GOOGL: horizon distributions, backtests, adversarial review

2026-07-30 · spot ~$334 · companions: deep dive · restrategy · options report · six-lens panel stress-tested and revised per adversarial review - analysis, not licensed financial advice
Post-adversarial headline: the six "independent" lenses are really ~2 independent votes (a consensus cluster + the bear), and the spread-selling edge is weighting-dependent: +$400-900/contract at equal weights, ≈ $0 with the bear at 50%, and ~-$800-1,200 under the market's own implied distribution (= the bid/ask cost plus the left-tail insurance premium you are being PAID to underwrite). The trade survives as a deliberately-sized directional bet - "I believe the bear deserves ~1/6 weight, not ~1/2" - not as an arbitrage. Backtests support the mean-reversion prior on 2016+ data but flag one live warning: three >20% drawdowns in 18 months (~4x the historical rate) plus first-ever negative FCF is exactly the pattern where single-name dip-buying history stops being the right reference class.
Curves: StratecheryBridgewater SemiAnalysisShort-seller CompounderDamodaran Equal mixture Bear-weighted 50% (dashed red) Market-implied (dotted)

1 · 6 months

$100$200$300$400$500$600spot
LensP5P25P50P75P95
Stratechery$251$300$346$371$405
Bridgewater$229$287$337$374$418
SemiAnalysis$256$302$347$389$435
Short-seller$218$293$323$359$402
Compounder$279$307$348$382$415
Damodaran$239$314$342$372$433
Equal mixture$240$300$341$376$422
Bear-weighted (50%)$230$297$334$371$416
Market-implied$212$283$330$383$485

Note the market-implied curve is FATTER than the panel in both tails (P5 $212 vs $240; P95 $486 vs $422) - empirical confirmation of the pseudo-independence critique: the panel under-disperses. Trust the market curve for tail sizing.

2 · 12 months

$100$200$300$400$500$600spot
LensP5P25P50P75P95
Stratechery$227$293$358$395$448
Bridgewater$197$271$341$395$461
SemiAnalysis$234$296$360$423$497
Short-seller$179$272$313$363$426
Compounder$265$304$362$414$465
Damodaran$211$310$350$394$489
Equal mixture$211$289$347$400$472
Bear-weighted (50%)$195$281$333$388$457
Market-implied$173$261$322$399$568

3 · 60 months (extrapolated - no lens was asked this horizon)

$100$200$300$400$500$600$700$800$900$1000spot
LensP5P25P50P75P95
Stratechery$171$303$473$588$781
Bridgewater$109$220$368$511$723
SemiAnalysis$185$313$485$698$997
Short-seller$69$177$241$337$479
Compounder$249$340$503$676$879
Damodaran$136$323$424$552$895
Equal mixture$124$263$406$575$859
Bear-weighted (50%)$90$211$323$478$793

60-month curves are a √t + drift extrapolation of a question no lens was asked. Use for shape intuition only, never for sizing. No listed market benchmark exists at this horizon.

4 · Backtests of the framework’s load-bearing theories

4a · "Far above the 200-day = poor forward returns" (Jessica’s regime signal)

Distance vs 200d SMAGOOGL med fwd-6mGOOGL win rateQQQ med fwd-6mQQQ win rate
<-10% (deep below)+20.1%65%-2.3%46%
-10% to 0%+16.5%82%+10.9%70%
0-5% (today: +2.8%)+11.0%76%+8.3%79%
>15% (far above)+7.1%69%+10.1%83%

Verdict: validated for GOOGL, inverted for QQQ. Single-name GOOGL dips below the 200-day have been the best buying regime in 22 years (fwd-6m +20.1% median). But the index shows the opposite at the extreme - deep QQQ breaks mark real bear markets. The unresolved question the adversarial review correctly flags: is the current shock (capex/FCF/antitrust/circularity, sector-wide) more "single-name pullback" or more "index regime break"? If GOOGL falls below its 200-day together with QQQ, the bullish table above is the wrong reference class.

