Research cutoff: Friday, September 4, 2026, around 7:25 a.m. ET.
Price cutoff: Thursday, September 3 regular-session close.
Horizon: roughly the next 1–15 U.S. trading sessions.
Premarket rule: premarket and after-hours moves are context only. They do not complete a catch-up signal.
Today’s answer: no clean setup
Thursday gave several laggards exactly the confirmation test I wanted. Most failed it.
Datadog is still the most interesting relationship on the board because it has badly lagged cloud software while Snowflake just delivered a strong AI/data-cloud demand read-through. But DDOG opened around $220, traded as high as $221, and faded to a $214.76 close. It did beat WCLD by a tiny amount for the day, but not enough to repair the larger 3-, 5- and 10-session deficit or clear yesterday's roughly $224 confirmation level.
UPS is more useful as a lesson than a trade. Wednesday's relative repair was mostly FedEx falling faster. Thursday, FedEx bounced 2.45% while UPS gained only 0.48%, immediately taking back almost two percentage points of the spread. The candidate-led repair rule added after the seven-session replay did exactly what it was supposed to do: it kept a mathematical spread improvement from being mistaken for a real UPS turn.
Super Micro looks superficially tempting because Dell has exploded higher on fresh AI-server earnings while SMCI has not. The problem is that the gap was created by Dell's own earnings shock, while SMCI carries its own financing/dilution and execution baggage. That is not a clean peer-laggard anomaly.
MongoDB has a huge software-relative deficit, but it was created by its own earnings reaction. Chevron trails crude materially, but comparing an integrated oil major with front-month crude ignores a large and persistent beta difference; CVX is also sitting near its 52-week high with its own Venezuela catalyst already in the tape.
And there is a major macro event only an hour after this research cutoff: the U.S. employment report is due at 8:30 a.m. ET. Reuters' consensus is around +56,000 payrolls, 4.1% unemployment and roughly 3.0% year-over-year wage growth. That is another reason not to manufacture a premarket entry from a marginal signal.
Bottom line: no clean U.S. relative-laggard buy before the September 4 open. DDOG remains a watch. UPS, SMCI, MDB and CVX are rejects under the current rules.
Image source: Wikimedia Commons.
Candidate summary
| Name | Genuine relationship | What the screen sees | Status | What would change my mind |
|---|---|---|---|---|
| Datadog (DDOG) | WCLD cloud-software basket; Snowflake is a real demand read-through for cloud data/AI workloads | Large 3/5/10/20-session deficit; Thursday sympathy move did not clear the regular-session reclaim | Watch only | Close above ~$224, then a second session holding the repair while outperforming WCLD |
| UPS | Direct parcel/logistics peer FedEx (FDX) | Old 20-session deficit is smaller; Wednesday's apparent repair reversed Thursday | Rejected today | UPS itself must reclaim ~$106.20 and lead FDX for two regular sessions |
| Super Micro (SMCI) | Direct AI-server infrastructure read-through from Dell (DELL) | Large fresh 3/5/10-session gap, but it was created by Dell's own earnings repricing; SMCI has separate financing/execution issues | Rejected | Dell's post-earnings spread must normalize, then SMCI needs its own close above roughly $40 on company-led strength |
| MongoDB (MDB) | WCLD software basket; database/AI application spending | Large 3/5/10-session deficit, but the break came directly from MongoDB's own earnings expectations reset | Rejected | Fresh base above roughly $399–400 plus two sessions of software-relative outperformance |
| Chevron (CVX) | Crude oil is a major earnings factor; cross-check against XLE/COP | Trails USO across every measured window, but the comparator has much higher oil beta and CVX is already near a 52-week high | Rejected | Crude must stabilize first; then CVX must begin leading integrated/producer peers rather than merely "catching" front-month oil |
Relative performance through Thursday’s close
The figures below are candidate price return minus comparator price return, in percentage points. Negative means the candidate lagged its comparator over that window.
| Candidate vs comparator | 1 session | 3 sessions | 5 sessions | 10 sessions | 20 sessions |
|---|---|---|---|---|---|
| UPS vs FDX | -1.98 pp | +0.24 pp | +0.08 pp | +1.43 pp | -2.38 pp |
| DDOG vs WCLD | +0.15 pp | -6.66 pp | -8.97 pp | -12.24 pp | -18.35 pp |
| SMCI vs DELL | -2.56 pp | -11.66 pp | -10.88 pp | -15.02 pp | +10.91 pp |
| MDB vs WCLD | -0.08 pp | -12.47 pp | -10.12 pp | -13.13 pp | -8.10 pp |
| CVX vs USO | -0.88 pp | -3.76 pp | -3.51 pp | -2.91 pp | -7.86 pp |
The table shows why the biggest negative number is not automatically the best trade.
