Value Complex

Value Complex — Avantis small + large value, by weight

This skill reads Avantis's two flagship value ETFs straight from TickerTrace's per-fund endpoint (/fund/<TICKER>) and lays out what changed under the hood, grouped the way the value factor actually moves:

• AVUV — small-cap value ($12.5B, ~794 holdings) — the flagship.
• AVLV — large-cap value ($3.2B, ~272 holdings).

This is the per-fund holdings view that the daily stock-recap skill does not pull. stock-recap only hits TickerTrace's cross-fund aggregates (/briefing, /institutional, /divergences). This skill is the deep, by-weight, per-fund value read.

What it groups, and why

For each fund (small tier → large tier), three lifecycle buckets:

1. 🆕 New / Exited — positions TickerTrace flags type=NEW or REMOVED. Passive Avantis funds rarely open or close a name day-to-day, so when they do it's meaningful. This is the literal "new adds."
2. 📈 Conviction tilt — active over/under-weighting, inflow-stripped. Every name in an inflow day shows a positive sharesDelta (new money buys the whole book pro-rata), so raw adds are noise. We compute each name's share growth sharesDelta/previousShares, take the fund-wide median as the inflow tide, and rank by the excess over that tide. Positive excess = the fund is genuinely leaning into the name beyond mechanical inflow; negative = leaning out.
3. 🔥 Up-streaks — names on multi-day winning runs (TickerTrace streaks). This is the cleanest momentum signal — independent of fund mechanics entirely.

Then the payoff, 🔗 cross-tier synthesis:

• Sector convergence — sectors that are streaking and/or tilting up in both the small and large fund. (Exact-ticker overlap is rare because the funds hold different cap tiers, so convergence is read at the sector/theme level — e.g. "transports leading in both.")
• Flow tell — all-buys with zero trims = inflows into value (creations, risk-on); a genuine mix of adds and trims = active rebalancing.

How to run it

node "$HOME/.claude/skills/value-complex/scripts/gather.mjs"

• Writes runs/<date_time>/report.md + raw/fund_AVUV.json, raw/fund_AVLV.json.
• Prints the report to stdout and a final JSON line with {reportPath, rawDir}.
• No key needed — /fund/<TICKER> is public on TickerTrace. Presents a browser User-Agent (the edge WAF 403s a bare Node UA). ~5–10s.
• Override the base with TICKERTRACE_API_URL; funds with VC_FUNDS=AVUV,AVLV,AVMV.

Present it to Mike

Read the generated report.md and give a tight read, in this order:

1. The flow tell first — are both funds taking inflows (value getting bought) or net trimming? That frames everything.
2. New / Exited names per fund — rare, so always call them out by name.
3. Conviction tilt — the inflow-stripped over-weights. These are the names the fund is actually leaning into, not just buying with the tide. Name the top 3–4 per tier and what sector they're in.
4. Streaks — the momentum leaders, longest runs first.
5. The cross-tier theme — the one or two sectors strong in both small and large value. This is the headline ("the whole value complex is leaning into X"). Tie it to the tape: what's it rotating out of (usually mega-cap growth).

Notes / gotchas

• weightDelta is not a clean signal on its own — it blends price drift with share-count change. The report leads with the inflow-stripped excess instead; it shows weightDelta as secondary context only. Don't present weightDelta as "conviction" without the excess check.
• Data is daily and can lag — the report prints TickerTrace's asOfDate; surface it so Mike knows how fresh it is.
• Cross-tier convergence is sector-level, not ticker-level — small and large funds rarely share a ticker; agreement shows up as a shared theme.
• Each run is a point-in-time snapshot under runs/; nothing is overwritten and old folders are safe to delete (gitignored under .claude/).