Price MFS and MME off their ETF proxies

[mspinola]

MFS and MME are ICE MSCI futures that Norgate carries no continuous series for, so there is no futures price to read. cotdata still holds their COT, which is why they are in the universe at all. Until now they had no prices whatsoever and every price-derived column for them was empty.

marketdata now registers EFA and EEM and the store holds their bars (mspinola/marketdata#17). This is the half that uses them: a PRICE_PROXIES map in market_data, and a price_symbol() that the three read paths go through.

The map is deliberately not the options one

options_data.ETF_PROXIES exists because a futures options chain is illiquid, so it names a proxy for markets that have perfectly good prices of their own: it maps ES to SPY, GC to GLD, CL to USO.

Reusing that map for prices would replace S&P futures prices with an ETF across the entire book, and the result would look plausible. So most of the new test file guards that blast radius rather than the feature:

• eight markets asserted never to proxy (ES, GC, CL, ZB, 6E, BTC, RTY, NQ)
• a check that the two maps share no keys
• an exact assertion on the map's contents, so growing it fails here and has to be argued for rather than passing quietly

The substitution is real, and the map says so

An ETF tracks its index net of fees, in USD, on US session hours. The future carries basis, financing and a currency treatment it does not. Levels are not comparable and neither are returns over a dividend date, so anything comparing these two markets against genuinely futures-priced ones needs to know. Both are Role: heldout, so this feeds display and indexing rather than anything selected or traded.

Verification

Cold cache, real 47-market universe:

MSCI EAFE              rows=792  priced_weeks=792   last_close=108.09
MSCI Emerging Mkts     rows=792  priced_weeks=792   last_close=65.43
Gold                   rows=1929 priced_weeks=1180  last_close=4441.10   <- control, unchanged

Both go from zero priced weeks to fully priced. 294 passed, ruff clean.

One existing test changed rather than being added to: test_the_rejection_scores_read_does_not_pin_a_tier asserted the read asks for MFS, which was correct before a proxy existed and is now EFA. Its actual point, that the tier must be one an equity accepts, is unchanged.

Order

Merge mspinola/marketdata#17 first. Without it there is no EFA/EEM to resolve to, and test_the_proxies_are_actually_readable will fail.

🤖 Generated with Claude Code