DDOG has the cleanest unexplained-looking recent lag, but it still needs price confirmation. SMCI's short-window gap looks dramatic until the 20-session number shows SMCI is actually still ahead of Dell over that longer window — a clue that this is a fresh Dell earnings event, not a long-standing disconnect. MDB's gap has a direct company-specific explanation. CVX's gap partly reflects using a commodity instrument with much higher oil sensitivity as the comparator. UPS's Thursday reversal shows why a candidate must lead its own repair.
Seven-session walk-forward: the stricter confirmation rule stays
I replayed the same selection logic at each of the prior seven completed U.S. decision cutoffs — August 26, August 27, August 28, August 31, September 1, September 2 and September 3 — using only the prices and information that were available at each cutoff.
I am not publishing raw hit rates. The useful part is the failure mode and whether the rules caught it.
The replay keeps pointing to the same trap: one-day relative repair is too easy to manufacture. A candidate can look better because its comparator falls, because a peer has a company-specific event, or because a one-day sympathy move briefly narrows a much larger deficit.
UPS is the cleanest live example. On September 2 UPS rose only 0.34% while FDX fell 2.57%. A simple spread screen would have called that repair. The candidate-led rule refused to do so. On September 3 FDX rebounded 2.45% while UPS gained 0.48%, and almost two points of relative performance disappeared immediately.
DDOG showed a similar reason to require a second close. It had brief relative strength earlier in the replay, then failed to sustain it. Thursday's Snowflake sympathy move also produced only +0.15 percentage point of one-day outperformance versus WCLD and no reclaim of the trigger area.
The event filters also did their job. MongoDB's September 2 drop is frozen as a company-event repricing rather than reclassified as "cheap software." Dell's earnings-driven jump does not make SMCI an automatic catch-up trade. Those are exactly the situations that contaminate naive pair screens.
So I am not adding another filter today. The replay does not justify loosening the method, and it does not show a new failure mode that yesterday's candidate-led plus second-session confirmation rules fail to handle.
1. Datadog: still the best watch, but Snowflake did not produce confirmation
Why the relationship matters
Datadog and Snowflake are not identical businesses, so I do not treat them as a literal pair trade. The economic link is still real: both benefit when enterprises move more data and workloads into the cloud, build AI applications, and spend on the infrastructure needed to operate those applications at scale.
That is why the main comparator is WCLD, with Snowflake's earnings used as a demand read-through rather than as the benchmark itself.
Datadog's fundamentals remain healthy. Q2 revenue was $1.12 billion, up 36% year over year. Free cash flow was $279 million, and the company had roughly 4,720 customers with $100,000 or more in ARR, up 23% from a year earlier.
Snowflake then gave the group a real catalyst. It raised its fiscal-2027 product-revenue forecast to $6.07 billion, said AI products represented roughly half of the acceleration in growth, and triggered a broad round of analyst target increases.
What happened to the catalyst
DDOG did respond Thursday — just not strongly enough.
The stock opened around $220, touched roughly $221, and faded to $214.76. WCLD also rallied, leaving DDOG only about 0.15 percentage point ahead of the basket for the session.
For a stock that is still down roughly 9 points relative to WCLD over five sessions and more than 12 points over ten, that is not catch-up confirmation. It is a bounce.
Entry / observation zone
| Item | DDOG plan |
|---|---|
| Observation zone | Roughly $210–$224 |
| First confirmation | Regular-session close above ~$224 |
| Cleaner setup | Second session holding the reclaim while DDOG beats WCLD again |
| Invalidation | Break below the recent ~$208.20 low without a quick regular-session reclaim |
Key risks
The stock still carries a premium software valuation, the 20-session deficit includes some older company-specific de-rating, and Snowflake's consumption/data-cloud economics are not identical to Datadog's observability/security model. Today's jobs report can also dominate every software-relative signal through rates.
Analyst / news / community sentiment
The sell side remains constructive. StockAnalysis' S&P Global aggregation shows 46 analysts, a Strong Buy consensus and an average target around $285. Recent target changes include both bullish $300-area calls and some reductions, which is a healthier picture than pretending every analyst agrees.
