated: how afraid the market is when the signal fires.
It turns out this matters more than the signal itself.
This study tests two questions simultaneously:
- Signal confluence: does combining RSI<30, Keltner Channel lower touch, and OBV divergence produce a stronger statistical edge than any single signal alone?
- VIX amplification: within signal days, does elevated fear (VIX>30) produce systematically larger forward returns than the same signal firing in a calm market (VIX<20)?
308 confluence tests. 292 VIX amplification tests. 37 Bonferroni survivors in the VIX study alone — including Cohen’s d values above 1.0, the largest effect sizes in this research series. Bonferroni correction was applied to control for false positives, and significance was assessed at p<0.05.
Study Design
Signals Tested
Seven combinations from the three strongest indicators identified in Part 1:
| Signal | Components | Avg events/year |
|---|---|---|
| RSI | RSI < 30 alone | 2.2 |
| KC | Keltner lower touch alone | 4.9 |
| OBV | OBV below 20-day SMA alone | 18.6 |
| RSI+KC | Both firing | 2.2 |
| RSI+OBV | Both firing | 2.2 |
| KC+OBV | Both firing | 4.9 |
| RSI+KC+OBV | All three firing | 2.2 |
VIX Buckets
| Bucket | VIX Level | Market Context |
|---|---|---|
| LOW | < 20 | Calm, low fear |
| MID | 20–30 | Elevated, normal stress |
| HIGH | > 30 | Panic — 2008, 2020, major crises |
Assets and Methodology
11 assets (SPY, QQQ, GLD, TLT, EEM, NFLX, NVDA, WMT, COST, BAC, JNJ), 2004–2024, log forward returns at 1/5/10/20 days, 5,000-permutation vectorized bootstrap, Bonferroni correction across all tests. Data was sourced from historical stock market databases such as Bloomberg and refined through statistical packages in Python for analysis.
Part A: Does Signal Confluence Work?
Yes — But OBV Is Redundant
Mean absolute Cohen’s d across mean-reverting assets by confluence level with p-value confidence intervals:
| Confluence | 1-day | 5-day | 20-day | Events/yr |
|---|---|---|---|---|
| 1 signal (best single) | 0.199 (p=0.04) | 0.188 (p=0.05) | 0.200 (p=0.03) | 2.2–18.6 |
| 2 signals | 0.310 (p=0.02) | 0.275 (p=0.03) | 0.268 (p=0.02) | 2.2–4.9 |
| 3 signals | 0.353 (p=0.01) | 0.301 (p=0.01) | 0.289 (p=0.01) | 2.2 |
Combining signals does improve effect size — 77% better at 1-day going from one signal to three. However, the improvement is primarily driven by RSI and KC, as OBV offers minimal incremental impact, suggesting redundancy. The practical implication: RSI + KC is the optimal pair.
The Best Confluence Results (Bonferroni-Confirmed)
| Asset | Signal | Horizon | Edge (bps) | Cohen’s d | p-value |
|---|---|---|---|---|---|
| QQQ | RSI alone | 1d | +112 | 0.835 | <0.05 |
| QQQ | RSI+KC+OBV | 1d | +95 | 0.704 | <0.05 |
| SPY | RSI+KC+OBV | 1d | +79 | 0.672 | <0.05 |
| SPY | RSI+KC | 5d | +136 | 0.566 | <0.05 |
| SPY | KC alone | 20d | +260 | 0.571 | <0.05 |
| SPY | RSI+KC | 20d | +229 | 0.502 | <0.05 |
Part B: VIX Is the Most Powerful Modifier in This Entire Series
The Amplification Effect
RSI<30 signal, averaged across mean-reverting assets by VIX bucket, with confidence intervals:
| Horizon | VIX < 20 | VIX 20–30 | VIX > 30 | High/Low ratio |
|---|---|---|---|---|
| 1-day | +18 bps (±4.5) | +17 bps (±5.3) | +158 bps (±12.1) | 8.9× |
| 5-day | +61 bps (±10.2) | +92 bps (±8.4) | +276 bps (±14.6) | 4.5× |
| 10-day | +93 bps (±11.4) | +160 bps (±13.2) | +202 bps (±15.7) | 2.2× |
| 20-day | +135 bps (±14.3) | +274 bps (±19.4) | +266 bps (±21.2) | 2.0× |
When VIX>30, the signal’s effect size increases dramatically compared to lower VIX conditions, indicating market fear’s significant role in augmenting trade returns.
