01
Keep the original research.
I keep the original paper and date each correction or change in view. An update explains what changed, what still holds and why I reached a different conclusion.
Research & measurement
How I review research
and measure results.
You should be able to read my original view, see what changed and follow the reasoning. Any performance record needs rules set before the result is known.
Methodology status / Draft
The model portfolio has not started.
These are the planned rules. Position sizing, exposure caps, instruments, execution timing, data sources and cost assumptions still need to be settled before the model starts. There is no active model return or independently verified performance record.
02
Separate research from positions.
A report can introduce a thesis, revisit it or discuss the wider market. The model will state which ideas qualify as positions and apply the same rules to each. Macro commentary alone cannot be assigned a simulated return, and several reports on one asset do not create several positions.
03
Define the actual exposure.
Each position needs a named instrument, direction and weight: the equity listing, commodity contract or ETF, how an FX view is implemented, and each leg of a relative-value trade. Futures also need a roll rule. The model must show both notional exposure and capital invested.
04
Set trading rules in advance.
Entry, position changes and exits must use a trading time set in advance, after publication. Each needs a timestamp, timezone and price source. A weight changes only when a portfolio change is recorded; a stronger or weaker view in a report does not change it on its own.
05
Account for the full return.
The planned model will account for dividends, corporate actions, currency effects, carry, futures rolls, cash and estimated trading costs. Short borrowing, financing and dividends owed on shorts need stated assumptions. Any return already included in the price data must not be counted twice.
06
Track the whole portfolio.
The index will start at 100 and compound the portfolio’s returns, including cash and every eligible holding. Averaging the returns of separate reports would ignore overlapping exposures, holding periods and weights. The record will continue across calendar years without restarting in January.
07
Keep the losses in the record.
Closed positions, losses, failed theses and delisted securities stay in the history and cumulative return. Reversing a view closes the old position and opens a new one. Missing prices or costs must be shown, with a stated rule for how they affect the portfolio valuation.
08
Use the same dates and show the risk.
The proposed benchmark is S&P 500 Total Return in USD. Portfolio comparisons will use the same dates and currency rules. Return differences, maximum drawdown, gross and net exposure, concentration and each thesis’s contribution will show the risks of a mixed-asset long/short portfolio.
Original papers added / 8 October 2026
How the original papers are tracked.
The original PDFs are unchanged. The research date printed in each paper is shown separately from the date it was added here. Earlier website publication times and executed trades have not been independently verified. Where a paper gives only a month, I show only that month. Figures link to the original paper and relevant pages.
The 25 September portfolio uses the first completed US daily close after the printed research date: 28 September 2026. The comparison is 7 October 2026, the last completed US session at the time of this review. This rule was chosen when the paper was added here, after the outcome was known. It is a retrospective price comparison, not a trading rule set in advance or proof of an entry.
Available observations use Nasdaq’s published USD Close/Last values for the identified US equity or ETF: (ending close ÷ starting close − 1) × 100. These are raw price changes. Dividends, reinvestment, trading costs, slippage, financing, taxes and investor currency effects are excluded. Corporate-action adjustment is unconfirmed; an action in the interval can distort a raw price comparison. Source values and the fixed dates are retained, and future refreshes require a new dated review.
The September PDF contains 13 exposures with printed weights totaling 100.01% because of rounding. On 9 October 2026, I excluded PSH and the Avanza USA / Spiltan fund sleeve from the website model. The remaining 11 US equities keep their original weights, totaling 93.92%, with the remaining 6.08% in uninvested cash. This selection differs from the full original portfolio. Historical price observations keep the original weights and fixed dates. Mentioning a company again in later research does not add a position or change its weight.
SPY and QQQ provide same-date ETF price context, rather than index total returns. No portfolio outperformance is inferred from individual winning or losing names. The July conditional setup has no automatic entry; the August framework has no security-level trade instructions; October lacks an exact date, weights and position events. Their eligibility decisions are displayed with the papers.
See the prices and sources ↗Research price index / rules recorded 9 October 2026
The September stock basket.
The retrospective price basket uses the eleven selected September stocks at their original weights, totaling 93.92%, with 6.08% USD cash. It starts from the 28 September 2026 close. I made the selection on 9 October, after that date, so the reconstruction contains hindsight and selection bias. It does not establish the start of a prospective model, an actual entry or my account performance.
Each stock’s opening weight buys a fixed fractional quantity at the reference close. Quantities stay unchanged, current weights drift with prices, and cash receives no interest. The index is 100 × [cash weight + Σ(opening stock weight × current close ÷ reference close)], with weights expressed as fractions. An asset’s contribution is its opening weight percentage multiplied by its price change as a fraction, expressed in percentage points. The contributions sum to the basket’s cumulative price change.
SPY supplies an S&P 500 ETF price reference on the same dates; QQQ provides optional Nasdaq-100 ETF price context. Each begins at 100 at the same reference close. These are price series. Dividends, reinvestment, costs, slippage, financing, taxes and FX effects are excluded, and corporate-action adjustment is unverified. The prospective net model and its proposed S&P 500 Total Return benchmark retain separate rules and history.
New prices can be published only with evidence of a market-data licence or permission covering this website. Each update needs all eleven stocks and both references from the same completed US session, with the exact instruments, source links and retrieval times. Incomplete dates cannot be filled in or quietly removed. Published observations stay in the record; any correction needs a source correction and a dated version review explaining the change. The index remains pending until the data is complete and checked.
View the basket and data status ↗Methodology changes
Changes to the rules.
The final rules will be dated before the first tracked position. Routine changes will apply from that point forward. Data corrections will explain their effect on the results. Reconstructed history will be labelled separately from the record built as decisions are made.
View model portfolio status ↗Weekly review / Sunday, 18:00 Stockholm
The weekly review.
Each published report appears in the timeline beneath the reports in its category. The weekly review compares new papers with earlier research, including related ideas from other categories. Links between papers cite pages in both original PDFs; a material change in the thesis gets its own review date and page references. Original PDFs and research dates stay unchanged. A link or change in view does not open, resize or close a model position.
The public index will track the research model separately from my personal portfolio. A cumulative comparison needs complete, checked model valuations and matching S&P 500 Total Return observations. Missing data will be shown, and earlier observations and losses will stay in the record.