Searchable retail/community discussion is much less useful right now. There is no fresh, broad DDOG consensus I would trust enough to make it part of the signal, so I give community sentiment essentially no weight today.
Verdict: watch only. Snowflake gave DDOG a fair test; DDOG has not passed it yet.
2. UPS: yesterday's false repair failed exactly as the replay warned
Why the relationship matters
UPS and FedEx remain one of the cleaner large-cap operating comparisons in U.S. equities. They share parcel volumes, business shipping, international trade, air and ground network utilization, labor, fuel and pricing discipline.
UPS itself is not in a financial crisis. Q2 revenue was $22.8 billion, adjusted EPS was $1.76, and management raised full-year revenue, operating-profit and EPS guidance. The Amazon volume glide-down and related network reconfiguration were also described as progressing to plan.
Why I reject it today
The price confirmation failed.
UPS's Wednesday relative repair was mainly FDX weakness. Thursday's rebound in FDX reversed most of it. The 20-session UPS deficit is now only 2.38 points, which is not large enough to force a catch-up thesis on its own.
There is also a structural reason to be less confident that the UPS/FDX relationship will behave exactly as it did historically. Amazon keeps expanding its own logistics network, and its internal projections reportedly expect it to handle an even larger share of U.S. package deliveries over the next few years. That affects UPS and FedEx differently depending on customer mix and network economics.
Entry / observation zone
| Item | UPS plan |
|---|---|
| Observation zone | Roughly $102–$106.20 |
| Revisit only if | UPS closes above ~$106.20 with positive absolute price action |
| Cleaner setup | Second regular session leading FDX, not merely surviving a peer selloff |
| Invalidation | Break below roughly $101.80 without a quick reclaim |
Key risks
Parcel demand, fuel, labor economics, Amazon's growing network and the possibility that FDX-specific moves distort the spread all matter. The most important risk to this strategy is simple: the old peer relationship may be getting less stationary.
Analyst / news / community sentiment
Analysts are mixed rather than uniformly bullish. StockAnalysis shows 29 analysts, a consensus Buy and an average target near $116, but recent calls include Hold ratings from Bank of America and Wells Fargo alongside Buy/Outperform calls elsewhere.
Retail discussion is similarly divided between a valuation/healthcare-logistics bull case and skepticism around domestic volume and Amazon. Most searchable threads are stale enough that I would not trade from them.
Verdict: reject today. A spread that repairs because FedEx falls is not a UPS catalyst.
3. Super Micro: Dell's blowout is a read-through, but not a clean pair signal
Why the relationship matters
Dell and Super Micro are both major suppliers of AI-server infrastructure. They compete for large GPU clusters, rack-scale systems, networking, cooling and enterprise/hyperscale deployments. A sharp Dell demand surprise is therefore economically relevant to SMCI.
Dell's latest report was genuinely exceptional. It lifted its full-year revenue outlook to roughly $192 billion from $167 billion and adjusted EPS guidance to about $25.50 from $17.90. Q2 revenue reached about $47 billion, up 58% year over year, with an AI-server backlog around $95 billion and roughly $60 billion of new orders in the quarter.
That is strong evidence that AI infrastructure demand remains enormous.
Why I reject the catch-up trade
The short-window spread is almost entirely Dell-event-created.
SMCI trails DELL by roughly 15 percentage points over ten sessions, but it is still ahead by about 11 points over twenty sessions. That is not the shape of a quietly ignored laggard. It is the shape of one peer being repriced after a company-specific earnings shock.
SMCI also fails the "no separate baggage" rule. The company has been using large equity-related financing to fund growth, and its latest results show a much larger diluted share base alongside substantial debt/convertible notes. Those funding needs matter in a strategy that explicitly avoids dilution and balance-sheet noise.
Entry / observation zone
| Item | SMCI plan |
|---|---|
| Observation zone | Roughly $36.50–$40 |
| Revisit only if | Dell's post-earnings spread normalizes and SMCI closes above ~$40 on its own news/volume |
| Cleaner setup | Two sessions of SMCI-led outperformance versus both DELL and a broader AI-server basket |
| Invalidation | Fresh financing/dilution pressure, material legal/compliance news, or a break below the recent mid-$35s base |
Key risks
Financing/dilution, working-capital intensity, historical compliance/accounting controversy, customer concentration and the fact that Dell and SMCI do not have identical business models all weaken a simple catch-up assumption.
