codepoet.me / market · Investment Journal

Holding a losing stock until you're back to even?

I was. When my gold and copper went into the red, I held on even after they went over their weight caps. While I waited to get back to even, they stayed over the cap — and the price, not I, was deciding when I'd sell. This journal records what I thought and decided on the way to turning that habit into rules, and how it has turned out.

What I learned

Breakeven is not a sell rule.

Without clear criteria, I kept telling myself, 'I'll sell once I'm back to even.' Now, before buying, I write down why I'm buying, what would make me sell, and what I'll monitor.

So how has it gone?

From September 27, 2026 to October 30, my portfolio (tradable holdings, price changes only) moved -7.8%. Over the same period the KOSPI moved -19.1% and the S&P 500, converted to won, -2.4%. So far that puts me ahead of the KOSPI and behind the S&P 500 (in won).

What I hold now, and how much

Weights by holding; amounts and share counts are left out. Some quantities are still the values I entered by hand on September 23, 2026, so later trades may show up late. There's no record of actual holdings before September 27, so earlier points show only the target allocation set at the time.

• Core (stability)
• Satellite (growth)
• Cyclical (waiting)
• Being exited
• Cash / unallocated

What happened after each decision

Price changes from the close on each decision date to the latest close. They don't reflect my actual trade dates, prices or realized results. This is an interim picture; on each review date I revisit the decision and make the call myself.

Market notes

One note per date: what happened in the market (Now), what I'm watching next (Outlook), and how I read it through my principles (My view). Newest first.

• Oct 30, 2026Oct 26 – Oct 30

  The week Alphabet fell 12.6% in a day — assessed with indicators set in advance, not price alone

  A week when big tech swung hard on a single earnings report. The decline was concentrated in a few large tech names, and the US investment-grade corporate spread held between 0.86 and 0.91 percentage points. I didn't sell on price alone; I judged by indicators I'd set in advance — favorable for gold, unfavorable for copper.

How my thinking changed — and the decisions I made

At first I thought a good product made a good stock. Now I believe a share price reflects the company's vision for its future, so I look first at the bottleneck in its way. Below, oldest first, are the changes in my thinking and the decisions I made along the way.

• Core (stability)
• Satellite (growth)
• Cyclical (waiting)
• Being exited
• Cash / unallocated

1. Jan 1, 2026View v1

   Starting from the view that power is AI's bottleneck

   Power is AI's bottleneck

   Target allocation (v1)

   • Berkshire (value-investing holding company) 50%
   • Power value chain 50%

   What I thought

   • As AI spreads, electricity demand rises, and grid capacity to carry power from plants to data centers becomes the first constraint.
   • The portfolio is split between a steady side and a future-growth side.
2. Apr 16, 2026View v2

   Deciding to invest where digital becomes physical

   Investing where digital becomes physical

   • I look for the points where software decisions move real equipment — factory robots, surgical robots, the power grid.
   • In those areas, I look for companies that address critical, urgent problems with products or services that are hard to substitute.

   Target allocation (v2)

   • Real-economy companies 50%
   • Power grid 25%
   • Copper 15%
   • AI power 10%

   What I thought

   • The more common AI becomes, the more value gathers where software moves real equipment, and in scarce resources like power and chips.
3. Aug 4, 2026

   Concentration limits are kept regardless of price

   What I thought

   • When one asset's weight exceeds the limit, I trim it. Even if the price keeps rising afterward, reducing the concentration was the right call.
   • Looking back in September, I should have held that stock longer. Instead of selling all at once based solely on its portfolio weight, I should have trimmed in stages while watching the news and signals.
4. Sep 20, 2026

   Re-examining companies through a 'production gets easy with AI' lens

   What I thought

   • When AI makes things easy to produce, supply floods in. I expect revenue to concentrate in distribution platforms that put products in front of people (app stores, search, retail networks) and in infrastructure that charges by usage (cloud, electricity). Products sold once, and software priced per seat, may find it harder to keep growing.
5. Sep 28, 2026

