Reference
Glossary
Every score, pattern, filing type, and measurement on this site, in plain English. Anywhere you see a dotted underline in the product, hovering or tapping it shows the same definition.
Scores
Notability
Our 0-to-1 score for how much a trade stands out from routine activity — based on things like dollar size, how big it is relative to the person's existing stake, whether it's their first-ever trade in the stock, whether several insiders acted at once, and their role (a CFO buying means more than a board member). High notability means 'unusual, worth attention' — it deliberately does NOT claim the trade predicts returns. That claim has to be earned separately (see Validation).
Validation
The earned half of every read. We track what happened to the stock after every past trade of the same kind (same pattern, same kind of actor), then run statistical tests to ask: did this KIND of signal genuinely beat the market, or is it luck? Bands run None → Preliminary → Low → Medium → High, and most signals honestly sit at None or Low today because real evidence takes time to accrue. Notability and Validation are kept separate on purpose: one says 'unusual', the other says 'proven' — we never blend them into one misleading number.
Actionability
The ordering used in 'What's New': the notability score, reduced as a signal ages (a disclosure loses relevance day by day) and reduced if the trade was pre-scheduled rather than a deliberate choice. A fresh moderate signal can outrank a week-old strong one.
two-axis read
Every signal shows two separate judgments: how unusual it is (Notability) and whether its kind is proven to work (Validation). We never merge them into one number.
skill-adjusted
Raw hit rates lie on small samples: someone who's 3-for-4 looks like an 75% winner but that's four coin flips. The skill-adjusted rate shrinks the raw number toward a coin flip in proportion to how little data there is, so a track record only looks impressive once there's enough history to deserve it.
thin sample
Too few resolved trades to judge skill — the record is shown, but treat it as anecdote, not evidence.
Signals & patterns
Convergence
The marquee multi-source pattern: distinct informed groups — corporate insiders, Congress members, big funds, the company itself via buybacks — net-agreeing on one stock's direction within the window. Each convergence carries a chance-alignment receipt: how often this same agreement would show up anyway given how bullish or bearish each group has been lately. Agreement against the grain scores high; agreement everyone's making scores low.
Divergence
Informed groups DISAGREEING on the same stock — one buying while another sells. Also worth knowing; the tension is shown, never netted away silently.
Hot hands
This specific person's past trades have actually worked — their resolved track record beats a coin flip by enough, on enough trades, to matter.
Aberration
A trade that's out of character — first-ever in the stock, unusually large, a big chunk of the person's stake, or a lone actor stepping out.
Timing & proximity
A burst of trades by the same person in the same stock within days — concentrated activity, flagged as descriptive (no prediction claim).
Informed exit
A sell that looks deliberate: a large slice of the stake, several insiders exiting together, or a proven performer reversing. Small routine sells are ignored — they're usually just cashing paychecks.
Cluster of buyers
Three or more different insiders buying the same stock in the same window — much harder to dismiss than one person's opinion.
Buying the dip
An insider buying while the stock sits well below its recent high — conviction into weakness.
Selling into strength
A meaningful, deliberate (not pre-scheduled) sale right near the stock's recent high — taking chips off at the top.
Crowded-short setup
Insider buying in a stock that short sellers are crowded into — if the stock rises, shorts may be forced to buy back, amplifying the move.
Plan terminated
The insider killed their own pre-set trading plan — a deliberate change of mind, one of the cleanest tells in the post-2022 rules.
Planned sale executed
The insider filed notice they intended to sell (Form 144), then actually sold — the bearish intent was followed through.
Planned sale abandoned
The insider filed notice to sell, then DIDN'T within 90 days — walking away from your own planned sale is a quietly bullish change of mind.
Buyback follow-through
The company announced a buyback AND its next quarterly filing shows it actually spent more buying shares — the program is real, not just a press release.
Buyback cheap talk
The company announced a buyback but the next quarterly filing shows no rise in actual buying — announcements without execution are a bearish tell.
descriptive
We're describing something that happened — we are NOT claiming it predicts returns. Prediction claims must pass the validation gate first.