Analyst / news / community sentiment
Community interest is very high and very noisy. Current SMCI Reddit discussion is split between "cheap pure AI infrastructure" arguments and recurring concern about dilution, governance/history and the stock's inability to hold rallies. A recent Dell-vs-SMCI thread also correctly points out that Dell's cash-flow profile and broader business mix are materially different.
That debate is useful as a risk map, not a buy signal.
Verdict: reject. Dell's earnings validate AI-server demand; they do not erase SMCI-specific financing and execution risks.
4. MongoDB: strong results, but the market explained the lag for us
Why the relationship matters
MongoDB belongs in a cloud/software comparison because Atlas is a major managed database platform used by modern application and AI workloads. WCLD is imperfect, but it is a reasonable broad cross-check for software risk appetite.
MongoDB's quarter was not weak. Q2 fiscal-2027 revenue was $771.8 million, up 30%, Atlas grew about 29%, and the company raised its full-year revenue outlook.
Why I reject it
The underperformance has a direct company-specific explanation.
Investors wanted a clearer reacceleration in Atlas. Instead, Atlas growth remained around 29% for another quarter, and the stock fell sharply even though headline earnings and revenue beat expectations and guidance moved higher.
That is an expectations reset, not unexplained peer lag.
The analyst reaction reinforces the point. Several firms raised targets after the quarter, while others cut targets or stayed neutral. The disagreement is precisely about the durability and quality of Atlas growth, not whether investors somehow failed to notice that software was strong.
Entry / observation zone
| Item | MDB plan |
|---|---|
| Observation zone | Roughly $375–$400 |
| Revisit only if | Close back above ~$399–400 after the earnings gap |
| Cleaner setup | Two sessions outperforming WCLD while holding the post-earnings base |
| Invalidation | Break below the post-earnings ~$375.50 low or another negative Atlas/guidance revision |
Key risks
High valuation, consumption volatility, competitive cloud databases, and the market's sensitivity to Atlas reacceleration all remain live. A large analyst target does not neutralize an earnings-day expectations reset.
Analyst / news / community sentiment
StockAnalysis shows 41 analysts, a consensus Buy and an average target around $453. Post-earnings target changes are mixed: some firms raised targets materially, while RBC, UBS, BMO and others cut theirs. That is a more informative picture than the consensus label alone.
Fresh community discussion is sparse and lower quality than the sell-side debate, so I give it little weight.
Verdict: reject. The lag is explained by the company's own earnings narrative.
5. Chevron: the oil lag is real mathematically, but the beta relationship is wrong for this screen
Why the relationship matters
Oil prices are obviously important to Chevron's upstream earnings. A sustained crude move can eventually flow into cash flow, asset values and analyst estimates.
But USO is not a clean one-for-one comparator for an integrated major. Front-month crude has much higher direct commodity beta, while Chevron also has refining, chemicals, LNG, long-cycle projects, hedges, capital allocation and company-specific execution in the stock.
Why I reject it
CVX trails USO over every measured window, including almost 8 points over twenty sessions. Normally that would get my attention.
The context kills the simple catch-up thesis. CVX closed at $211.32, already close to its $214.71 52-week high, and has its own major fresh catalyst: Chevron announced plans to invest more than $7 billion in Venezuela over five years and more than double production there to around 600,000 barrels a day.
Analysts have already reacted with multiple target increases. In other words, the stock is not being ignored. The remaining spread mostly says crude has moved faster than a lower-beta integrated oil major.
Entry / observation zone
| Item | CVX plan |
|---|---|
| Observation zone | Roughly $207–$214 |
| Revisit only if | Crude stabilizes rather than spikes, while CVX begins leading XLE and large producer peers |
| Cleaner setup | Two sessions of CVX-led relative strength with the commodity no longer doing all the work |
| Invalidation | Another oil spike with CVX simply tracking integrated peers confirms the beta mismatch rather than a catch-up anomaly |
Key risks
Oil can reverse violently around geopolitical headlines, Venezuela projects carry political/execution risk, and integrated-major earnings do not move one-for-one with spot crude.
Analyst / news / community sentiment
StockAnalysis shows a Buy consensus and an average target around $221, with fresh target raises to roughly $230–$243 from several firms after the Venezuela announcement. That is supportive fundamentally, but it also means the catalyst is already well recognized.
There is no community signal strong enough to override the factor-beta problem.
Verdict: reject. A lower-beta equity does not owe a front-month commodity ETF a full catch-up move.
