I bought on the market's worst day in a month

I bought on the market’s worst day in about a month.

Thursday was ugly.

The S&P 500 fell 1.2%. The Nasdaq dropped 2.2%. Alphabet lost 7.1%, and Tesla fell 14.5%.

The market received its first major answer to the AI-spending question—and decided that the scale of the spending and cash burn mattered more than the growth being delivered.

Oil made the backdrop harder.

Brent closed above $100 as geopolitical and shipping risks intensified, while the 10-year Treasury yield moved toward 4.70%.

Higher oil creates inflation pressure.

Higher inflation pressure supports firmer yields.

Firmer yields make investors less willing to pay extreme prices for profits that may arrive years from now.

That was the day I deployed capital.

Ten positions—not one market call

I opened eight new positions and added to two I already held.

The new entries were:

APH — Amphenol: connectors and sensors; the unglamorous hardware that everything else plugs into.
CBOE — Cboe Global Markets: an exchange operator. Volatility is part of its business, and there is currently no shortage of it.
DAL — Delta Air Lines
FCX — Freeport-McMoRan: copper and gold exposure.
FIX — Comfort Systems: mechanical and electrical contracting; part of the physical build-out behind data centres and industrial construction.
JBL — Jabil: electronics manufacturing.
KEYS — Keysight Technologies: test and measurement equipment.
TER — Teradyne: semiconductor and automation testing.

I also added to two existing positions:

GLD: topped up. Gold continues to provide diversification while markets debate AI valuations, inflation and geopolitical risk.
SPY: added broad index exposure rather than increasing my dependence on any single company or narrative.

This was not one large mega-cap technology bet divided into ten ticker symbols.

It was a collection of businesses with different revenue models, customers and economic sensitivities: an exchange, an airline, a copper miner, an industrial contractor, electronics manufacturing, test equipment, gold and broad-market exposure.

Every company first had to clear a quality screen.

Then it had to rank on opportunity.

Then the portfolio construction rules had to decide whether adding it would leave the book too dependent on one outcome.

Diversified does not mean uncorrelated

There is an important qualification.

APH, FIX, JBL, KEYS and TER all retain some exposure to electronics, industrial investment or the physical AI build-out. They are not interchangeable businesses, but neither are they five completely independent bets.

That means I do not count ticker symbols and call the portfolio diversified.

I look through the names to the common economic driver beneath them.

The portfolio is less concentrated in mega-cap platform companies. It has exposure to several different layers of the real economy. But part of the book would still be affected if infrastructure and technology capital spending slowed broadly.

That shared exposure has to be treated as a cluster.

This is the distinction between diversifying company risk and eliminating theme risk.

The first is achievable.

The second is often an illusion.

What Thursday’s market confirmed

While the market repriced AI spending, industrials were the strongest S&P 500 sector, gaining roughly 1.8%. Communication services and consumer discretionary—home to Alphabet and Tesla—were the weakest.

Capital was not reacting uniformly.

It was moving between themes.

That supports the rotation argument I have been developing throughout July. Thursday’s market figures

My Alphabet post-earnings review reached the same conclusion from the company level: Alphabet’s operating business passed the test, but the stock was repriced because the capital bill increased again.

Thursday expanded that lesson across the market.

The AI debate is no longer simply about whether demand exists. Demand is visible.

The argument is now about how much capital must be committed, how long the cash-flow pressure lasts and which parts of the value chain capture acceptable returns.

Oil added a second repricing mechanism. Brent closed near $100.55 after attacks on Saudi tankers intensified concerns around global supply routes. The 10-year yield reached approximately 4.70%. Reuters market recap

That combination—expensive AI investment, higher oil and firmer yields—is difficult for long-duration growth assets.

It does not affect every business in the same way.

That dispersion is where portfolio construction matters.

I did not predict the selloff

On the timing, I want to be completely honest.

I did not predict Thursday’s decline.

