Twenty-nine new buy signals and not a single sell: the week that closed on Friday pushed everything one way. Below are twelve of them, chosen as the most representative, and the horizon throughout is weekly swing trading. Two families hold them together: software and IT services, and European defence.

Wednesday split Wall Street in two. The Dow Jones set a new all-time high and the Russell 2000 ran 1.8% at the open, while the Nasdaq shed 0.8%, breaking four straight gains, and the S&P 500 pulled back after touching an intraday record. This was not money leaving the market: it was money changing sector.

The direction reads in the sectors. Healthcare and materials led at 1.27% and 1.23%, energy trailed at -2.07%, communications and utilities lost around a point each. Technology, which is almost a third of the US index and half of the Nasdaq 100, lost half a point: small in itself, very large when it is the only engine switching off. On a day like this price action trading on single names matters more than the index, which is why the cards below look at stocks.

Our model came into the session with the ledger in good order on semiconductor stocks, and that is the part worth telling. The sell on Arm HoldingsShort is worth 12.91%, on MarvellShort 10.51%, on TeslaShort 15.32%, on AppLovinShort 12.43%. On Astera LabsShort the short recovered almost fifteen points in a single session, from -19.12% to -4.88%. We remain in the red on BroadcomShort and NvidiaShort, and we say so. On the buy side AMDLong is up 96.72% since the April signal and Arista NetworksLong 16.29%.

The star of the session was ShopifyLong. Revenue of 3.58 billion dollars, up 34% year on year against 3.45 expected, earnings per share of 0.42 dollars, 115.6 billion in gross merchandise volume and third-quarter guidance well above expectations: Shopify stock closed around 17% higher, the best session in e-commerce. Our model has been long since 3 July and the position is worth 20.74%. Shopify stocks were the most watched name of the day.

Healthcare did well too, and it was the sector that carried the day. Eli LillyLong rose 4.86% with revenue up 48% on diabetes and obesity drugs and full-year targets raised to 85-87 billion; AmgenLong added 4.57%. We are long both, with 16.41% and 14.82% earned since the signal, and Eli Lilly stocks led the healthcare tape. DisneyShort gained 3.65% with quarterly profit above estimates, parks up 10% and a film past a billion at the box office: there our weekly signal is a sell, and the day went against us.

On travel the ledger is the strongest on this card: BookingLong up 6.56% with 20.51% since the signal, ExpediaLong up 2.44% with 21.64%. Two positions open for weeks that walked through earnings without flinching, and that is the kind of trade this model looks for. MercadoLibreLong passed 10 billion in quarterly revenue for the first time with 50% growth, but the operating margin fell to 6.7% from 12.2% and the stock slipped after the close.

On the other side of the session sits a group of companies that beat estimates and were punished anyway. AMDLong lost another 7.04% in the second day of reaction to its quarter, with capital spending almost tripled squeezing free cash flow. Western DigitalShort fell more than 5% in the session and another 11% after the close despite revenue up 44% and a 54.4% gross margin; SandiskShort followed at -5.40%. AppLovinShort dropped around 19% after the close on third-quarter guidance judged weak, and AppLovin stocks were among the day's biggest losers.

What links them is not revenue, it is the cost of producing it. Capital spending, legal provisions, free shipping: the market has stopped reading the top line and started asking what it takes to get there. It is the same question hanging over artificial intelligence spending, and it bears directly on ai trading and on whoever builds the infrastructure it runs on.

For our model this card reads in our favour: we are short Western Digital with 3.68%, short Sandisk, short AppLovin stock with 12.43% and short UberShort with 18.71%, and all four lost ground on Wednesday. On AMD we are long since April and the signal stays open: after two heavy sessions the gain since the signal eases to 96.72%, yet the weekly Point of Inversion was never even approached.

The geopolitical thread pulled again. Houthi rebels claimed a missile attack on a Saudi tanker off Yanbu in the Red Sea, and crude rebounded after two sessions of collapse, with Brent heading towards 80 dollars. The nervousness went straight into safe havens: gold rose more than four points on our reading, silver passed 61 dollars and copper touched a two-month high.

On stocks the move was immediate. Barrick MiningLong gained 7.43% and NewmontShort 6.71%, both with a move worth about twice their typical swing. Our model is long Barrick Mining and short Newmont: on the first the day worked for us, on the second against, and the short narrows to -4.59%. Barrick reports on Monday, so the system has already cut the size by itself.

Two things are worth keeping apart. Gold trading these days is born of an event outside the chart, and outside events fade as fast as they arrive: our sell on the gold ETF, opened in mid-March, has walked through plenty of sessions like this one without the weekly level ever being touched. That is why the model watches Friday's close rather than Wednesday's reaction.

The Milan market held its highs, with the FTSE MIB up fractionally and a session that passed without shocks. The notable fact is UniCreditLong overtaking Intesa SanpaoloLong by market capitalisation: the former is now worth about 128 billion euro against 119 for its rival, a measure of how far the rerating of Italian banking has rewarded it. We are long both, and they are among the oldest and most profitable positions on our book.

Leonardo stock added another 1.54%, with defence still the strongest structural theme on the market: it is the fourth gain in five sessions and the signal agrees across all three horizons — monthly, weekly and daily. Among the other names we follow, FerrariLong and FincantieriLong stay on the buy side, while on STMicroelectronicsShort and PrysmianShort our model sits on the sell side: on Italian semiconductor stocks that position is worth 12.77% since the mid-July signal.

