Study 8: Sector ETF and Mega-Cap Profiling — TSLA's Momentum Archetype and AAPL's Defensive Secret

Overview

Study 8 profiled 9 assets: three mega-caps (AAPL, MSFT, TSLA) and six sector ETFs (EEM, XLV, XLU, XLE, XLF, XLK). The headline discovery: TSLA is a new archetype — momentum_continuation — where BB upper touch predicts CONTINUED upside, not reversal. This is the opposite of every other asset in the study.

Secondary finding: AAPL behaves like a defensive stock, not a growth stock. Previously assumed high_vol_growth; it’s actually fundamental_reverter with the best KC midline exit in the entire study (100% win rate).

Results by Asset

AAPL — fundamental_reverter (reclassified from assumed high_vol_growth)

Signal	Cohen’s d (10d)	bps	Win%
mfi_oversold	+0.437	3.5	70%
kc_lower	+0.098	1.8	64%
bb_lower	+0.031	1.4	51%

Optimal RSI: < 20 (selective — only extreme oversold triggers)
Best exit: KC midline → 4.3 bps, win=100%, Sharpe=2.44
Stop: -15.59%
Bonferroni survivors: 0

AAPL’s KC midline exit with 100% win rate is the second-best exit in the entire two-study batch (behind XLV’s KC midline at 1.586 Sharpe). Every time AAPL gets extreme oversold and enters KC midline as target, it reaches it. This is a high-quality signal that should override the previous high_vol_growth assumption.

Why was AAPL assumed high_vol_growth? Study 6’s hypothesis: high retail participation → BB lower signal. But AAPL’s 2019-2026 window shows MFI overwhelmingly dominant (d=+0.437 vs KC’s d=+0.098). AAPL is institutional-grade, not retail-driven.

MSFT — fast_index (3 Bonferroni survivors)

Signal	Cohen’s d (10d)	bps	Win%
bb_lower	+0.123	1.6	60%
kc_lower	+0.029	1.1	56%
mfi_oversold	+0.048	1.1	67%

Bonferroni confirmed (1-day/3-day):

bb_lower 1d: d=+0.435, p=0.0002
kc_lower 1d: d=+0.433, p<0.0001
kc_lower 3d: d=+0.408, p=0.0006
Optimal RSI: < 32 (broad threshold — 56 entries/period)
Best exit: KC midline → 3.3 bps, win=86%, Sharpe=0.935, ~30d hold
Panic VIX kills it: -2.5 bps, 25% win (same pattern as QQQ)
Stop: -16.1%

MSFT has three Bonferroni survivors — all short-horizon (1d and 3d). Same pattern as QQQ: signals are real, but at short horizons. KC midline is the correct exit despite the signal being at 1-3d — the entry is good and MSFT drifts up over 30d if you hold.

TSLA — momentum_continuation (new archetype)

Signal	Cohen’s d (10d)	bps	Win%
bb_upper	+0.466	7.6	62%
mfi_oversold	+0.116	3.4	58%
kc_lower	-0.142	0.2	56%
rsi_oversold	-0.171	-0.2	54%

Bonferroni confirmed (5d, 7d, 10d):

bb_upper 7d: d=+0.522, bps=6.8, win=71%
bb_upper 10d: d=+0.466, bps=7.6, win=62%
bb_upper 5d: d=+0.444, bps=5.0, win=70%
Optimal RSI: < 20 (rare trigger — 17 entries/period)
Best exit: BB midline → 4.4 bps, win=75%, Sharpe=0.447, ~30d hold
VIX structure: calm VIX (3.2 bps, 67%) > elevated (-2.3 bps) > panic (-3.5 bps, 40%)
Stop: -26.68% (median MAE -12.7%, 90th pct -20.97%)

TSLA violates every mean-reversion assumption:

RSI oversold INVERTS (d=-0.171) — buying dips is unprofitable
KC lower INVERTS (d=-0.142) — another false signal
BB upper (overbought) PREDICTS CONTINUATION — Bonferroni confirmed at three horizons
Panic VIX hurts (40% win) — calm market is best for TSLA breakouts
BB midline (target of a prior rally) is the optimal exit

This is a momentum stock: when TSLA breaks out through the upper Bollinger Band, it tends to continue. The signal is the breakout itself, not an oversold condition. The mean-reversion framework simply does not apply.

