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.

On Thursday the seven baskets we follow all moved within half a percent: the S&P 500 basket shed 0.16%, the Nasdaq 100 basket 0.37%, small caps 0.51%, while Paris and Milan added 0.34% and 0.46%. Looking only at the indices you would conclude nothing happened.

A great deal happened underneath. Nineteen stocks in our universe made a move far outside their own typical swing, and the average sign of those moves was -3.8%. DatadogLong lost 19.03%, AppLovinShort 19.66%, PelotonLong 15.57%, FigmaLong 14.80%, Beyond MeatShort 13.50%, Western DigitalShort 13.03%, FastlyLong 12.90%; on the other side UnityLong gained 15.05% and SafiloLong 16.90%. This is dispersion, and it is a condition distinct from both a rally and a correction. Datadog stocks and AppLovin stocks were the day's heaviest losers.

For anyone doing swing trading the change is concrete: the day's risk was not getting the direction wrong, it was getting the name wrong. In a market like this risk management matters more than reading the trend, because the same sector holds a +15% and a -19% in the same session. Our model came into the day short AppLovin stock, short Trade DeskShort, short Dutch BrosShort and short Western Digital, and long Unity stock: four of the five positions worked in our favour.

What links the stocks punished on Thursday is not revenue: almost all of them beat estimates. DatadogLong lost 19.03% on a revenue outlook judged weak. Western DigitalShort fell 13.03% despite a quarter above expectations on the top line, and SanDiskShort followed at -6.81%: this is the reset of expectations across memory after triple-digit gains since January, and semiconductor stocks have been carrying it for three sessions. CloudflareLong grew revenue 36% and raised full-year guidance, yet the stock still gave ground.

The cleanest case is ExpediaLong. A sixth straight quarter with earnings per share above expectations, full-year targets raised, and the stock lost 4.09%. Our model has been long since 22 June with 16.66% earned and the signal has not moved a decimal: what moved was the reaction, not the structure. Same story on SiemensLong, which raised its earnings outlook for the second time and fell 4.49% because orders at its digital industries arm merely met expectations.

On the short side the ledger reads in our favour: we are short AppLovin stock, Western Digital, SanDisk and Trade DeskShort, and all four lost ground. On Datadog stock, Fastly, Figma and Peloton stocks we are long instead and the day went against us: these are positions open for weeks, the weekly Points of Inversion are holding, and we say it as it is.

The star of the session was UnityLong, up 15.05% on a quarter above expectations. It is also July's best-performing new monthly buy, 28.70% since the month-end entry, and our model has been long since April. It has to be said plainly, though, that Thursday's move is worth almost seven times its typical swing: anyone entering after a jump like that pays the full price of the headline. Unity stocks were the most-bought name of the day.

PayPalLong also did well, up 3.19% after two brokers raised their price targets: we are long since 6 July with 29.06% earned, one of the strongest positions on this card, and PayPal stocks have been steady all week. ShopifyLong added 2.22%, consolidating the previous day's earnings jump, with 23.42% since the 29 June signal.

Closing the group is ModernaShort, down 4.26% even though the US regulator approved its mRNA flu vaccine for adults aged fifty and over. There our weekly signal has been a sell since 20 July, and the day proved us right on news that read as positive: a case where chart structure weighed more than the headline.

The Milan market closed at record highs and the day was all about results. The best performer was Banco BPMLong, up 3.93%: a half-year above expectations, full-year profit guidance raised beyond 1.95 billion euros and a promised dividend of at least one euro. Our model has been long since 7 April and the position is worth 35.54%. Worth noting that the numbers had come out the day before and the stock moved 0.52%: the report was priced twenty-four hours late.

Record half-years also for GeneraliLong and MediobancaLong, both released after the close. Generali beat consensus with adjusted net profit of 2.54 billion, up 13.7%, a 500 million buyback and a solvency ratio of 216%; Mediobanca closed with revenue at 1.95 billion, profit at 711 million and return on tangible equity at 14.9%, while moving towards its merger with Monte dei Paschi. We are long both, with 24.87% and 39.19% earned. BPERLong also did well, long since 8 June with 12.86%.

The sour note came from TenarisShort, bottom of the board at -7.08% on a second quarter below expectations: there our weekly signal has been a sell since 22 June, and the day proved us right. EniLong added 1.36% in the wake of the oil and gas sector, with both horizons aligned.

In Germany the main basket closed just above the line but with sharp internal divergences. Deutsche TelekomLong was the best performer at +6.31%: a quarter above expectations, free cash flow guidance raised towards 20 billion euros and a buyback expanded by around 3 billion. We are long since 13 July with 7.52% earned, and Deutsche Telekom stocks led the board. Going the other way, SiemensLong fell 4.49% and InfineonShort 5.7%, sold despite record quarterly revenue.

RheinmetallLong gave up 3.46% in the broad retreat of defence names after the Iran-Oman easing. This is one of the cases where we are long and the day went against us, and the cause is geopolitical: it can reverse as quickly as it arrived.

In London BPLong posted quarterly profit of 5.7 billion dollars and a dividend up 4%, but the absence of a buyback disappointed; the stock still added 2.48% in the wake of the sector. Miners ran, with Rio TintoShort and GlencoreLong lifted by copper strength: on Rio Tinto we are short and the day went against us. In Paris HermesShort rebounded 5.17% after its half-year miss, helped by a broker raising its price target: there too our signal is a sell, and this is a bounce inside a trend our model reads to the downside.