Filings & sources
Form 4
The SEC form corporate insiders (executives, directors, 10%+ owners) must file within 2 business days of trading their own company's stock. Our core insider feed.
Form 144
A notice an insider files BEFORE selling restricted stock — intent to sell, filed ahead of the sale itself. The leading edge of the sell signal.
PTR
Periodic Transaction Report — the disclosure members of Congress must file within 45 days of a trade. Amounts come in ranges, not exact figures; we use the midpoint.
13F
The quarterly filing where big investment funds (over $100M) reveal their stock holdings — 45 days after each quarter ends. How we track what the whales own.
13D/G
Filed when an investor crosses 5% ownership of a company — 13D signals activist intent, 13G passive. A loud stake-building signal.
8-K
A company's 'something material happened' filing — buyback announcements, executive changes, big contracts. We read these for catalysts.
10b5-1 plan
A pre-set automatic trading schedule insiders adopt in advance. Trades under a plan are mechanics, not opinions — we down-weight them (and flag when a plan is killed).
insider
In the legal sense: an officer, director, or 10%+ owner who must disclose trades in their own company. Everything here comes from those public disclosures.
whale
A large fund whose quarterly 13F holdings we track — activists, famous stock-pickers, and a few giants for context.
source licensing
Some data sources are public-domain (SEC, Congress); others are licensed and can't be shown publicly. Public pages only ever show the former.
Market terms
abnormal return (vs SPY)
When we say a trade was 'right', we don't mean the stock simply went up — in a rising market almost everything goes up. We estimate how much of the stock's move the market itself explains (using each stock's own sensitivity to the S&P 500), and score only the excess. A 'right' bullish trade beat the market; a 'right' bearish exit dodged a stock that lagged it.
20-session window
We score each trade by the stock's market-beating return over the 20 trading days AFTER the disclosure became public — never from the (earlier, private) trade date.
Net flow
Buys minus sells, in dollars, across all disclosed trades in the window. Positive = accumulation, negative = distribution.
informed flow
Net flow with pre-scheduled (10b5-1) trades counted at reduced weight — deliberate money counts more than automatic money. This is what ranks the boards.
Acceleration
Is the informed money speeding up? Recent-window flow minus the prior window's — positive means building momentum.
breadth
How many DIFFERENT people/institutions are behind a signal. Five separate buyers beats one buyer five times — and it's the anti-manipulation check.
days-to-cover
How many days of normal trading it would take short sellers to buy back everything they've borrowed. High = a crowded short — fuel for a squeeze.
drawdown
How far the stock sits below its recent high — bought '30% below the high' means bought well into weakness.
unscheduled
A trade the person CHOSE to make now (vs one executed automatically under a pre-set plan). Choices carry information; schedules mostly don't.
pre-planned
A trade executed on a pre-set schedule (10b5-1 plan or regular cadence) — mostly mechanics, so we down-weight it.
How we measure
disclosure day
The day the trade became PUBLIC — always our starting line. Measuring from the earlier trade date would use information nobody could have acted on.
point-in-time
Every historical read uses only what was publicly knowable on that date — no hindsight allowed to leak in. The discipline behind every backtest here.
q-value (FDR)
When you test hundreds of strategies, some look great by pure luck. The q-value corrects for that: q=0.03 means ~3% of discoveries at this bar are expected to be flukes.
chance alignment
The receipt on every convergence: how often this same agreement would occur anyway, given how bullish/bearish each group has been lately. Low = genuinely surprising.
resolved trade
A past trade whose 20-day scoring window has fully closed, so its outcome is final and counted in the track record.
receipts
Every figure links to the underlying public filing it came from. Nothing here is asserted without a source you can click.
catalyst
A disclosed company event near the trade — an 8-K, a contract award, an FDA decision — shown as context for WHY the timing might matter.
held (quarantine)
Data we couldn't verify is HELD with a reason code, never guessed at and never silently dropped. You'll sometimes see counts of held rows — that's honesty, not breakage.
How the numbers are computed end-to-end — including what we deliberately do NOT claim — lives on the methodology page.