# The MTG Portfolio, and what to fund next

**TO:** The CEO · **FROM:** Product · **RE:** turning one-offs into one compounding line
**Original:** 2026-07-08 · **Status revision:** 2026-07-18

Live page: https://drwu-htmls.vercel.app/mtg-ceo-pitch/

---

## Status revision, 2026-07-18

Ten days on, **two of the eight pitches are delivered and live**. Every route below returned HTTP 200 on 2026-07-18.

| Pitch | Status | What shipped |
|---|---|---|
| **03** Generalize the craft ranker | **SHIPPED** | [MTGA Craft Planner](https://drwu-htmls.vercel.app/mtga-craft-planner/): any color combo, 25 archetypes ranked by mythic win rate, collection-aware craft path over 60 staples |
| **08** Creator winrate database | **SHIPPED, with a caveat** | [Creator Standard Decks](https://drwu-htmls.vercel.app/mtg-creator-decks): 83 decks, 13 creators, plus new price-to-build and wildcards-to-build sorts |
| 01, 02, 04, 05, 06, 07 | open | Unchanged. Pitch 01 remains the recommendation. |

## The finding that changes Pitch 08

Pitch 08 asked for **"a slow polite Moxfield harvester."** Politeness is not the obstacle.

| Host | Decks | Result |
|---|---|---|
| MTGGoldfish | 7 | **200** via `/deck/download/<id>` |
| Moxfield | 52 | **403** behind Cloudflare, on `api2` and on the public page after its 301 |
| Aetherhub | 23 | **403** behind Cloudflare |
| Untapped | 8 | 200, client-rendered behind an encoded deckstring |
| No link | 22 | nothing in the description |

Rate limiting does not move a 403. Only **7 of 83 decks** expose a machine-readable list, so **76 read "list walled"** rather than carry a fabricated price. Unwalling them needs a signed-in browser pass, which is a different and more expensive ask than this memo originally made.

A second gap: the **transcript-claim confidence model was never built**. YouTube's transcript API returned RequestBlocked, so colors and archetype come from video descriptions, not from what the creator says. The opinion-weighted signal that made Pitch 08 interesting is still open.

## Three constraints, priced

| Constraint | Evidence | Cost |
|---|---|---|
| Decklist hosts are walled | 403 Moxfield · 403 Aetherhub · 200 MTGGoldfish | 7 of 83 priceable. Bulk decklist access needs a browser budget, not a delay. |
| Transcripts blocked | RequestBlocked | No creator-claim confidence score. |
| Collection snapshot ages | stamped 2026-04-22 | Wildcard math drifts on every craft. Mitigated with an editable budget field. |

The honest read: the data layer we own outright is **Scryfall**, which is why price-to-build works wherever a list is visible, while win rate and creator opinion stay out of reach.

## If you fund one thing

**Still Pitch 01**, the post-release accuracy backtest. Low effort, credibility unlock, and with 03 and 08 delivered it is the only remaining item that raises the value of everything else at once.

Revised sequence:

1. **Backtest the Marvel grades** against real post-release prices, publish the scorecard
2. **Parametrize the pipeline**, pilot on the next spoiled set
3. ~~Unpin the craft ranker~~ **shipped 2026-07-08**
4. **KC Events "today" button** and ICS subscribe
5. **Package Badgering** print-and-play, then wire the MTG hub

A pitch to fix decklist access would unlock Pitch 08's real value and is not on this slate yet.

---

Sources: `~/Projects/drwu-htmls/public/`, `~/Projects/_outputs/mtg-marvel-feedback/`, `~/Projects/_outputs/creator-winrate-db/`.