These entries came from a scheduled deployment cycle, not from reacting to Alphabet, Tesla, oil or the afternoon headlines.

The process ran when it was scheduled to run.

It happened to run during the market’s ugliest session in about a month.

That is the actual argument for having a system.

The system deploys when the rules say to deploy—not when the mood feels comfortable.

On Thursday, the mood said stay away.

The rules said these companies qualified.

I followed the rules.

That is different from buying simply because the market was red.

A red market is not a setup.

A scheduled process that produces qualified names, defined exposure and portfolio-level controls is a setup.

That is the inverse of FOMO: the criteria existed before the discomfort rather than being invented after the excitement. Lesson 19 explains the decision architecture; Thursday’s deployment is the live case study.

Why every entry was partial

Each position was opened at partial size rather than full size.

Building in tranches means I do not need to identify the exact market low.

A full-sized entry makes a large timing claim, whether the trader admits it or not. It assumes the opportunity is sufficiently developed that most or all of the intended exposure should be committed now.

A partial entry says something more modest:

The company qualifies.
The price is acceptable for an initial position.
The market may still move lower.
Additional capital will require additional evidence or a better level.

That structure matters in a falling market.

It separates participation from certainty.

I can begin building exposure without pretending that Thursday marked the bottom.

Seven positions are already underwater

One day later, seven of the ten positions are below their entry prices.

FIX is down about 5%.

TER is down roughly 4%.

KEYS is down around 3%.

Only DAL is meaningfully positive so far.

That is what buying during a falling market can look like.

The immediate P&L does not prove that the process worked, and it does not prove that it failed.

One day is too short to evaluate a thesis designed for a longer holding period.

If I had needed every position to work immediately, I would have sized them differently.

I did not.

So I have not.

The purpose of partial sizing is not to avoid seeing red. It is to ensure that seeing red does not control the next decision.

Ten small trades can still create one large exposure

My 1% rule limits the planned risk of any single idea.

But ten simultaneous positions make another control more important: aggregate risk.

Several individually acceptable positions can create an unacceptable portfolio if they share the same driver or are likely to fail together.

Lesson 18 made this point in the context of revenge trading: multiple correlated positions can reproduce size escalation even when every trade appears compliant on its own.

This deployment was scheduled, not emotional. But the portfolio mathematics are the same.

The questions are:

How much total capital was committed?
How much planned loss exists across all positions?
Which holdings share the same economic exposure?
What happens if the next earnings reports trigger another broad repricing?
Is there enough cash to respond without breaking the original risk plan?

The answer to the last question matters most now.

Cash remains around 58% of the account.

There is still room to add if prices continue lower and the setups remain valid.

There is also room to do nothing if they do not.

Cash is not evidence that the deployment was timid.

It is what keeps the next decision independent of Thursday’s entry prices.

The next test is already scheduled

The AI-spending debate is not finished.

Meta and Microsoft report Wednesday. Apple and Amazon report Thursday.

Thursday may have been the first major repricing of this earnings cycle—not necessarily the last. Earnings calendar

That does not mean I should delay every decision until the calendar is empty. There is always another report, inflation release or geopolitical risk ahead.

It means the portfolio must be built to survive new information.

Partial positions do that.

Theme limits do that.

Cash does that.

Defined invalidation does that.

Prediction does not.

What the process actually proved

On Wednesday I wrote:

I do not predict the print. I trade what it does to price.

Price came down.

The scheduled process found qualifying opportunities.

I bought partial positions.

Seven went immediately underwater.

Cash remained available.

There was no prediction, no dramatic market call and no need to identify the bottom.

The value of the system is not that it made Thursday comfortable.

The value is that discomfort did not get a vote.

Rules before emotion. Diversification before concentration. Cash retained for what comes next.

This is a record of my process and opinions, not investment advice. I hold positions in securities mentioned or related. Copy trading involves risk, including loss of capital. Past performance is not an indication of future results.