Small stocks that gap up hard before the open, on retail hype or a pump, tend to deflate during the session. Traders call it "gap and crap". Short at the open, cover at the close: no overnight borrow and no exposure to the market's overnight drift. The counterparty is the late buyer chasing the gap at the bell.
intraday / small-cap gap fade
Intraday gap-and-crap short
The idea
How it was tested
Pre-registered on 2026-06-04 as a systematic version of that lore. Signal: an open at least 20% above the prior regular-session close, with at least $1M of pre-market dollar volume. Universe: every US stock with 1-minute data and a prior close of $5 or more. Short at the 09:30 ET open, cover at the 16:00 close or at a +15% intraday stop. Costs: 25 bp slippage per side and $2 commission per round trip. One-minute SPY history was too short, so the benchmark was the same day's median open-to-close return across the universe.
Where it died
On the first run it passed, and clearly: 7,630 trades, alpha of 2.285% per trade with t 11.96, and +1.854% net with clean fills. The verdict that day was "passed".
The next day a yearly breakdown found the alpha shrinking as the number of qualifying gappers grew (r −0.937). The day after, the stops were stressed. The backtest assumed the +15% stop filled exactly; the stress filled it halfway between the stop and the day's high, doubled slippage and dropped the least liquid quarter of the gappers. Net per trade became −4.09% with t −8.08, and the worst 5% of trades lost 43.5% or more. The verdict changed from passed to not capturable on 2026-06-06, before any real-time intraday infrastructure was built.
The record
Thesis
Small stocks that gap up hard pre-market on retail hype or a pump deflate during the session. Shorting the open and covering at the close captures the fade, with no overnight borrow and no exposure to the market's overnight drift.
Specification
- signal
- open ≥ 20% above the prior regular-session close, pre-market dollar volume ≥ $1M
- universe
- all US stocks with 1-minute data, prior close ≥ $5
- entry
- short at the 09:30 ET open
- exit
- cover at the 16:00 close, or a +15% intraday stop
- costs
- 25 bp slippage per side, $2 round-trip commission, no borrow (intraday)
- benchmark
- Same-day cross-sectional median open-to-close return of the universe — 1-minute SPY history was too short to use SPY.
Gates
Pre-registrationpass
Was the hypothesis, its parameters and its pass/fail rule written down before any data was looked at?
Market-adjusted alpha (legacy)pass
Does the strategy beat SPY over the same holding windows, assuming beta = 1?
t ≥ 1.65 vs the same-day cross-sectional median
t stat+11.96alpha / trade %+2.29trades7,630
Not a Jensen alpha against SPY — the benchmark is the universe's own intraday median.
Gate F — Frictionfail
Is the alpha per trade large enough to pay for its own execution?
net > 0 and alpha t ≥ 1.65 under a stressed stop-fill model
net per trade pct clean fills+1.85net per trade pct stressed-4.09t stressed-8.08p5 trade pct stressed-43.50
Clean fills at +15% pass. The stressed scenario fills each stop halfway between the stop and the day's high (a proxy for halts and gap-throughs), doubles slippage and drops the least liquid quarter of gappers.
Why it died
The edge is real in the data — alpha t 11.96, +1.854% net per trade with clean fills — and it does not live in the illiquid names (alpha holds when they are dropped). What kills it is the stop. The backtest fills the +15% stop exactly; real gappers halt and reopen through it. If stops fill even a quarter of the way from the stop to the day's high, net per trade goes from +1.85% to −0.80%; in the combined stress it is −4.09% with t −8.08.
What it taught
A stop level in a backtest is a promise the market does not have to keep. The cheapest possible proxy — the day's high — was enough to show the edge could not survive realistic stop fills, before any intraday real-time infrastructure was built. The yearly breakdown also showed the alpha falls as the number of qualifying gappers rises (R² 0.88 on full years): crowding, not decay.
Verdict history
- 2026-06-04Single run: alpha t 11.96 vs cross-sectional median, net +1.854%/trade with clean fills Passed all gates
- 2026-06-05Yearly breakdown and regime fit: alpha vs gapper count r −0.937, R² 0.88 — edge shrinks with retail participation
- 2026-06-06Tail stress: edge flips negative with leaky stop fills; liquidity filter is not the problem Not capturable