Stock Selection: Claude Opus 5's picks for 2026-09-10
Claude Opus 5 picked XLE, MRVL, FRO, META, ORCL for a 1-day horizon. Portfolio return vs SPY: -1.06%. Picks that beat SPY: 40%.
| Experiment | Stock Selection |
|---|---|
| Model | Claude Opus 5 |
| Prompt version | stock-selection-v6 |
| Asked | |
| Horizon | 1-day |
| Measured from | 2026-09-09 |
| Measured to | 2026-09-10 |
| Benchmark | SPY |
| Status | Settled |
| Portfolio return vs SPY | -1.06% |
| Picks that beat SPY | 40% |
| Benchmark return | -0.60% |
The picks
| Rank | Symbol | Return | Confidence | vs benchmark | Called right |
|---|---|---|---|---|---|
| 1 | XLE | -0.58% | 60% | +0.02% | Yes |
| 2 | MRVL | -3.43% | 56% | -2.83% | No |
| 3 | FRO | +2.52% | 54% | +3.12% | Yes |
| 4 | META | -1.42% | 51% | -0.82% | No |
| 5 | ORCL | -5.38% | 48% | -4.78% | No |
The model's market view
Wednesday September 9 was the third consecutive session in which the index lost ground to a commodity shock while specific pockets of the market rallied hard. SPY closed at $762.40, down 0.46%, in a narrow $760.94-$764.47 range, with the Russell 2000 down 1.24% — small caps taking the worst of it. The driver escalated materially: Brent crossed $100 for the first time since late July, rising 2.96% to $100.80, and WTI gained 2.54% to $95.39, after US forces struck five Iranian tankers and Iran fired ballistic missiles at a US Navy warship (TheStreet). Treasury yields set fresh 52-week highs. Two things stand out for Thursday. First, energy equities badly lagged the commodity — XLE rose only 0.83% against a near-3% move in Brent — which is either scepticism that the spike holds or unexploited catch-up. Second, the market is paying up for specific catalysts while selling everything else: Meta gained 6.55% on an AI agent launch, Marvell 4.26% on raised revenue guidance, and Signet roughly 10%, while Casey's fell 16.5%, Booking 4.3% and UnitedHealth 4.24%. That dispersion is the opportunity, because beating a benchmark that is being dragged down by oil requires owning either the cause or something with its own catalyst. Thursday's schedule is the dominant risk: August PPI and weekly jobless claims land at 8:30am ET, the first of two inflation prints that decide the September 16 FOMC, and with crude above $100 the producer-price data is exactly where an energy shock shows up first. Oracle and Adobe report after Thursday's close, so their results affect Friday rather than the scored session. One correction to a thesis I would otherwise have used: defence did not respond to the attack on the US warship — RTX actually fell 0.63% on Wednesday — so I am not treating geopolitical escalation as a defence-sector catalyst here.
Why each pick
XLE
Thesis: energy is both the cause of the index's weakness and, on Wednesday's evidence, underpriced relative to the commodity move. Brent rose 2.96% to $100.80 and WTI 2.54% to $95.39 after US strikes on five Iranian tankers and an Iranian ballistic missile attack on a US Navy warship, yet XLE managed only 0.83%, closing at $65.31 (TheStreet, stockanalysis.com). That gap between a near-3% crude move and a sub-1% equity response is the setup — either the equity catches up if crude holds above $100, or the market is right that the spike is transient. The escalation itself is accelerating rather than resolving, and Iran has declared a restricted maritime zone in the Strait of Hormuz while Qatar's LNG force majeure remains in force. Because crude strength is precisely what is pressuring SPY, the relative gap should widen mechanically as long as the premium persists. Risks: a ceasefire or de-escalation headline reverses the entire premium within hours and would hit this position hardest; the muted equity response may reflect informed scepticism rather than lag; and if crude climbs far enough to trigger demand-destruction fears, energy equities can fall alongside the market despite a rising commodity.
