Portfolio allocation research

The candidate uses dependence-robust CVaR and is not an order ticket.

The allocation uses the validated cached Schwab snapshot identified above. The solver starts from all cash; live positions and margin debit are intentionally excluded from the candidate trade deltas. The decision uses the coherent developing-candle barrier. The independent and estimated-correlation solutions are ablations. Terminal direction mass uses the validated calibration artifact e0dfe2f435db647b4dbac55606972811bdda69cc6c0e67281d343c82e29e8d24; magnitude and cross-symbol dependence remain explicit modeling assumptions.

Suggested shares · research candidate

Suggested cash$292,488

Gross exposure2.00x

Net exposure50.3%

Three-day turnover$1,174,701

Worst 95% CVaR loss$13,193

Symbol	Suggested	Trade	Exposure	Weight	Model mean	MC radius	Side
MU	0	+0	$0	0.0%	-4.1%	±2.8%	flat
SNDK	-63	-63	-$113,038	-19.2%	-14.8%	±3.8%	short
NVDA	0	+0	$0	0.0%	-2.7%	±0.8%	flat
PLTR	0	+0	$0	0.0%	-4.9%	±1.4%	flat
MRVL	-168	-168	-$40,809	-6.9%	-7.5%	±3.9%	short
INTC	-104	-104	-$11,327	-1.9%	-7.1%	±3.0%	short
ADBE	806	+806	$200,630	34.1%	3.7%	±0.9%	long
IBM	278	+278	$64,007	10.9%	4.0%	±1.9%	long
GOOGL	0	+0	$0	0.0%	-0.8%	±0.8%	flat
AAPL	-702	-702	-$235,114	-39.9%	-7.3%	±0.9%	short
MSFT	0	+0	$0	0.0%	-2.2%	±0.8%	flat
TSLA	646	+646	$235,337	40.0%	2.7%	±0.7%	long
TSM	-90	-90	-$39,009	-6.6%	-4.1%	±1.1%	short
AMZN	0	+0	$0	0.0%	-1.0%	±0.5%	flat
AEVA	15,822	+15,822	$235,431	40.0%	13.9%	±1.8%	long
QCOM	0	+0	$0	0.0%	-0.3%	±1.2%	flat
SKHY	0	+0	$0	0.0%	2.1%	±1.8%	flat
SPCX	0	+0	$0	0.0%	-0.9%	±1.1%	flat

Position	Independent	Estimated correlation	Robust suggestion
MU	0	0	0
SNDK	-62	-68	-63
NVDA	0	0	0
PLTR	0	0	0
MRVL	-72	-74	-168
INTC	-628	-120	-104
ADBE	945	876	806
IBM	202	202	278
GOOGL	0	0	0
AAPL	-702	-702	-702
MSFT	0	-76	0
TSLA	599	646	646
TSM	-17	-31	-90
AMZN	0	0	0
AEVA	15,822	15,822	15,822
QCOM	0	0	0
SKHY	0	0	0
SPCX	0	0	0
Cash	$292,822	$292,474	$292,488
Gross exposure	2.00x	2.00x	2.00x
Turnover	$1,174,995	$1,174,540	$1,174,701

Robust allocation under every coupling

Coupling	Mean P/L	5th percentile	Candidate CVaR loss	Current CVaR reference	Profit probability
IndependentFour independent marginal permutations; optimistic diversification baseline.	$87,876	$11,741	$11,263	$0	96.5%
Estimated CorrelationExponentially weighted, 30% identity-shrunk historical correlation coupling; SKHY, SPCX lack sufficient aligned history and are independently paired.	$87,876	$18,254	$6,462	$0	97.9%
Recent RegimeEmpirical rank dependence from the recent regime block.	$87,876	$30,161	-$11,018	$0	98.8%
Prior RegimeEmpirical rank dependence from the prior regime block.	$87,876	$13,408	-$3,447	$0	98.0%
Older RegimeEmpirical rank dependence from the older regime block.	$87,876	$18,276	$4,812	$0	97.3%
High Volatility RegimeEmpirical ranks from the 256 highest-volatility aligned four-hour bars.	$87,876	$31,432	$108	$0	97.7%
Common ShockComonotonic stress: every forecast marginal occupies the same percentile.	$87,876	$40,725	-$31,726	$0	100.0%
Semiconductor ShockMU, NVDA, and TSM share one percentile shock; other names remain independent.	$87,876	$10,982	$13,193	$0	96.5%
Factor ReversalSemiconductor and non-semiconductor percentiles move in opposite directions.	$87,876	$34,918	-$22,119	$0	98.8%

Fixed holdings

Symbol	Account	Shares	Exposure	Reason
Live holdings are outside the all-cash starting-state assumption.

Cost and constraint assumptions

Starting state: all cash.
10% annualized margin debit cost over three days.
1% annualized borrow cost on short notional over three days.
1 basis point one-way transaction charge; no hard turnover or per-name trade cap.
Cash may be deployed and aggregate short and gross exposure may change.
PLTR may be held flat or long, but never short.
META is excluded from optimization because trading it is prohibited.
Absolute limits are 2.0x gross exposure, 75% short exposure, 40% per name, and net exposure between -75% and 100% of individual-account equity.
CVaR is penalized in the objective but is not hard-capped at the starting portfolio's value.
Terminal up/down mass is calibrated; conditional return magnitudes retain the observed Monte Carlo marginal shape.
An optional ticker with a stale or missing calibrated barrier is omitted until aligned.

Forecast provenance

MU	`20260918T215304Z-step5000-mu-0c3da810`
SNDK	`20260918T215140Z-step5000-sndk-5d236960`
NVDA	`20260918T215419Z-step5000-nvda-0dd0ed5e`
PLTR	`20260918T215101Z-step5000-pltr-b24de96f`
MRVL	`20260918T214555Z-step5000-mrvl-817537f0`
INTC	`20260918T215335Z-step5000-intc-b13d5dd2`
ADBE	`20260918T214717Z-step5000-adbe-f7747e7f`
IBM	`20260918T214719Z-step5000-ibm-eb01aa8c`
GOOGL	`20260918T215429Z-step5000-googl-658bbdee`
AAPL	`20260918T215428Z-step5000-aapl-65d477dc`
MSFT	`20260918T215210Z-step5000-msft-9c08d0e6`
TSLA	`20260918T214702Z-step5000-tsla-c3550331`
TSM	`20260918T214553Z-step5000-tsm-395215c3`
AMZN	`20260918T215500Z-step5000-amzn-24642b4d`
AEVA	`20260918T215303Z-step5000-aeva-4b53e7ec`
QCOM	`20260918T214545Z-step5000-qcom-82fe6353`
SKHY	`20260918T215130Z-step5000-skhy-a1c9cc11`
SPCX	`20260918T215254Z-step5000-spcx-c5923a52`

Return to forecasts