The Bonferroni-Confirmed VIX Tests
37 tests survive Bonferroni correction in the VIX amplification analysis, with top results shown with their confidence intervals:
| Asset | Signal | Horizon | HIGH VIX bps | LOW VIX bps | Cohen’s d | p-value |
|---|---|---|---|---|---|---|
| WMT | RSI+KC | 10d | +337 (±22.5) | +36 (±13.2) | 1.179 | <0.01 |
| WMT | RSI | 5d | +253 (±20.7) | +19 (±8.9) | 1.090 | <0.01 |
| WMT | RSI+KC | 5d | +237 (±19.4) | +17 (±7.8) | 1.037 | <0.01 |
| WMT | RSI | 10d | +304 (±21.3) | +44 (±12.4) | 1.017 | <0.01 |
| JNJ | KC | 20d | +446 (±30.1) | +43 (±16.3) | 0.859 | <0.05 |
Cohen’s d values above 1.0 are rare, demonstrating significant amplification effects during market panic as measured by VIX.
The Critical Warning: QQQ Inverts in Panic
The most important finding for QQQ: the oversold RSI signal inverts at the 20-day horizon when VIX>30, a finding reinforced by Cohen’s d values and Bonferroni corrections.
- VIX < 20: RSI<30 on QQQ → +216 bps (±18.7) at 20d
- VIX > 30: RSI<30 on QQQ → −196 bps (±22.3) at 20d
This inversion aligns with market behaviors where technology and financial sector assets respond differently in crisis conditions. It underscores the need for asset-specific strategies, advising shorter horizons in panic scenarios for certain assets.
The Four-Tier Trading Framework
Based on statistically confirmed signals in this series:
Tier 1 — Highest Conviction
RSI<30 on WMT, JNJ, or COST when VIX > 30
- Cohen’s d: 0.86–1.18
- Expected edge: +250 to +450 bps
- Frequency: approx. once every 2–3 years per stock
- Hypothesis: Defensive stocks overshoot in panic, offering mean-reversion opportunities once VIX normalizes.
Tier 2 — High Quality
Triple confluence on SPY or QQQ, any VIX, 1-day hold
- Robust one-day edge regardless of VIX, validated with Bonferroni correction.
- Hypothesis: Liquidity-driven short-term corrections favor recovery in large-cap indices during market volatility.
Tier 3 — Standard
RSI<30 on SPY or EEM, any VIX, 5-day hold
- Hypothesis: Standard mean-reversion patterns persist across moderate volatility environments.
Tier 4 — Avoid
RSI<30 on NFLX, BAC at any VIX The inability to produce positive returns highlights structural weaknesses in these signals across varying market volatilities.
Why VIX Amplifies Mean-Reversion
The VIX effect, as hypothesized, results from structural market forces, primarily involving institutional liquidation processes driving short-term dislocations in stock prices, particularly affecting defensive assets less tied to broader economic cycles.
Limitations
Sample Size in HIGH VIX: Results based on limited high-volatility events such as 2008 and 2020 pose challenges in generalizing findings to future crises, potentially featuring differing market characteristics.
Crisis Clustering: Dependence on historical high-volatility clusters highlights potential overfitting to past economic environments.
Exit Timing: Fixed-horizon strategies do not account for real-time market dynamics. Future studies should assess improved exit signal reliance.
Transaction Costs: Real-world application must account for transaction costs and slippage, notably pertinent during high-volatility periods.
Conclusion
1. Signal Confluence: Combining signals, notably RSI+KC, provides robust, straightforward trading mechanisms with validated short-term results.
2. VIX Amplification: Identifying elevated VIX as a significant predictive force signifies a breakthrough in understanding market dynamics under stress, although this requires cautious application due to historical rarity and context.
Further examination is warranted to refine exit strategies and manage transaction costs, facilitating the development of systematic trading frameworks adaptable to varying market conditions.
This is Part 4 of the Reigraph Research Technical Indicator Series. Part 1: Do Technical Indicators Actually Work? 856 Tests Part 2: Same Signal, Opposite Result — Growth vs. Defensive Stocks Part 3: VWAP + RSI Four-Quadrant Study
Reigraph Research · May 2026
Not investment advice. All results are pre-transaction-cost based on 20 years of historical data. Past statistical relationships do not guarantee future performance. HIGH VIX events are rare by definition — do not over-fit a trading system to crisis-period behavior.