   Putting sell-by dates on positions I'd been putting off

   Actual holdings (Sep 28, 2026, tradable assets)

   • Gold 25.4%
   • Copper ETF 24.2%
   • AI power ETF 10.5%
   • Cash 10.3%
   • Waste Management (WM) 9%
   • Alphabet (GOOG) 8.8%
   • Power grid ETF 6.1%
   • KBR 1.5%
   • US tech/bond balanced ETF 0.9%
   • Other small positions 3.3%

   What I thought

   • Some positions were still there not because the thesis held, but because I hadn't cleaned them up. Those get a sell-by date.
6. Sep 30, 2026View v3

   Setting my own investment principles for the first time — core and satellite

   A core-satellite portfolio

   • I use a core-satellite approach: the core prioritizes capital preservation (stores of value such as gold, essential-service stocks, safe assets), while satellite positions seek growth through AI-platform companies and the power sector, today's bottleneck in AI.
   • Right now capital preservation comes first, but part of the portfolio still goes to growth.
   • I put the growth portion into companies I expect to benefit as today's supply constraints ease. Example: AI can't scale without enough power for data centers, however many chips exist.
   • I don't buy cyclical assets (commodities, industrials) while the economy is heading down; I buy after signs that it has bottomed.

   Target allocation (v3)

   • Gold / store of value 20%
   • Essential-service companies 29%
   • Safe assets 1%
   • AI power & grid 30%
   • AI platform 10%
   • Search allocation (left empty) 10%

   Actual holdings (Sep 30, 2026, tradable assets)

   • Gold 25%
   • Copper ETF 24.3%
   • AI power ETF 10.7%
   • Cash 10.4%
   • Waste Management (WM) 9%
   • Alphabet (GOOG) 8.8%
   • Power grid ETF 6.1%
   • KBR 1.6%
   • US tech/bond balanced ETF 0.9%
   • Other small positions 3.3%

   What I thought

   • Principles come from accumulated study and experience. When I didn't know enough, I couldn't set any.
   • Diversification makes losses easier to sit through: with gains elsewhere, I feel less pressure to sell a losing position in a hurry. Whether I keep it still depends on its thesis and weight.
   • Whether I can survive everything falling at once is checked separately.
   • The sequence matters for investing too. In power, the bottleneck clears in order: generation → transmission → distribution. I invested in distribution first, but distribution only opens up once there's enough generation capacity. So I look at the current bottleneck first.
   • Buying cyclical assets at the top of the cycle loses money. Assets that move with the same economic variable are treated as one group.
   • I worry that ongoing monetary expansion will erode the purchasing power of cash over time. That concern is why I hold gold, commodities and shares of essential-service companies rather than picking a currency.
   • I find it hard to earn returns from bonds and gold. I treat them as holdings that carry me through periods of volatility.
   • For content companies, I first check who owns the IP. If the creators own it, the company is closer to a distributor or broker.
   • Every holding gets a thesis, an exit rule and indicators to watch.
   • Target weights can change, but each comes with a reason. A number without a reason is not a rule.

   What I decided

   • Zscaler — closing a small position I kept just to watch — Zscaler +19.3% · Nasdaq-100 (QQQ) -6.7% (Oct 30) · the stock sold has beaten the index
   • Which money counts when calculating portfolio weights
   • Pinterest — closing a small position held without a thesis — Pinterest +2.5% · Nasdaq-100 (QQQ) -6.7% (Oct 30) · the stock sold has beaten the index
   • KBR — exiting in two steps because the thesis was wrong — KBR +10.1% · S&P 500 -0.9% (Oct 30) · the stock sold has beaten the index
   • Gold — setting a weight cap and pausing new buys — gold futures -6.9% (Oct 30) · below the decision-date close
   • Currency — not picking between won and dollar, and no FX hedging — won/dollar rate -1.6% (Oct 31) · the won strengthened against the dollar
   • Copper — setting a weight cap and pausing new buys — copper futures -1.5% (Oct 30) · below the decision-date close
7. Oct 28, 2026

   Turning a habit of holding losers into a selling rule

   Actual holdings (Oct 28, 2026, tradable assets)

