The largest single-session moves across US equities, and underneath them the full record of every candidate our own filters removed, with the number that removed it. Most sites publish the board. The second table is the one worth reading.
Session 2026-11-06, measured against 2026-11-05. 2,951 equities cleared the liquidity gate. 46 exclusions recorded across 12 sessions.
What made the board
Symbol
Change⊙
INFERENCE — derived by arithmetic
SourceClose against the previous close, Databento EQUS.SUMMARY daily bars, every US equity
As of2026-11-06 against 2026-11-05
HowRaw traded prices, not split-adjusted, which is why a move of 25% or more must come with heavy volume to be listed.
Close⊙
EVIDENCE — filed/published fact
SourceSession close, Databento EQUS.SUMMARY daily bars, every US equity
As ofclose 2026-11-06
HowRaw traded price, delayed, not real-time.
Turnover⊙
INFERENCE — derived by arithmetic
SourceClose times shares traded in the session
As ofsession 2026-11-06
Volume vs median⊙
INFERENCE — derived by arithmetic
SourceSession volume divided by the stock's own median volume
As ofsession 2026-11-06; median over the completed sessions before it, this one excluded
Session 2026-11-06: of the 12 names on the board, 10 had a filed record new to the session, 0 a headline only, 2 nothing on file after all three checks answered, and 0 could not be read. Checked 2026-11-07 07:45 UTC.
Session
Filed record
Headline only
Nothing on file
Not read
Names
2026-11-06
10
0
2
0
12
2026-11-05
10
0
2
0
12
2026-11-04
7
2
3
0
12
2026-11-03
6
0
6
0
12
2026-11-02
8
1
3
0
12
2026-10-30 (in progress)
0
0
0
6
12
Pooled over 5 finished sessions (60 names): 68.3% filed record, 5% headline only, 26.7% nothing on file, 0% not read.
Three checks run on each name, and a record counts only when it went public between the previous close and this close: SEC filings (the company’s own submissions file: 8-K, 6-K, Form 4, 10-Q and 10-K, placed by the time EDGAR accepted each), halts (our archive of the Nasdaq, NYSE and Cboe halt feeds; a volatility pause is listed but not counted, because it follows a fast move rather than explaining it) and headlines (TickerTick, by publish time). A record public before the previous close is not counted, since earlier sessions traded on it too. A name is “not read” when a check did not answer, and nothing is claimed about it. A filed record is on file, not a proven cause. These are only the names the board’s filters let through: how a name reaches the board and what the filters removed. Each name’s read is kept at a dated link from the board above, and the live check for any ticker is at why is it moving.
What we filtered out
These are candidates that cleared the price and turnover gates and then failed one of the two judgement guards. Names that never cleared a $1 price, $5M of daily turnover, or the equity-type filter are not logged individually, and neither are the thousands of stocks that simply did not move much. Each session records at most 40 exclusions, and the log keeps 12 sessions.
2026-11-06 · 13 exclusions
Symbol
Would have shown⊙
INFERENCE — derived by arithmetic
SourceClose against the previous close, Databento EQUS.SUMMARY daily bars, every US equity
As of2026-11-06 against the session before it
HowRaw traded prices, not split-adjusted, which is why a move of 25% or more must come with heavy volume to be listed.
Turnover⊙
INFERENCE — derived by arithmetic
SourceClose times shares traded in the session
As ofsession 2026-11-06
Volume vs median⊙
INFERENCE — derived by arithmetic
SourceSession volume divided by the stock's own median volume
As ofsession 2026-11-06; median over the completed sessions before it, this one excluded
No volume confirmation · volume confirmation 2.7x, needed 3x
How a name reaches this board
Every US equity with a daily bar is ranked, not a curated list. Instruments that are not stocks are removed first: preferreds, warrants, units and rights, along with exchange-traded funds. Then a $1 price floor and a $5M daily turnover floor, both chosen as the least restrictive settings that still removed illiquid prints when the alternatives were measured against each other.
Two judgement guards follow, and only these two produce the log above. The first is volume confirmation. Daily bars are raw traded prices, so a split or a bad print looks like a move; what separates them is that a mechanical change moves price without volume, while news brings volume. A move of 25% or more must come with at least three times the stock’s own median volume across the prior completed sessions, with the current session excluded from that median. The second guard rejects a malformed bar, one carrying a zero open or close, on either end of the two-day comparison.
Both thresholds are opinions, which is exactly why the removals are published. The record has already paid for itself: in August 2026 the malformed-bar guard was scanning the entire comparison window instead of the two bars actually being compared, and it removed a stock that rose 137.9% on $285.5M of turnover, the second largest gain in the market that session. Reading this log is what found it. The guard was narrowed and the board rebuilt the same day. Educational information, not investment advice.
Exclusion log FAQ
Why publish the names you filtered out?
Because a filter nobody can inspect is indistinguishable from a broken pipeline. Two guards stand between the market and this board, and both are thresholds we chose. Publishing what they removed, with the number that removed it, is what makes them checkable rather than something you have to take on trust.
What does the volume rule actually test?
Daily bars are raw traded prices, so a stock split or a bad print can look identical to a huge move. The signature that separates them is volume: a mechanical price change moves price without unusual trading, while real news brings trading. A move of 25 percent or more must come with at least three times the stock’s own median volume from the previous completed sessions, with the current session excluded from that median.
Why the median rather than the average?
An average baseline is contaminated by the very spike it exists to detect. One name traded 24,000, 11,000 and 11,000 shares, then 57.8 million on a 194 percent day. Against the average, its genuine 54.9 percent reversal the next session scored 0.1x and was rejected. The guard was systematically removing day two of every two-day move, which is when a reversal is most worth seeing.
Does this log contain every name that did not make the board?
No, and the page says so above the table. It records candidates that cleared the liquidity and price gates and then failed one of the two judgement guards. Names that never cleared a one dollar price, five million dollars of daily turnover, or the equity-type filter are not individually logged, and neither are the thousands of stocks that simply did not move much. Each session records at most 40 exclusions and the log keeps 12 sessions.
Have these filters ever been wrong?
Yes, and reading this log is how it was found. In August 2026 the malformed-bar guard scanned the whole comparison window rather than just the two bars being compared, which removed a stock that rose 137.9 percent on 285.5 million dollars of turnover, the second largest gain in the market that session. The guard was narrowed the same day and the board rebuilt.