This defines a fifth archetype: momentum_continuation. Any stock exhibiting BB upper Bonferroni confirmation with RSI inversion should be classified here and traded on breakouts, not dips.

EEM — fundamental_reverter (emerging markets ETF)

Signal	Cohen’s d (10d)	bps	Win%
obv_weak	+0.168	1.0	57%
mfi_oversold	+0.091	0.9	44%
kc_lower	+0.076	0.7	63%

Optimal RSI: < 20
Best exit: fixed_7d → 1.8 bps, win=82%, Sharpe=1.023
Stop: -6.28% (tight — EM stocks don’t recover from large drawdowns)
Bonferroni survivors: 0
VIX sweet spot: elevated (59% win, 0.9 bps), panic hurts (40%, -1.9 bps)

EEM is the only asset where OBV (volume-based) is the #1 signal, outperforming MFI. This makes sense: EM fund flows are dominated by institutional capital allocation decisions. When institutional money withdraws (low OBV), it’s a contrarian entry signal. The short 7d exit also differs from other reverters — EM assets bounce quickly or not at all (illiquid, political risk, currency effects).

XLV — fundamental_reverter (Healthcare sector — best sector ETF)

Signal	Cohen’s d (10d)	bps	Win%
bb_lower	+0.307	1.5	66%
mfi_oversold	+0.300	1.6	69%
kc_lower	+0.293	1.4	62%

Optimal RSI: < 25
Best exit: KC midline → 2.9 bps, win=100%, Sharpe=1.586
Stop: -8.58%
VIX panic is the sweet spot: 2.9 bps, 75% win (elevated is negative!)

XLV is the best sector ETF in the study. Three signals near-equal strength around d=+0.30 — this is a multi-signal confirmation cluster. The KC midline exit at 100% win rate means healthcare stocks in stress always recover to their midpoint.

Note: VIX elevated is slightly negative for XLV (-0.5 bps) but VIX panic is strongly positive (2.9 bps). This is the opposite of most assets. Theory: healthcare is a defensive sector — it sells off during market panics but immediately becomes the preferred safe haven as investors seek non-cyclical exposure. The true panic entry (VIX>27) is a quality entry, not a false alarm.

XLU — fast_index (Utilities sector — surprise classification)

Signal	Cohen’s d (10d)	bps	Win%
mfi_oversold	+0.444	2.2	71%
rsi_oversold	+0.272	1.8	68%
bb_lower	+0.103	1.0	73%

Bonferroni confirmed:

rsi_oversold 1d: d=+0.418, p<0.0001
Optimal RSI: < 20
Best exit: fixed_20d → 3.6 bps, win=90%, Sharpe=1.217
Stop: -6.67% (very tight — utilities don’t move much)
VIX panic preferred: 3.7 bps, 75% win

XLU classified as fast_index because its 1-day RSI Bonferroni confirmation (d=+0.418) triggers the fast_index gate. This is accurate: utilities bounce sharply the day after going extreme oversold, driven by dividend-hunting retail traders. MFI is the dominant 10d signal (d=+0.444), but the 1d bounce is real and confirmed.

The fixed_20d exit is longer than typical fast_index assets — utilities have a slow recovery arc after the initial bounce.

XLE — fundamental_reverter (Energy sector — weakest in study)

Signal	Cohen’s d (10d)	bps	Win%
rsi_oversold	+0.172	1.9	71%
kc_lower	+0.095	1.4	63%
obv_weak	+0.072	1.2	60%

Optimal RSI: < 35 (wide — energy sells off hard)
Best exit: BB midline → 1.7 bps, win=66%, Sharpe=0.263
Stop: -23.86% (widest in study — oil crashes don’t recover quickly)
Bonferroni survivors: 0

XLE is the hardest sector to trade mean-reversion. Energy price dependence creates regime changes that break RSI signals: oil supply gluts (2020), OPEC decisions, geopolitical shocks. The -23.86% stop is the widest in the study — holding through an energy crash requires extreme conviction. The Sharpe of 0.263 is the lowest, confirming this sector’s signal weakness.