The agreement between Iran and Oman on the coordinates of a navigation route through the Strait of Hormuz eased the tension that had held energy hostage since late February. Crude fell for a third straight session, down 10-12% in five sessions while still sitting well above pre-crisis levels. In our series oil is nonetheless up 2.8% on the day and is the only macro indicator with a genuine trend: +5% on the month and three consecutive weeks of gains.

Energy was the best US sector of the day at +1.48%, and it holds the best monthly reading on the whole grid. OccidentalLong gained 4.14% after a quarter above expectations released post-close the day before, Exxon MobilLong added 2.12% and ConocoPhillipsLong 1.50%, with adjusted earnings per share of 3.24 dollars against 2.85 expected and revenue up 32%. We are long all three, but on Occidental and Exxon the daily signal is still a sell: the sector is rising and the short horizon has not followed yet.

On the macro front there was no shock: equity volatility, the BTP-Bund spread, the dollar and gold were all broadly flat, with gold still up 4.8% on the week. The one uncomfortable reading is tail risk, down 6.6% on the month: very little protection has been bought ahead of the US jobs report, which is out today and is the only event of the week.

Among US sector ETFs the best of the day was XLELong, the energy ETF, up 1.48% and holding the best monthly reading on the grid at +6.44%. Our model has been long since 13 July. At the other end sits XLKShort, the US technology and semiconductor ETF: -0.31% on the session, but more importantly a weekly sell signal since 27 July with the trade down 5.69%. It is the same problem weighing on the Nasdaq basket, and we say so even when it costs us.

XLFLong, the US financials ETF, shed 0.33% but remains the basket with the best structure on the grid: setup quality at 76 and a buy signal active for sixteen weeks with 10.26% earned. XLVLong, the healthcare ETF, closed broadly flat at +0.18% after leading the day before: the rotation into that sector lasted a single session.

In Europe the move of the day belongs entirely to EXV2Short, the European telecoms ETF, up 2.97%: it is the move of one stock, Deutsche Telekom, and our signal on that basket is a sell. EXH1Long, the European oil and gas ETF, gained 1.05% with the best monthly reading on the continent at +6.84%, and there we are long. EXV1Long, the European banks ETF, added 0.54% with relative strength improving: it is the only banking basket across the two continents that still has the weekly structure on its side.

The bulk of the wave has passed and the coming week thins out: seven stocks in our universe report between today and next Thursday. Next to each one you'll find our model's position, so you know which side we are on when the numbers land.

Friday 7 · Berkshire HathawayLong · DiageoLong · Munich ReLong · Take-TwoLong

Monday 10 · Plug PowerShort · Rocket LabShort

Thursday 13 · JD.comLong

The measure we apply to these cases comes from the July season and is worth repeating: entering with the signal already active ahead of results returned 2.12% on average against 2.03% for the control group. Earnings do not move the average return, they widen the dispersion, and Thursday supplied the proof. That is why a report is neither a reason to enter nor a reason to stay out: it is a reason to cut the size.

The four European baskets we follow are all on a buy signal, and that is the first piece of information: while America splits apart, Europe is compact. CAC 40Long and FTSE MIBLong closed at record highs, up 0.34% and 0.46%, both with signals active since 7 April. DAXLong added 0.04% with its signal open since 13 April; FTSE 100Long, the UK basket, added 0.04% with its signal since 7 April.

The basket that diverges from the others is the German one, and the reason lies in its composition: it closed marginally higher only because Deutsche Telekom's +6.31% offset Siemens at -4.49% and Infineon at -5.7%. Beneath a four-hundredths move sits an internal spread of more than ten points between best and worst: the same American dynamic, in a basket that hides it better.

The Italian basket carries the cleanest momentum, which is no surprise after a day of results that saw Banco BPM, Generali and Mediobanca all beat expectations. The UK basket is the flattest of the four: miners pulled, banks disappointed on buybacks, and the result is a draw.

SPYLong, the S&P 500 ETF, has been on a buy signal for seventeen weeks with 13.11% earned and sits 1.07% below its all-time record. The SPY ETF this week reached the third and last take-profit window of the system, in the very week that window expired: all three areas have now been taken, and from here the plan has no targets ahead, only a rising stop, the weekly Point of Inversion, 3.83% below current levels.

QQQShort, the Nasdaq 100 ETF, tells the opposite story. Our weekly signal has been a sell since 20 July and is down 4.45%; the price sits 4.54% below its own record. The number that matters, though, is another one: the weekly Point of Inversion is less than a point above the price, 0.84%, and it coincides to within two hundredths with the upper edge of the daily Ichimoku cloud. A weekly close above that level cancels the sell and reopens the buy.

The difference between the two reads in the weights. The S&P 500 basket carries 12.59% in financials, where our buy has been active for sixteen weeks and is up 10.26%; the Nasdaq 100 basket carries 0.24%. Where the first has a cushion, the second has a void, and on US technology, half the basket, our signal is a sell down 5.69%: exactly the same problem. On the daily, by contrast, both have their Point of Inversion on a buy since 31 July: the short horizon agrees, the weekly does not.