   • Copper ETF 22.8%
   • Gold 22.3%
   • Alphabet (GOOG) 16.4%
   • Cash 9.9%
   • Waste Management (WM) 9.3%
   • AI power ETF 8.4%
   • Power grid ETF 5.4%
   • KBR 1.7%
   • US tech/bond balanced ETF 0.8%
   • Other small positions 3%

   What I thought

   • Without clear criteria, a losing position hardens into 'I'll sell when I'm back to even.' That's exactly what happened: once I was at a loss, I just held on.
   • Even if the long-term call is right, how much to hold is a separate decision.
   • Anything above the weight cap is trimmed or held based on one market signal chosen in advance for each asset: for gold, 10-year breakeven inflation; for copper, the gold-to-silver ratio — each compared with three months earlier.

   What I decided

   • When to sell a losing position — market signals instead of breakeven
   • Trimming a copper ETF — keep the deadline even below my purchase price? — copper futures -2.8% (Oct 30) · below the decision-date close

Your turn

What sell rule — other than breakeven — would you write for your losing position? Try one line each: why you bought it, and what change would make you sell.

The rules I use now

• If I can't explain it, I don't buy it

  I only invest in companies I can explain in one sentence, and I judge on value, not price.

• Count diversification by risk, not by ticker

  Gains elsewhere take away the pressure to sell a losing position in a hurry, but whether I keep it still depends on its thesis and weight. Positions exposed to the same risk (interest rates, the economic cycle) count as one, even under different tickers.

• Check separately whether I can survive the worst case

  There are times when everything falls together. Whether I have the cash and time horizon to ride that out is checked apart from picking assets.

• Find the bottleneck in the value chain first

  Industries get stuck and unstuck in stages. AI, for example, runs chips → data centers and power → the grid → services. Before buying, I write down which stage the asset is in and which stage is the bottleneck now.

• Write the thesis, exit rule and indicators first

  Every holding gets three notes: why I bought it, when I'll sell, and what I'm watching. If any is missing, I don't add to it.

• There are only three reasons to sell

  ① The weight is over its cap (rebalancing) ② the thesis is broken ③ a better opportunity. A falling price alone is not a reason.

• Every target weight needs a reason

  Weights can change, but each one comes with a line on why it's that size. A number without a reason is not a rule.

How I keep this journal — the path taken and the paths not taken

Every decision with money on the line gets an entry, and I revisit it on a set date. The decisions below follow this format; for some September decisions, the options I didn't choose were filled in looking back in October.

1. Write the expected outcome first

   When deciding, I write one sentence on what 'right' would look like, and set a review date.

2. Record the options I didn't choose

   Writing down only the chosen option makes it easy to credit my own judgment for the result. Comparison needs the rejected options too.

3. On the review date, only facts

   I record what actually happened to both the chosen and rejected options — facts only. If something can't be checked, it stays 'unverifiable'.

4. I make the call on right or wrong

   AI may draft the review, but I decide whether I was right and rewrite it in my own words.

5. Wrong calls change the rules

   Each lesson is kept in three parts — when, what I learned, evidence — and lessons from wrong calls are used to fix the next set of criteria.

Study notes

Long reads that follow one question to the end. Company names are examples of a structure.

• Pricing Power and Business Control — Two Questions for Every Company

  Start with two questions when assessing a business. ① Can it set its own prices (pricing power)? ② Can it operate without relying on someone else's permission or resources (business control)? A company that holds both can sustain its earnings. When others hold both, even strong sales may primarily benefit someone else.

• Can the Space Industry Earn More Than It Costs? Where the Money Comes From and Who Captures It

  Almost all revenue from space comes from services delivered from space for people on Earth: communications, navigation, and observation. Revenue from bringing extraterrestrial materials back to Earth is close to zero, and those materials are unlikely to compete with terrestrial mining costs anytime soon. The real value of space lies less in minerals than in control over the systems that observe Earth, connect people, and provide navigation.

A personal record, not investment advice. Amounts, share counts and portfolio weights are left out.