MFI notably underperforms RSI here (MFI not even in top 6). Energy volume behavior differs: high volume during energy crashes doesn’t reliably predict recovery the way it does in equity oversells.

XLF — fundamental_reverter (Financials sector)

Signal	Cohen’s d (10d)	bps	Win%
kc_lower	+0.193	1.7	77%
rsi_oversold	+0.137	1.3	73%
bb_lower	+0.093	1.0	71%

Optimal RSI: < 32
Panic VIX threshold: > 35 (banks only trigger in true systemic fear)
Best exit: fixed_30d → 5.3 bps, win=72%, Sharpe=0.537
Stop: -17.61%
Bonferroni survivors: 0

XLF’s VIX panic threshold (>35) is the highest in the study. Banks are the source of systemic risk — they only revert mean after true panic (VIX>35 is SVB-grade or COVID-grade crisis). In elevated VIX (30-35), banks may still be in the process of failing. The 75% win rate in both elevated and panic regimes shows this is a high-quality setup when triggered.

KC lower is top signal — financial stocks hug their lower KC bands during liquidity stress. The 77% win rate for KC lower is notable.

XLK — fundamental_reverter (Tech sector ETF)

Signal	Cohen’s d (10d)	bps	Win%
bb_lower	+0.243	2.2	62%
kc_lower	+0.222	2.1	67%
rsi_oversold	+0.104	1.5	49%

Optimal RSI: < 28
Best exit: rsi_40 → 3.4 bps, win=89%, Sharpe=1.131, ~4d hold ← fastest in study
Elevated VIX: 3.0 bps, 60% win — panic VIX only 0.5 bps
Stop: -15.74%
Bonferroni survivors: 0

XLK’s rsi_40 exit at 4 days is the fastest exit in the entire study. Tech stocks recover quickly from oversold conditions because algorithmic rebalancing brings them back to weight almost immediately. The rsi_40 exit (wait until RSI recovers to 40) catches this fast bounce. BB lower and KC lower both show strong signals (d≈+0.23) because XLK components are widely tracked and oversell events get bought quickly.

Cross-Asset Summary — All 15 Profiled Assets

Asset	Archetype	Top Signal	Best Exit	Sharpe	Stop
AAPL	fundamental_reverter	MFI d=+0.437	KC midline 100%!	2.44	-15.6%
XLV	fundamental_reverter	BB/MFI/KC ~d=+0.30	KC midline 100%!	1.586	-8.6%
COST	fundamental_reverter	MFI d=+0.303	fixed_30d	1.499	-12.7%
XLU	fast_index	MFI d=+0.444	fixed_20d	1.217	-6.7%
XLK	fundamental_reverter	BB lower d=+0.243	rsi_40 4d	1.131	-15.7%
EEM	fundamental_reverter	OBV d=+0.168	fixed_7d	1.023	-6.3%
MSFT	fast_index	BB lower d=+0.435 Bonferroni	KC midline	0.935	-16.1%
NVDA	fast_index	KC lower d=+0.379	rsi_55	0.624	-25.8%
JNJ	fundamental_reverter	BB lower d=+0.222	rsi_45	0.596	-7.6%
XLF	fundamental_reverter	KC lower d=+0.193	fixed_30d	0.537	-17.6%
WMT	fundamental_reverter	MFI d=+0.122	fixed_30d	1.169	-7.7%
QQQ	fast_index	MFI d=+0.718 Bonferroni	KC midline	0.613	-14.0%
TSLA	momentum_continuation	BB upper d=+0.466 Bonferroni	BB midline	0.447	-26.7%
SPY	fundamental_reverter	BB lower d=+0.093	rsi_40	0.457	-20.1%
XLE	fundamental_reverter	RSI d=+0.172	BB midline	0.263	-23.9%

Key Findings

1. Fifth archetype confirmed: momentum_continuation (TSLA)

TSLA has BB upper Bonferroni confirmation at three horizons (5d, 7d, 10d) with d=+0.466 at 10d. RSI oversold and KC lower both invert. This stock does not mean-revert — it trends. The tradeable signal is the BB upper breakout in calm VIX, not an oversold entry. This required adding a new archetype to the tree farm.