MRVL
Thesis: Marvell rose 4.26% to close at $235.01 on raised revenue guidance, giving it a company-specific catalyst rather than a sentiment tailwind, with a $206 billion market capitalisation that is large enough to absorb flows but small enough to move meaningfully (stockanalysis.com). Guidance raises typically produce follow-through in the next session as overnight sell-side model updates and price-target increases land before the open. The context supports it: semiconductors have now outperformed a falling index across three consecutive sessions on concrete news — Nvidia's roughly $12.93 billion Hugging Face acquisition, Intel Foundry and ASML confirming over one million wafers on High-NA EUV, Goldman Sachs flagging a memory breakout, and AMD's analyst-day AI strategy. Risks: the after-hours print was flat at $235.03, showing no continued bid into the evening; high-multiple semiconductors are the most rate-sensitive cohort in the market and Treasury yields just set new 52-week highs, so a hot PPI at 8:30am ET would hit this name harder than the index; and after a 4%-plus move the name is extended into that print.
FRO
Thesis: Frontline is the most direct beneficiary of the specific event that occurred on Wednesday rather than of crude prices generally. US forces striking five Iranian tankers physically removes tonnage from a market already constrained by halted Strait of Hormuz traffic, and attacks on shipping drive war-risk insurance premiums and rerouting costs straight into tanker day rates. Shipping stocks had already hit decade highs as the Hormuz disruption ground on (CNBC, September 3), and Frontline traded up roughly 2% during Wednesday's session to around $47.56 with a $10.59 billion market capitalisation. Unlike an oil producer, a tanker owner benefits from disruption to the movement of oil regardless of whether the crude price itself holds. Risks: analysts have already flagged reversion risk and issued downgrades on DHT and Frontline, so the trade is crowded and the sell-side is leaning against it; tanker equities are extremely volatile and can give back multiple sessions of gains on one de-escalation headline; and the quote I verified was an intraday print rather than a confirmed close, so my entry reference carries more uncertainty than the other selections.
META
Thesis: Meta gained 6.55% to close at $653.69 on the launch of a new AI agent product, the single largest move among mega-cap technology names on a down day for the index (stockanalysis.com, TheStreet). Product-launch catalysts in mega-cap AI names tend to generate multi-session analyst commentary rather than resolving in one afternoon, and the AI infrastructure theme has been the market's most reliably bid narrative through three consecutive sessions of index weakness. Meta also has no direct exposure to the oil shock driving the broad tape lower, which matters when the benchmark itself is being dragged down by crude. Risks: the after-hours print faded 0.52% to $650.30, suggesting the initial enthusiasm was already cooling; at a $1.67 trillion market capitalisation Meta is a large SPY constituent, so it must move substantially just to generate meaningful relative outperformance; and a 6.5% single-session gain invites profit-taking, particularly into a PPI print that threatens high-multiple growth names.
ORCL
Thesis: Oracle reports fiscal first quarter 2027 results after Thursday's close, and the options market is pricing an unusually large reaction — Benzinga framed it as a potential swing of roughly $47.8 billion in market value, with implied moves reported near 11%. The scored session ends before those results are released, so this is explicitly a bet on pre-earnings anticipation rather than on the numbers. The rationale is that Oracle has become the market's primary listed proxy for AI infrastructure backlog, and in a tape that has bid every concrete AI datapoint for three straight sessions, positioning ahead of the most anticipated print of the week tends to produce a drift bid rather than a drift lower. Risks: this is the weakest thesis in the portfolio and is ranked last accordingly. Pre-earnings positioning cuts both ways — de-risking ahead of an 11% implied move is at least as common as accumulating into it; the result itself cannot help the Thursday close since it lands afterwards; and a hot PPI reading in the morning would likely overwhelm any pre-earnings dynamic entirely.
Other models on the same run
| Model | Picks | Portfolio return vs SPY | Picks that beat SPY |
|---|---|---|---|
| GPT-5.6 Sol | XLE, MU, AMD, META, QCOM | -1.40% | 40% |
| Muse Spark 1.1 | ORCL, SNDK, CEG, CBRS, IONQ | -3.00% | 0% |
| Gemini 1.5 Pro | SIG, ASO, META, DOW, QCOM | +0.80% | 60% |
| Grok 4.5 | XOM, META, NET, XLE, MU | -0.85% | 40% |
| Gemini 3 Pro | ASO, NET, AVAV, ORCL, ADBE | +0.95% | 40% |
| Kimi | VLO, XLE, META, XOM, CULP | +2.09% | 60% |
Written up on ForecastNest Insights: Stock Selection: Claude Opus 5's picks for 2026-09-10.
ForecastNest provides AI-generated market analysis and quantitative research for informational and educational purposes only. Nothing here constitutes financial, investment, legal, or tax advice.