Other relationships screened and rejected quickly
A few other areas looked interesting but failed before the main table.
Gold miners do not qualify: GDX and major miners have already participated strongly in the gold move rather than lagging it.
Airlines are not a clean catch-up trade against one another while crude and jet-fuel costs are moving sharply; the factor shock affects the whole group.
Semiconductor equipment still fails the uniqueness test. KLA, Lam Research and Applied Materials have been weak together, which looks more like a subindustry rotation/policy issue than one company being left behind.
HPE has its own earnings, margin and supply commentary dominating the tape. Broadcom is also in an earnings-event window. Neither belongs in a clean peer-laggard screen today.
That is the point of the process: rejecting contaminated gaps is more valuable than filling a daily quota.
Market context: the jobs report comes first
The August U.S. employment report is due at 8:30 a.m. ET. Reuters' current economist survey expects roughly:
- Nonfarm payrolls: +56,000
- Unemployment rate: 4.1%
- Average hourly earnings: about +3.0% year over year
Thursday's market rally was helped by Federal Reserve Governor Christopher Waller pushing back against an immediate rush toward another rate hike. That reduced rate-hike expectations, but today's payroll and wage data can reprice yields very quickly.
For a short-horizon catch-up strategy, that matters because relative spreads can look "fixed" for one hour simply because duration, cyclicals or commodities all move together after macro data.
So even DDOG — the one name still on the watch list — needs a regular-session, post-data confirmation. I would rather miss the first few percentage points than call a macro squeeze a repaired company relationship.
What I would actually watch after the data
The cleanest information today is not a list of buys. It is a set of tests.
DDOG: can it reclaim roughly $224 and outperform WCLD after the jobs reaction, not just before it?
UPS: can it produce two sessions where UPS itself rises and leads FDX? Until then, the old spread is not actionable.
SMCI: does Dell's post-earnings lead start to normalize without another SMCI-specific funding or compliance headline? If not, there is no reason to force the pair.
MDB: can it hold the earnings low and build a fresh base above $400? That would be a new setup, not a "catch-up" on the old gap.
CVX: does it begin to lead other integrated/producer equities after crude stops accelerating? If not, the USO comparison is simply the wrong beta.
The discipline here is deliberately boring:
A relative deficit becomes interesting only when the economic relationship is intact, the lag is not explained by company-specific news, and the candidate itself starts repairing the spread. Today I do not have all three conditions in one liquid U.S. name.
Sources
Checked September 4, 2026.
Market / macro
- Reuters — U.S. job growth expected to rebound in August; unemployment seen at 4.1%
- Reuters — Nasdaq, S&P 500 futures climb ahead of key jobs report
Datadog / Snowflake / cloud software
- Datadog — Q2 2026 financial results
- Reuters — Snowflake shares surge as AI demand powers growth and lifts outlook
- StockAnalysis — DDOG price, analyst consensus and current news
- MarketBeat — Software names among the most upgraded stocks
- StockAnalysis — WCLD
UPS / FedEx
- UPS — Q2 2026 earnings
- Reuters — Amazon opens its logistics network to outside businesses
- Business Insider — Amazon's internal delivery-network projections
- StockAnalysis — UPS analyst forecast
- StockAnalysis — FDX
Dell / Super Micro
- Reuters — Dell shares gain after strong AI-server demand boosts annual forecast
- Super Micro — Q4 and full-year fiscal 2026 results
- Reuters via Yahoo Finance — Super Micro plans equity-related financing for AI-server orders
- Reddit — Recent SMCI discussion on the sub-$40 tape, dilution and historical risks
- Reddit — Recent Dell vs SMCI earnings discussion
- StockAnalysis — SMCI
- StockAnalysis — DELL
MongoDB
- MongoDB — Q2 fiscal-2027 financial results
- MarketWatch — MongoDB shares tumble despite strong earnings as Atlas growth disappoints
- Investing.com — MongoDB post-earnings expectations reset
- StockAnalysis — MDB price, analyst revisions and current news
Chevron / oil
- Chevron — Venezuela expansion announcement
- Reuters — Chevron records highest quarterly profit in six years
- StockAnalysis — CVX price, analyst targets and current news
- StockAnalysis — USO
Price-history references
Market research only, not individualized financial advice. Relative-laggard trades can fail abruptly when a peer relationship changes, a company-specific event explains the gap, or macro data moves the whole factor. Observation zones are not guarantees or automatic entries.