2. AAPL is fundamental_reverter, not high_vol_growth

Study 6 hypothesized AAPL might be “BB lower” driven (high retail participation). Study 8 proves this wrong: MFI d=+0.437 dominates, KC midline exit achieves 100% win rate, and the pattern (patient, 30d hold, defensive bounce) matches fundamental_reverter exactly. The previous high_vol_growth assumption was based on sector membership (tech), not behavior.

3. MFI shows strong results in this dataset — with regime and methodology caveats

MFI top signal: QQQ (d=+0.718 ✓), AAPL (d=+0.437 ✓), XLU (d=+0.444), COST (d=+0.303), XLV (d=+0.300), WMT (d=+0.122), JNJ (d=+0.209), NVDA (d=+0.172). Exception: XLE (RSI dominates), EEM (OBV dominates), TSLA (BB upper dominates).

Two caveats:

This study covers 2019–2026, a period with two fast V-shaped recoveries (COVID, 2022 rate-shock). The 20-year multi-asset study (Do Technical Indicators Actually Work?) ranks MFI 9th of 14 indicators. A 21-year QQQ-specific study finds MFI < 20 alone has p=0.882 on that asset. The strong results here may be partly regime-specific.

MFI thresholds were also not optimized. RSI was swept across multiple thresholds per asset; MFI used the standard MFI < 20 throughout. This makes direct comparisons between MFI and optimized RSI somewhat unfair. Additionally, the QQQ deep study shows the better entry signal is the MFI recovery zone (20–30) after a dip below 20, not the raw oversold condition — a refinement not yet tested in this study series.

4. KC midline exit dominates higher-quality setups

Assets with Sharpe > 1.0 (AAPL 2.44, XLV 1.586) both use KC midline exit. KC midline (a dynamic, adaptive target) outperforms fixed-day and RSI-threshold exits on assets with clean recovery profiles. When the exit target is “price returns to equilibrium” (KC midline), win rate is 100%.

5. Panic VIX patterns differ by sector type

Sector	Panic VIX effect
Healthcare (XLV)	Positive (2.9 bps, 75% win) — flight to defensive
Financials (XLF)	Positive but requires VIX>35
Tech (XLK, QQQ, MSFT)	Negative — panic overwhelms tech buyers
Utilities (XLU)	Positive — dividend hunters absorb selling
Energy (XLE)	Insufficient data

This sector-VIX matrix is actionable: in a market panic, buy healthcare and utilities, avoid tech.

6. Energy (XLE) is systematically weak for mean-reversion

XLE Sharpe=0.263, stop=-23.86%, no Bonferroni survivors. Energy is regime-dependent in a way that breaks all our signals: oil supply shocks and OPEC decisions override any technical pattern. For energy, macro/fundamental analysis dominates technical signals. The tree farm should note: XLE requires a commodity regime filter, not just VIX.

Tree Farm Updates (Study 8)

New archetype node momentum_continuation — added to generate_tree() and archetype table
TSLA tree regenerated as momentum_continuation with BB upper entry + calm VIX filter + BB midline exit
AAPL tree — fundamental_reverter, MFI in inputs, KC midline exit
Archetype table — AAPL→fundamental_reverter, MSFT→fast_index, TSLA→momentum_continuation, all sector ETFs added
mfi_oversold node metadata updated (v1.1) — expanded valid_archetypes, per-asset d values

Next Studies

Study 9: AMZN, META, NFLX, GOOGL (high_vol_growth archetype deep-dive — validate BB lower signal)
Macro regime filter: XLE needs a commodity regime node; SPY needs a rate-shock detection node
Composite signal gate: MFI < 20 AND RSI < 30 simultaneously — how many opportunities? What’s the Sharpe?
Sector rotation signal: panic VIX → buy XLV and XLU, fade XLK and XLF