Enel stock remains one of the most followed names on the market and the signal is a buy; on EniLong we have been long since 20 July, and Wednesday's rebound in crude is the first piece of good news after three sessions in which European oil gave ground.

In London the FTSE 100 closed broadly flat but with two stories beneath the index. GlencoreLong rose 4.12% after half-year profit jumped 86%, driven by commodity trading on geopolitical volatility, with a special distribution and a 500 million dollar buyback. AstraZenecaShort gained 2.52% after a deal with Bristol Myers Squibb was ruled out: there our signal is a sell worth 4.43%, so the day went against us, though the weekly level was never touched. AstraZeneca stocks were among the FTSE's main contributors.

In Frankfurt Siemens EnergyShort posted a solid third quarter, with net profit of 1,188 million euro and full-year targets confirmed towards the top of the range. The most delicate case is Novo NordiskShort, down about 5% despite raising guidance and delivering the strongest US launch ever for a drug in its class: the market looked at pricing pressure, which erodes the quality of that growth. We are short and the position has just returned to flat.

Among European banks the day was heavy instead: HSBCLong lost 4.7% the day after half-year profit rose 23%, judged insufficient on the buyback, and PrudentialShort fell 6.4%. The two are not on the same side for us: we are long HSBC, and that is the second session against us in a row, while on Prudential the weekly signal turned to a sell this very week, so Wednesday’s drop found the model already on the right side.

Two readings on the US labour market, both pointing to a slowdown. Private payrolls in July came in at 44,000 against 75,000 expected, sharply below the revised 95,000 of June, and the services survey slipped to 54.1 from an expected 54.5: still expansion, but cooling. The combination strengthens rate-cut expectations and keeps the dollar below 100.

On stress indicators the day was one of retreat: equity volatility fell 4.2% and bond volatility 5.1%. The number worth watching is tail risk, back up 5.5% after Monday's 9.7% collapse. It measures what the market pays to protect itself against extreme events, and its rebound on the eve of the jobs report says that whoever dumped insurance is buying it back.

Everything hangs on Friday's official employment figure, which the options market prices as the event of the week. For any trading strategy built on weekly swing trading the practical consequence is simple: positions are judged on Friday's close, and that close comes after the number.

When money changes sector, the most direct way to be in it is through baskets. Here are the five that mattered on Wednesday.

The gold ETF is the protagonist: up 4.1% in a session, a move worth twice its typical swing, built entirely on Red Sea tension. Our model has been short since mid-March and the day went against us, though the weekly level that would close the trade remains far away. The US healthcare ETF led the day at 1.27%, and it is the basket that holds Eli Lilly and Amgen: we are long, and here structure and movement agree. The US materials ETF follows at 1.23% and carries Barrick and Newmont.

On the other side, the semiconductor ETF and the US technology ETF gave back part of Monday's surge, the latter at -0.53%: we are short both, and the weekly picture still reads unfavourable despite the flare at the start of the week. Last comes the US energy ETF, worst on the grid at -2.07% — while still the best of the month at almost 10% — because it is handing back the geopolitical premium built up in July.

The bulk of the wave has passed: twelve names from our universe report between today and Monday. Next to each one you'll find our model's position, so you know which side we are on when the number lands.

Thursday 6 · AirbnbLong · ConocoPhillipsLong · FiservLong · Unity SoftwareLong · Deutsche TelekomLong · UnipolLong · PetrobrasLong

Friday 7 · Munich ReLong · AllianzLong · Take-TwoLong

Monday 10 · Rocket LabShort · Barrick MiningLong

Earnings around the corner on a fresh entry mean gap risk, and our system cuts the size by itself when results fall within three weeks. The measurement made in July says earnings do not move average returns, they widen dispersion: hence halving the position rather than skipping it.

The four European baskets closed close to one another, and that in itself is information: none diverged from the rest. CAC 40Long gained 0.11%, FTSE 100Long was flat, DAXLong and FTSE MIBLong lost 0.23% and 0.20%. Our signal is a buy on all four.

The difference shows in duration and return rather than in the day. On the Italian basket the signal has been open since April and is the most profitable of the four; on Paris and Frankfurt the positions are younger. What I watch is the distance from each weekly Point of Inversion: while it stays wide on all four, the week of US data ahead is not enough to change the European picture.

It is worth noting that the European session was compressed by the wait for the US jobs number rather than by any continental headline: when four baskets move less than half a point in the same direction, they are usually waiting for something happening elsewhere.

The comparison between the two US baskets tells the day better than any commentary. SPYLong, the ETF on the S&P 500, touched a new all-time high at the open and gave it back, closing at -0.20% and staying less than 1% below its record. QQQShort, the ETF on the Nasdaq 100, lost 0.90% and remains 4.19% below its own.

The reason is compositional and purely arithmetic: technology is 50.5% of the Nasdaq basket and 32.9% of the S&P 500 one, while financials are 0.24% of the first and 12.6% of the second. When money leaves technology for healthcare, materials and banks, one basket has somewhere to go and the other does not.

Our model is long SPY since 6 April with 13.29% earned, and short QQQ since 20 July. On the latter the game is open: Wednesday's close came back below the weekly Point of Inversion by 0.46% after passing above it two days ago. The model works on weekly closes, so it decides on Friday — and Friday brings the jobs number.