Original Research Lead hypothesis (not seeded). New instrument class: no hypothesis in the corpus has traded preferred shares, and CAEQ has 3 of 5,022 experiments. Canadian preferreds (BCE, ENB, TRP, PPL, MFC, SLF, GWO, BMO, TD, POW series) are a retail-dominated market with median daily turnover of about C$50k per series. Their prices share one main driver: the 5-year Government of Canada yield that sets rate resets, plus issuer credit. The median trailing 250-day correlation of a series' daily return with the equal-weight preferred basket is 0.58. When one series drops sharply against that basket while its sibling series from the same issuer do not, the drop is most likely a liquidity shock (a retail seller or fund outflow hitting a thin book), not news. Issuer news or an issuer-wide ex-dividend date moves every series of that issuer. Liquidity providers who absorb the sale are paid by the subsequent recovery (Grossman-Miller liquidity provision; Nagel 2012, 'Evaporating Liquidity'). This also answers lesson L269: it is a narrow peer panel whose members share one common driver, with the homogeneity measure and the exact-name forward returns pre-registered below. Pre-study on our CAEQ daily bars, 2018-01 to 2026-10, frozen baseline rule (z<=-2.5, sibling-series median z>-1.0, 20-day median traded value >=C$25k, at most one series per issuer and 3 entries per date, next-open to open 5 sessions later): 585 events on 505 distinct entry dates. Mean gross return +0.36% per event, median +0.18%, 5th percentile -4.2%. Eight of nine calendar years were positive; 2022 was -0.38% during the rate shock. The per-date t-stat is 2.95. The stricter 3.0-sigma variant gives 288 events and +0.52% per event. Cost arithmetic: CAEQ charges zero commission plus a modelled 3 bp half-spread, about 0.06% round trip. Real preferred spreads are wider, so the plan sets a 10 bp per side slippage floor, about 0.26% round trip all-in. That leaves an expected net of about +0.10-0.15% per trade, a gross edge roughly 1.4x cost; this is the main risk and is pre-registered honestly. Known biases, stated up front: (1) dividends are not in PnL, and the platform holds no historical ex-dividend dates for CA preferreds (the dividends and adjclose feeds hold only the latest row). Holding through an ex-date therefore books the roughly 1.2-1.5% quarterly (or 0.4% monthly-floater) drop as a loss, with no credit for the dividend. Issuer-wide ex-date drops are mostly removed by the sibling filter, but single-series issuers (BMO-PE, GWO-PN, POW-PE) are not protected. The bias is against the strategy, and the pre-study already includes it. (2) The universe is Yahoo's current top-N list. Redeemed series that Yahoo still serves (TD-PFA, TD-PFJ, SLF-PK, TRP-PA) are included and close at their last bar through the platform's listing calendar. Series redeemed earlier and dropped from the list are missing; redemptions happen at par, usually after price gains, so this residual bias is small but unsigned. (3) Bid-ask bounce: entering at the next OPEN, not the signal close, removes the close-at-bid bounce. The open print itself can still sit at the bid, so the 10 bp per side floor is required. Execution: next_open_fill is supported on session venues, so a decision at the session close fills at the next TSX open. No stops are used (route (b)).