4b · Valuation mean reversion (P/S percentile vs forward returns, 2006-2025, n=20)

Year-end P/S bucketnMedian fwd 1-yrMedian fwd 2-yr
Top 30% (rich) - 2026 is here (9.26x)7+1.9%+0.3%
Middle7+30.9%+31.9%
Bottom 30% (cheap)6+52.5%+70.6%

Verdict: strong signal (corr -0.56/-0.76), and it argues AGAINST the panel’s own P50. Rich-P/S starting years produced ~flat 1-2yr forward returns; the panel’s equal-mix P50 (+9% by Dec-2028) sits above what this signal predicts. Reconciliation: the margin-adjustment (fair P/S ≈ 28.1 x net margin) says today’s 9.26x is only ~13% rich, not 40% - the raw percentile overstates richness because the margin regime changed. Both readings are shown; the honest range is "flat-to-modestly-up," which is precisely the zone where put-spread selling beats holding stock. Caveats: n=20 overlapping-era annual points; part of the correlation is Google’s life-cycle derating, not a stationary law.

4c · Drawdown & recovery - era-split per adversarial fix

Era>20% drawdownsMed depthMed days to troughMed recoveryWorst
Pre-20162-47%222d811d-65% (2008)
2016+6-27%62d117d-44% (2022)
Current (from 5/13 peak)ongoing-21% @ trough 7/2371d?-

Verdict: modern-era drawdowns resolve in ~4 months median, and the current one is tracking the modern pattern so far. The live warning stands: 3 separate >20% drawdowns since Feb 2025 (~4x the base rate) + first-ever negative FCF = two features with zero historical precedent in the sample. History supports buying this dip; history has also never seen this exact stressor.

5 · Adversarial review: what survived, what changed

Attack (ranked)VerdictFix applied
Six lenses = one model + one data pack, not independentSustained - 5 P50s cluster $350-407Bear-weighted mixture added everywhere; market-implied curve treated as the tail benchmark (it IS fatter both sides)
Equal weighting bakes in the bullish answerSustained - edge collapses at 50% bear weightEV now reported under 3 schemes (table below); headline reframed
"Panel beats market" has no mechanismSustainedReframed: this is a priced directional bet (bear ≤ 1/6 weight), plus the structural argument that long-dated single-name skew embeds an insurance premium sellers systematically collect - real but modest, and unverifiable for this specific chain
Backtests: overlap, n=20, regime drift, QQQ inversionPartially sustainedEra-split stats; contradictions surfaced in-line (4a/4b verdicts); thresholds acknowledged as untuned defaults, not optimized
√t scaling wrong for jump/event riskSustained for near tailsMarket-implied curves overlaid at 6m/12m (they price the jumps); 60m flagged as extrapolation; event calendar retained as the real timing driver
Timing triggers underspecified (IV measure, whipsaw, reserve repricing)SustainedTriggers pinned: IV trigger = Dec-2028 ~30-delta put IV > 43% (weekly close basis); price trigger = daily close < $317 with next-day confirm; reserve tranche EV haircut acknowledged

The revised EV table (per contract, Dec-2028, executable fills)

StructureEqual-weight panelBear at 50%Market-implied
340/240+$874+$37-$1,228
310/250+$521+$35-$752
310/240+$604+$62-$870
320/250+$532-$50-$955
300/240+$416-$33-$842

Read the market column as the price of the insurance you’re selling (bid/ask + left-tail risk premium), not as "expected loss" - risk-neutral probabilities overweight crash states by construction. The equal-vs-bear spread is the honest measure of how much of the edge is panel composition. 310/240 is the most weighting-robust structure (positive under both panel schemes).

What this means for sizing: the framework earns conviction for the STRUCTURE (capped spreads, laddered, Dec-2028, GOOGL class, managed by the playbook) and for the FLOOR analysis (valuation floors at $185-235 are twice-tested facts). It does NOT earn conviction that the market is mispricing the left tail - that part is a belief. Size the program to what the belief deserves: max loss across all tranches within the single-digit % of portfolio you’d assign to one directional idea, not the "it’s all +EV" sizing the equal-weight table would tempt you toward.
Method: six persona models (same base model, curated data pack - correlation acknowledged), quantiles time-scaled by log-linear drift + √t; market-implied curves from Breeden-Litzenberger fits of live GOOGL chains (2027-01-15, 2027-06-17); backtests on GOOGL 2004-2026 / QQQ 1999-2026 daily closes (yfinance, stored in data/) and year-end P/S 2006-2025. Adversarial review by an independent agent pass; fixes applied as documented. Probabilities are judgment. Not licensed financial advice.