Methodology
How the book works — study, value, thesis, trade, track
This is the book — a daily, full-transparency practice run in public: I study companies and their financial statements, value them, form theses, trade them with real money, and track the real results. This page explains how that loop works, what the numbers mean, and how to read every signal.
I publish wrong calls and losing days as loudly as the right ones. The point of a public track record is the record, not the highlight reel.
The daily practice
Study → value → thesis → trade → track
- Study. Most days start with a company and its financial statements — the filings, the unit economics, the moat. The learning is logged in the daily journal whether or not it leads to a trade.
- Value. Where it earns the work, a Damodaran-style DCF or a relative-value read turns the study into a number — a price I'd pay, and why.
- Thesis. A view with a direction, conviction, timeframe, and target. Falsifiable. The ones that clear the bar graduate to a public call.
- Trade. Positions are taken in a real IBKR account. The book is real money — not a paper simulation or a model portfolio.
- Track. Every day the book is frozen and published. The cumulative return becomes the track record; resolved calls become the scorecard.
The numbers
Real money, percentages only
The headline track record is the cumulative return on the real book — a time-weighted % curve built from each day's snapshot, sourced from the IBKR account.
Everything is published percentages only: position P&L %, % of book, market price. Never absolute dollars, share counts, or account size. The return is fully verifiable; the sizing stays private. This floor is enforced in the data layer — absolute figures are stripped before anything is stored, not just hidden in the UI.
Direction
Which way the thesis cuts
I expect the price or thesis to be validated higher within the timeframe.
I expect the price or thesis to be invalidated lower within the timeframe.
Range-bound or no clear directional edge — surfaced because the setup matters.
No call yet — I am tracking a setup until conviction crystallises.
Conviction
How sure I am
Strong evidence, multiple converging sources, and a clear catalyst path. I expect to be right; if I am wrong, I want to know quickly.
A reasoned position with meaningful uncertainty. Worth tracking; not worth betting the house.
A directional lean I publish for transparency, not endorsement. Often calls I am watching evolve.
Conviction is also tracked over time — every revision is logged on the call detail page so you can see when and why I changed my mind. A drift from HIGH to LOW is a signal in itself.
Timeframe
The window I'm committing to
Tactical positioning — earnings windows, near-term catalysts, sentiment shifts.
Cyclical positioning — sector rotations, capex cycles, mid-term thesis testing.
Strategic positioning — full earnings cycle, secular trends with mid-term inflection.
Structural positioning — long-cycle theses, technology adoption, multi-year compounders.
A call without a timeframe is a wish. The clock starts at publication and ends at the stated horizon — at which point the call resolves to Validated, Invalidated, or Expired.
Status
Where the call is in its lifecycle
Position is live. I track and update it with new evidence as it lands.
The thesis played out within the timeframe. Outcome documented and added to the track record.
The thesis was wrong. I say so explicitly — wrong calls teach more than right ones.
The timeframe lapsed without resolution. Marked closed; not counted as right or wrong.
Brier scoring
How I score probabilistic forecasts
For probabilistic claims (“65% chance HBM4 ships at scale before Q4 2027”), I use the Brier score — the squared error between my predicted probability and the actual outcome.
Lower is better. A perfect predictor scores 0; a coin-flip-as-everything scores 0.25. Every resolved forecast contributes to the running Brier on /kb/calibration, along with a calibration curve showing whether my 70% forecasts actually resolve true 70% of the time.
The point: confidence calibration is a discipline. Saying “90% sure” means you should be right ~9 times in 10. If you're not, your scale is broken — and Brier shows it.
What I don't do
Boundaries and disclosure
- No financial advice. This is research and opinion, published for transparency. It is not a personalised recommendation.
- No hidden positions. The book itself is the disclosure — if I hold it, it's in the published positions. The whole point is that you see the real book.
- No silent edits. Every revision to a published call — conviction change, status change, target change — is logged on the call page with a timestamp and reason.
- No retroactive scoring. A call counts in the track record only at its stated horizon, by its stated thesis. I don't move the goalposts.
Who writes this
One operator, in public
The book is mine — one operator studying, valuing, trading, and writing it up daily, in public. No desk, no committee, no ghostwriters. When I'm wrong, that's on the record too.