A3 · Positioning Extreme

A squeeze setup: a heavily-shorted name whose price refuses to break down, hinting the crowded short is wrong.

The short code A3 is display metadata only. The canonical identifier this archetype is keyed by everywhere — config, the screener_candidates.archetype column, the scorer emit — is the descriptive string positioning_extreme.

Dated note (2026-09-01) — primed gate retuned. Before this date primed was structurally unreachable across the universe: the options-skew leg gated it, and options data covers only the watchlist. Skew is now additive confirmation only, and primed rests on the two universe-wide legs (short-extreme + price-resilient), so the primed cohort begins accruing — and its calibration begins fresh — from 2026-09-01.

Thesis

High short interest by itself is not an edge — short sellers are often informed, and most hated stocks deserve it. The non-obvious setup is crowded-short-INTO-resilient-price: shorts are piled in, the bearish thesis is consensus, and yet the price won’t confirm. When a name carrying heavy short interest holds its ground instead of making new lows, the positioning has gotten ahead of the fundamentals, and the fuel for a squeeze is loaded. The edge is the divergence between extreme bearish positioning and a price tape that refuses to validate it.

Direction

LONG. The setup anticipates a short-covering reversion higher.

What it looks for

The evaluator evaluate_positioning_extreme reads its thresholds from screener_config.json § archetypes.a3:

• Short extreme (this is the partial rule) — crowding on EITHER input: short_extreme = (short_pct_float >= short_pct_float_min=15.0) OR (days_to_cover >= days_to_cover_min=5.0). days_to_cover is the universe-wide leg — it comes from the FINRA bulk short-interest feed (wired June 2026) and is present across the whole sweep, which is what lets A3 fire universe-wide. short_pct_float (short interest as a percentage of float, already in percentage points — 15.0 = 15%, normalized by each producer rather than re-normalized here) is a deep-path-only field: NULL on the sweep, supplied only when a name is scored individually. The evaluator abstains (none) only when BOTH short inputs are absent — the price leg alone can never fire A3.
• Price resilient (gates primed) — ret_20d >= primed_ret_20d_floor=+0.05 (flat-to-rising over the 20-session window — a genuine refusal to break, not merely “not in freefall”) AND above the 52-week low (dist_52w_low > 0). This is the divergence the setup rests on: crowded short into a tape that won’t confirm. The softer price_holding_ret_20d_floor=-0.05 is retained only as an informational trigger, not for primed.
• Options skew (additive confirmation only) — surfaced as a trigger where watchlist options data exists (put_call_ratio > skew_pcr_floor=1.5 OR skew_25d > skew_25d_floor=0.05, the 25-delta risk reversal on a 0–1 scale so 0.05 = 5 vol points of downside-fear premium), but it never gates primed. It reads the latest co_options_daily row, which is watchlist-only, so for the vast majority of universe names it ABSTAINS cleanly — no score contribution, no error, no penalty. Gating primed on this sparse leg was what left primed structurally unreachable universe-wide before the 2026-09-01 retune.

How it scores

The state machine turns on the two universe-wide legs; skew, where it exists, is confirming colour, never a gate:

• none — quality fail, or not short-extreme.
• partial — short-extreme alone (crowded short, but the tape has not yet refused to break).
• primed — short-extreme AND price-resilient (the divergence is confirmed). The firing candidate ranked toward S2; primed_ret_20d_floor is the operator’s lever if the primed pool floods the board.

Pitfalls / when it fails

• Shorts are often right. This is the headline risk: piling into a crowded short opposite informed sellers only works when the price genuinely refuses to break. The price-resilient leg is what gates primed, and it is the entire safeguard against a screen that just lists hated stocks — respect it.
• No options data for most names. Because co_options_daily is watchlist-only, the additive skew confirmation is unavailable for nearly the whole universe. That is by design now — primed stands on the two universe-wide legs (short-extreme + price-resilient); on a non-watchlist name you are simply missing the options-premium colour, not a gate.
• Short-interest staleness. Float short percentage is reported on a lag; the figure may not reflect covering that’s already underway.

Related

• The Screener System (vision)
• Screener Pipeline & Lifecycle (operations)
• Mechanical reference cards: Score · State · Triggers
• Sibling archetypes: A6 · Vol Mispricing · A2 · Oversold Sympathy