Built from the conviction ranking (what we've actually analyzed), sized by risk-adjusted expected return × screen agreement, capped at 8% single / 25% sector. The universe screen is a weight haircut, not a hard gate — 6 holdings carry a screen disagreement.

E[r] is the authored return, frozen at each name's analysis-date price. E[r] live re-anchors the same bull/base/bear targets to the last close (2026-07-28) — hover a cell for distance-to-base-target. A split? badge means the price broke vs the analysis (likely a split, e.g. ROCK-B) and the name needs a re-stream; its live figure is suppressed and the sizing falls back to the authored weight. The sizing module below can size on either basis.

Two-stage: the conviction (value) layer selects the universe; a momentum/trend gate then sizes the book. The price spine is 12-1m momentum (trailing-year return, skip-month) z-scored across names; the 40-week MA times entries (benching cheap names below a falling 40w MA — value traps still falling; overweighting those reclaiming a rising 40w MA with positive momentum). The 200-week MA & 52-week-high flag the over-extended (stretched on the 4-yr cycle) for trimming.

Fundamental-momentum leg (35%): position size blends price momentum (65%) with realized fundamental momentum — ROIC & operating margin vs each name’s own 5-year baseline, plus 5-year revenue growth (Novy-Marx fundamental momentum, less crash-prone). It modulates size only — deteriorating fundamentals shrink a high-price-momentum name; improving ROIC lifts a quiet one. The ΔROIC column shows the headline driver. Optional overlay; not backtested (forward-return validation matures ~Aug 2026).

Of 28 eligible value names — confirmed 8, basing 8, weakening 2, extended 1, broken 7, no-data 2. Active book 19; 9 value names benched awaiting trend confirmation. Versus pure value: 9 dropped, 0 newly included.

Concentrated by design. The active book is 19 of 30 target slots. The gap is not a missing-analyses gap — the analyst has reviewed the top of the universe and chosen HOLD/OBSERVE on names like NONG.OL, TRAC-B.ST, AKSO.OL, ABG.OL etc., consistently citing cyclical-peak or quality concerns (see the verdict_rule badges). Forcing the book to 30 would dilute conviction; the smaller, higher-conviction book is the honest reflection of available BUY-rated names. See VERDICT_RULES.md for the rule taxonomy.

Fully automated: eligibility comes straight from the nightly universe screen (rating ≥ Attractive, data-integrity ok, one share class) — no analyst dependency. The same 40-week-MA trend gate and 12-1m momentum sizing as the Value+Momentum book. Hysteresis: a held name is kept while it stays in the top 50 by natural rank (an extended holding is kept while it would rank inside the band without the extended sizing haircut, so its shown rank can exceed 50) — new names enter only when a held name exits the band, breaks trend, or loses screen eligibility. This cuts turnover; only the top 30 are sized.

Screen-eligible 212 — confirmed 50, basing 57, weakening 30, extended 18, broken 53 (benched). Active 155; book 30; buffer (31–50) 22. Turnover this run: +1 / −1. Movement vs 2026-07-23 (same universe, same screening method). Not backtested; forward validation matures ~Aug 2026. The known trade-off vs the conviction book: this holds screen-cheap cyclicals while their trend is intact (the analyst HOLDs them) — the 40w gate is the exit, not analyst judgment.

Exits this run: DNO.OL (rank fell to 53 (outside top 50))

Weights shown use the default sizing basis (inverse volatility). To size an actual allocation, switch basis, and download the .xlsx, use the Position sizing module below — it follows this portfolio when the NC-Momentum toggle is active, and explains the inverse-volatility calculation.