What every scan hides
Research tool · not investment advice. Full disclaimer →
What every scan hides
every published scan counted three ways, including the ways that loseSource: a committed measurement artefact · not yet generated in this deployment.
A dated study, not a live number. Measured once against the full archive with the tape through 2026-08-28. Forward window 20 sessions; a flat 0.32% round trip; signals floored at close > 20 and turnover > 10,000,000. It does not update on its own.
Every row runs through the identical machinery — same archive, same forward window, same liquidity floor, same corporate-action check, same two universes. Only the scan differs, which is what makes this a comparison rather than a pile of separate studies. Reported is today’s survivors applied backwards, the way a scanner’s backtest is normally built. After cost is the same signals on the universe as it actually stood, minus the fee.
| Scan | Signals | Reported | After cost | Overstatement | Survivorship ± noise floor | From vanished names |
|---|---|---|---|---|---|---|
| 52-week low breakdownToday's close is the lowest close of the last 252 sessions, and yesterday's was not. | 33,321 | 1.051% | 0.124% | +0.926 pp | +0.606 ± 0.088 | 18.1% |
| Gap up 3%The open is at least 3% above the previous close. | 111,569 | 1.272% | 0.469% | +0.803 pp | +0.483 ± 0.070 | 26.9% |
| RSI(14) oversold — Cutler'sCutler's RSI over 14 sessions crosses up through 30. Cutler's uses a SIMPLE mean of gains and losses, not Wilder's exponential smoothing; the two disagree and this row is the simple one. | 108,872 | 0.855% | 0.378% | +0.477 pp | +0.157 ± 0.043 | 18.2% |
| Volume shockerVolume is at least 3x its own 20-day average, on a day the close is up. | 195,306 | 0.792% | 0.342% | +0.450 pp | +0.130 ± 0.047 | 22.8% |
| Golden crossThe 50-day simple moving average crosses above the 200-day. | 10,193 | 1.161% | 0.713% | +0.448 pp | +0.128 ± 0.171 · inside the noise | 17.6% |
| Within 2% of a 52-week highClose is within 2% of the highest close of the last 252 sessions. | 361,811 | 1.497% | 1.131% | +0.366 pp | +0.046 ± 0.029 | 17.6% |
| Bollinger upper-band breakoutClose above the 20-day average plus two standard deviations of the last 20 closes, having been below it yesterday. | 159,020 | 1.445% | 1.088% | +0.357 pp | +0.037 ± 0.045 · inside the noise | 20.4% |
| 52-week high breakoutToday's close is the highest close of the last 252 sessions, and yesterday's was not. | 87,068 | 1.777% | 1.423% | +0.353 pp | +0.033 ± 0.066 · inside the noise | 18.4% |
| Volume dry-upThe 20-day average volume has fallen to half or less of the 100-day average. | 297,227 | 0.980% | 0.651% | +0.329 pp | +0.009 ± 0.027 · inside the noise | 17.5% |
| Death crossThe 50-day simple moving average crosses below the 200-day. | 8,972 | 1.197% | 0.902% | +0.295 pp | -0.025 ± 0.138 · inside the noise | 17.1% |
| Bullish moving-average stackClose above the 20-day average, which is above the 50-day, which is above the 200-day — entering that state today. | 22,884 | 1.754% | 1.466% | +0.288 pp | -0.032 ± 0.139 · inside the noise | 17.9% |
| Delivery shockerDelivered quantity as a share of volume is at least twice its own 20-day average, on a day the close is up. | 4,250 | 0.968% | 0.682% | +0.286 pp | -0.034 ± 0.298 · inside the noise | 33.0% |
Rows are ordered by the size of the counting gap, not by return. That is deliberate: an ordering is an editorial claim, and the only claim this page is entitled to make is about how far apart the two ways of counting are. Nothing here says a scan is worth running.
What is not on this table, and why
Four scans were considered and deliberately not built. They are listed because a table that quietly stops at what was convenient reads as coverage.
- supertrend — Needs a recursive band that carries its own previous value forward. Not expressible as a SQL window function, and a re-implementation that differs from the reader's tool would be worse than not shipping it.
- macd-cross — Needs exponential moving averages, which are recursive for the same reason. Deliberately not approximated with an SMA and labelled MACD.
- wilder-rsi — The Wilder-smoothed form of the RSI row above, for the same recursion reason. Cutler's is shipped and labelled instead of shipping Wilder's approximately.
- candlestick-patterns — Doji, engulfing, hammer and the rest are definable, but every vendor uses different body/wick tolerances. There is no standard definition to reimplement, so any number we published would be a measurement of OUR tolerances.