APEX factor · Short Interest
Short interest reads the size of the bearish bet against a stock relative to its float. High and rising short interest is bearish — many sophisticated traders have done the work and concluded the stock is overpriced. But extreme short interest with rising fundamentals is the precondition for a squeeze, which flips the signal.
Short selling is expensive, risky, and time-pressured. If a hedge fund is paying borrow fees of 8%/year and holding overnight margin against a stock, they have done the homework. A high short ratio is the market's collective bearish thesis aggregated. Most of the time it works as a fade — the bears were right, the stock declines slowly, the shorts cover with a profit. Sometimes — when fundamentals improve faster than the bears expected, AND the stock has high days-to-cover, AND retail momentum picks up — the shorts cannot exit fast enough and the price gaps up. Both regimes exist; the factor reads which one a ticker is in.
si_ratio = short_interest / float days_to_cover = short_interest / avg_daily_volume_30d inst_ratio = institutional_ownership / shares_out crowd_score = si_ratio · inst_ratio · (1 + log(1 + days_to_cover)) squeeze_flag = (si_ratio > τ_high) AND (momentum > 0) AND (days_to_cover > τ_dtc) si = -1 · z_score(crowd_score) + squeeze_flag · κ_sq short_interest_score = z_score(si)
Two mechanisms in one factor. The crowd score is bearish-when-high (sign-flipped to bullish-when-low). The squeeze flag adds a bullish bonus only when the three squeeze preconditions co-fire — extreme short ratio, positive momentum, high days-to-cover. Thresholds τ_high, τ_dtc and the squeeze coefficient κ_sq are not disclosed; the structure follows Asquith-Pathak-Ritter 2005 plus Lamont-Stein 2004.
Short interest comes from FINRA's bi-weekly settlement reports plus daily updates from Finnhub. Institutional ownership comes from SEC 13F filings (quarterly cadence). Days-to-cover is computed from the 30-day average daily volume on Yahoo Finance EOD data. The factor refreshes when any input changes — typically twice a month for the SI baseline, daily for momentum and volume inputs. The squeeze setup re-evaluates every day during the 06:00 UTC universe sweep.
Short Interest interacts with Momentum to form the SHORT SQUEEZE SETUP confluence pattern — extreme short interest plus rising momentum plus high days-to-cover is the Asquith-Lamont signature. With Quality and Value, it composes the CROWDED-SHORT VINDICATION pattern — when bears are right (high SI plus deteriorating quality plus a value trap signature), the bearish thesis is doubly anchored. The factor's regime amplifier is conservative — short-squeeze setups depend on retail flow, which is regime-driven; in risk-off the squeeze flag down-weights automatically.
The factor's two regimes are inherently hard to separate ex-ante, and the signal noise around the squeeze threshold is high. Real-world example: between 2020 and 2022 we would have seen GME, AMC, BBBY at extreme short levels with positive retail momentum — the signal would have flagged squeeze setups correctly on GME and AMC, but the same heuristic on BBBY produced a false positive that crashed back. We accept this — the false-positive rate near the squeeze threshold is the cost of capturing the rare but extreme upside. A second limitation: the FINRA bi-weekly cadence means the SI baseline is up to 14 days stale at the moment of evaluation, which materially affects fast-moving names.
Every ticker page shows the per-factor decomposition. The Short Interest score is one of twelve composing the 0–100 APEX composite.