How to read the numbers
How the scores work
Tickertail puts two numbers on things. Both run 0-100. This page is the reader's guide: what each number means, what feeds it, and how seriously to take it. One thing it is not: the recipe. The exact weights are ours - what you get instead is every definition, every source, and a live report card we can't edit.
The Signal Score
"should I care about this one trade?" · per event · 0-100Every disclosed trade gets one. The Signal Score measures how much a single filing deserves your attention: a large, deliberate, out-of-character buy scores high; routine portfolio shuffling scores low.
It weighs the things you'd weigh yourself: whether it's a buy or a sell, how much money is on the line and how unusual that is for this filer, whether the filer's day job overlaps the industry they're trading, whether colleagues pile into the same name around the same time, and whether the pattern looks like a person making a decision or an advisor rebalancing on autopilot. Conviction is marked up; mechanical noise is marked down - and the card says which is which.
Reading the number: under 40 - probably noise. 55+ - worth a look. 80+ - the reason this site exists. The score caps at 100.
The CAT Scan
Conviction Across the Tape · per stock · 1 to 99One number per stock, read like a dial. 50 means the filings say nothing either way. The further above 50, the more the people who must tell the public what they did with this stock have lately been buying it on purpose; the further below, the more they have been selling it on purpose.
- Three groups, nobody else. Company insiders' Form 4 trades, the quarterly 13F reports of the funds we track, and members of Congress's periodic transaction reports.
- Decisions, not paperwork. Open-market purchases and sales count. Grants, option exercises, tax withholding, gifts and trades scheduled in advance under a 10b5-1 plan do not.
- Measured against the filer's own habits. A habitual small purchase counts less than an unusually large one.
- From the day it became public. A filing counts from the day it was filed, not the day of the trade, and fades with time. With no new filings the number drifts back toward 50. Several people acting the same way count more than one, and the strongest readings need more than one group to agree.
- The receipts, not the recipe. Every stock page lists which groups lean which way and the filings behind them. How much each filing weighs stays ours.
| Number | Word | What it means |
|---|---|---|
| 70 to 99 | Heavy buying | More than one group bought with conviction, and recently. Rare. |
| 59 to 69 | Buying | A decision-grade purchase or two, still fresh. |
| 42 to 58 | Mixed | Quiet, routine or stale filings, or buying and selling that cancel out. |
| 31 to 41 | Selling | Deliberate selling by people who did not have to sell. |
| 1 to 30 | Heavy selling | Several insiders selling together, often with a fund exit beside them. Rare. |
Thin tape and silence. When one filer stands behind a number barely off 50, it shows grey with the words "Thin tape" instead of a band. A stock with no filing that counts in the last year has no number at all: we print nothing rather than a 50.
Recipe v2.0, recomputed every night from the filings alone: it never reads the price, the news or a model. Not yet validated against later returns, so it is a reading of what filers did, not a forecast or a signal.
The Copy Score
"what would copying them have earned me?" · per memberUnlike the two scores above, this one is not a recipe: it is arithmetic you could redo yourself from the same filings. Buy every purchase a member disclosed, at the closing price on the day it became public, and hold. The Copy Score is the average return of those positions, against the S&P over the very same windows. Four things to know about how it reads:
- It never counts a price you could not have paid. The clock starts at disclosure, not on the trade date, a price you could not have paid.
- Buys only. Copying a member means buying what they bought. We do not claim their personal profit or loss: congressional disclosures are dollar ranges with no cost basis, so nobody can compute it honestly, us included.
- The number of buys is always printed next to it. An average over 177 purchases and an average over one are not the same claim, and they will never look alike on this site.
- Under 5 buys we call it a small sample. The figure is still real and we still show it, because it is the member's own number. But it does not rank on the Congress or Stocks boards, it does not count toward anyone "beating the S&P", and it never reaches the leaderboard. One lucky purchase is not a track record.
What feeds the scores
official filings and closing prices · nothing elseBoth scores are computed from public, official filings plus end-of-day prices. Every card that shows a real disclosure links the filing it came from, so you can always check our reading against the record.
| Feed | Source | How fast it arrives |
|---|---|---|
| House trades | Periodic Transaction Reports, Clerk of the U.S. House of Representatives | by law, up to 45 days after the trade |
| Senate trades | Electronic financial disclosures (eFD), U.S. Senate Office of Public Records | by law, up to 45 days after the trade |
| Insider trades | SEC Form 4 filings, via EDGAR | two business days after the trade |
| Fund holdings | SEC Form 13F filings, via EDGAR | quarterly, up to 45 days after quarter end |
| Prices | Marketstack end-of-day closes | after each trading day |
The math - scores, returns, every number about a real company, person or fund - is computed from those filings and end-of-day prices by regular code. It is never AI-generated. Two things on the site are made up and labelled so: the example stock on the locked stocks trail, and the worked example on the home page.
Does it actually work?
we grade our own homework - live, receipts attachedEvery priced congressional buy on this site already carries its return against the S&P 500. So here is our own score, marked against reality: if higher-scored buys don't out-perform lower-scored ones, you deserve to see that too.
| Signal Score bucket | Positions | Avg vs S&P (pts) | Median (pts) | Beat the S&P |
|---|---|---|---|---|
| 0-39 (noise) | 2,457 | -3.5 | -7.6 | 36% |
| 40-54 | 509 | -0.9 | -6.4 | 37% |
| 55-79 (worth a look) | 192 | -3.8 | -8.1 | 36% |
| 80-100 (screaming) | 6 | +12.2 | +3.8 | 67% |
3,164 positions (one member × one ticker = one bet) folded from 5,582 priced congressional buys - members often refile the same bet many times, and counting filings instead of bets would let one repeated position masquerade as a trend. A position sits in the bucket of the hottest score it printed; its return is measured from the close on the day its filing became public, the first price a reader could have paid, to the latest close, minus the S&P 500 over the same window. Recomputed hourly. Caveats, honestly: short history, wide disclosed ranges, option bets are measured on the underlying stock (understates leverage), and the top bucket is still small. A diagnostic, not investment advice.
From the trade date: the same report card measured from the member's own trade date
| Signal Score bucket | Positions | Avg vs S&P (pts) | Median (pts) | Beat the S&P |
|---|---|---|---|---|
| 0-39 (noise) | 2,330 | -6 | -13.1 | 33% |
| 40-54 | 480 | -2.6 | -8.6 | 37% |
| 55-79 (worth a look) | 191 | -5.2 | -10.1 | 31% |
| 80-100 (screaming) | 6 | +15.1 | +7.9 | 67% |
3,007 positions from 5,367 priced buys, each measured from the member's trade date to the latest close, minus the S&P 500 over the same window. Nobody reading the filing could have bought at that price; the gap between the two tables is what the disclosure lag costs.
Our honesty rules
the moat- Copy math starts at disclosure.When we say what copying a trade would have earned, the clock starts the day the filing became public, at the closing price - never the day of a trade you couldn't have seen. A trade's details on the stocks trail also show the stock's move from the day the filing names to the day it became public, and from each to the latest close: the trade day for Congress, the filed price for an insider, the quarter's closes for a fund. Those figures sit beside the disclosure figure, are each labelled as a price a reader could not have paid or acted on, and are never called anyone's profit or loss.
- The lag is drawn on every card. Every trade card shows the gap between the trade date and the day it became public. The delay is part of the story, so it lives on the card, not in a footnote.
- Missing data shows as missing.No score is ever invented to fill a gap: if a piece of data can't be confidently resolved, the signals that need it simply don't fire, unscored trades show a pending ring, and thin tape is labeled thin tape. We never guess.
- What a price figure in a trail trade's details measures.For a congressional trade: the close on the trade day the filing names, the first close on or after the day the filing became public, and the latest close, with the stock's move between each of them, each beside the S&P 500 over the same days. On a purchase, the move from the day the filing became public to the latest close is the copy return. When the same member or insider traded the stock the other way later, the details also show the stock's move between the two trade days, beside the S&P 500 over the same days, labelled as prices a reader could not have acted on. For an insider: the same, with the first price the one stated on the Form 4. For a fund: a 13F names a quarter, not a day, so the figures run from the quarter's lowest, average and highest daily close, each to the close on the filing date and to the latest close, beside the move from the filing date to the latest close. A filing can arrive after the closing bell; the figure starts at the close of the day the filing is dated. Percentages use split-adjusted closes; dollar prices are shown as they traded that day, except where a stock split falls inside the dates the details cover; they then say their prices are split-adjusted. Price moves only: dividends are not counted, for the stock or for the S&P 500. When a close is missing, is out of line with the days around it, or the stored history is not comparable across a date, the details say so and print no figure.
- We grade our own homework. The back-test above recomputes from live data - it is allowed to embarrass us.
- The recipe is versioned. The weights are hand-set and versioned internally. When they change, every score on the site recomputes under the new version, so numbers stay comparable.
A note on what changed: until August 2026 this page printed the full weight table. It no longer does - the recipe is now proprietary; that's the honest trade-off of running a business on it. The definitions, the sources, and the live back-test stay public.
The scores are a reading of public filings, not a recommendation. Tickertail’s data and scores are for information only, not investment advice. Investing involves risk, including the possible loss of the money you invest. Consult a licensed financial professional before making investment decisions.