What changed, and when.

Methodology & corrections

Changes to the measurement, and things we got wrong

Every entry below is dated to the day the change merged. If the way a factor works changes, the change is logged here.

1. 2026-10-05correction#1089

   The per-stock blog pages no longer suggest how much a score should count

   The per-stock blog pages had a section headed "Confidence" that said a data-confidence figure was close to how much a reading should weigh in a portfolio. That was advice about position sizing, which Tapeline does not give, and it is gone. The section is now headed "How complete is this score?" and says how many of the six checks had data, the same count the stock pages show. The rest of those pages no longer mentions sizing, conviction or trades.

2. 2026-10-05disclosure#1088

   The public pages use the same plain words as the app

   How it works, the glossary, the data sources page, the transparent-screener and press pages, What's new, the highest-scored list page, the per-stock blog pages and the market regime page now use the app's words: the score levels Very high to Low, the six checks by their plain names, how many checks had data, and the market mood (Calm, Normal, Nervous or Fearful). Headings, page titles and web addresses did not change, and nothing about how a score is built changed.

3. 2026-10-05disclosure#1087

   The market regime is now the market mood

   The dashboard's Market Regime page is now Market mood. Its word reads Calm, Normal, Nervous or Fearful (it read Bull, Neutral, Cautious or Bear), from the same fear-gauge cut-offs: under 15, 15 to 20, 20 to 25, and 25 or more. The page now shows what changed since last week and the last 30 days of the fear gauge, the long-term interest rate and the dollar, from FRED's daily readings. A tooltip that said the regime multiplies every score was wrong and is gone: the market mood is one of the checks in every score.

4. 2026-10-05disclosure#1085

   Stock pages count the checks that had data

   The stock pages now say how many of the six checks had data behind a score, for example 5 of 6, in place of a data-confidence percentage that our two data sources worked out in different ways. The dashboard stock page now leads with the score, what is helping and hurting it, the chart, the news and the next earnings date, with the rest folded under Full details.

5. 2026-10-05disclosure#1083

   Score levels and factors have plain names

   The six score levels now read Very high, High, Above average, Average, Below average and Low (they were High conviction, Strong setup, Constructive, Neutral, Caution and Weak). The six factors are now called price trend, strength vs the market, company finances, insider trades, market mood and recent momentum. The levels, where they start and end, and how a score is built did not change, and the API keeps its labels. In the stock list, the Conf column became How complete: how many of the six checks had data, for example 4 of 6. The sort by confidence was removed with it, and an old link that asks for it sorts by score.

6. 2026-10-05scope#1081

   All signals joined the stock list

   The dashboard's All signals page is now part of Find stocks, in the stock list, and its old links open that list with the same level filter. On the Free plan the list shows the top 10 rows of any scan, as it always has. The dashboard menu went from seventeen items to six, and a new Today page shows the ten highest scores.

7. 2026-10-04scope#1075

   AI assistants no longer look up tickers by name

   Search is for signed-in members, so the AI-assistant server no longer offers a name lookup. An assistant can still read any covered ticker's score by its exact US symbol. Searching by name or symbol now needs a signed-in session.

8. 2026-10-04scope#1058

   Search moved inside the dashboard

   Search is now for signed-in members, inside the dashboard. The public search page and the search icon are gone, and a web address that is not a ticker opens the list of scored stocks instead. Every ticker's own page stays public.

9. 2026-10-04disclosure#1023

   The track record moved inside the dashboard, for paying members

   From this date the track record is for paying members only, inside the dashboard. Its old public pages send you there after you sign in. No recorded entry was changed.

10. 2026-10-04disclosure#1022

    Files you could save from the site were removed

    On 3 October 2026 we removed every file you could save from the site: the record as a spreadsheet or raw data file, and the scanner and watchlist files. Some entries below named those files, and those words were taken out, so no entry points to a file that is gone.

11. 2026-09-26disclosure#939

    We no longer say cancelling takes one click

    Until this date, several of our pages, emails and release notes said a subscription or trial could be cancelled in one click. It takes two: Cancel subscription on the billing page opens a screen that also offers a pause or a discount, and Just cancel my subscription on that same screen completes the cancellation, with no survey required. It can also be done in the Stripe customer portal. Every current page and email now says you can cancel online any time. The release notes below keep the words they shipped with, and the refund policy carries dated notes. Nothing about how cancelling works changed.

12. 2026-09-26scope#921

    Most crypto prices now come from CoinGecko and are re-read about once a minute

    CoinGecko prices began at 23:54 UTC on 25 September 2026. Since then the price shown for most crypto pairs comes from CoinGecko and is re-read about once a minute; a few pairs still show a daily closing price, which can be several days old. Before that, every crypto price was a daily close. For a pair priced by CoinGecko, the one-day move shown is its change over the last 24 hours, since crypto trades around the clock. CoinGecko does not supply any score. Crypto scores are still built from daily closes, on four of the six factors, and rebuilt about once a day where the data allows, so a score can be older than the price beside it: each crypto page shows when its score was last computed, and a pair whose score stops updating leaves the ranked lists. Stock and ETF prices are unchanged. No coin has ever been listed on the published record, and no recorded entry was changed.

13. 2026-09-20scope#897

    The ranked scanner no longer takes price filters or market-data sorts without a signed-in session

    Our ranked scanner answers visitors who are not signed in, capped at ten rows. From this date a request from such a visitor that sets a minimum or maximum price, or asks to sort by one-day, five-day or one-month move or by volume, is refused rather than answered. The reason is that ten rows at a time is enough to recover a price that we are not licensed to publish in bulk: a caller who can ask "which tickers cost between $41 and $43" can narrow that range until only one name remains, and a caller who can order the whole universe by yesterday's move publishes that ordering ten rows at a time. Signed-in visitors, at every plan including the free one, keep every filter and sort, and the default ten rows a logged-out visitor sees are unchanged. This does not mean the scanner shows no prices to a logged-out visitor: it still returns the price, daily move, volume and market cap of the rows it answers with, as the logged-out preview on our home page shows. What changed is only that the filters and orderings above are no longer available without an account. No recorded entry was changed.

14. 2026-09-20correction#898

    Ticker pages for notes, preferred shares and warrants no longer show their issuer's company figures

    A listing that trades like a stock but is not the company's common shares (such as an exchange-listed note, a preferred or depositary share, a warrant, a right, a unit or an exchange test symbol) was shown with its issuer's company-wide figures as if they were its own: market cap, beta, P/E, EPS, dividend yield and, in the app, the Financials tab. A data vendor answers such a symbol with the company's figures, so a preferred share's page could show the company's market cap and its common stock's dividend yield instead of its own coupon. From this date those fields are blank on such a listing, with a line saying why, the market-cap column of the ranked table is blank for it too, and the Financials tab says why it shows nothing. The listing's own trading figures, its prices, ranges and volumes, are unchanged. Listings are recognised by name and symbol, the same rule as the entry below titled "Notes, preferred shares and warrants no longer qualify for the daily record", which does not catch every one. No recorded entry was changed.

15. 2026-09-19methodology#879

    Insider filings now count for every common-stock class of a company. An entry added on 17 September said GOOG and NWSA had none

    From this date a company's Form 4 filings count for every listed common-stock class of that company, and for none of its other listed securities: its preferred shares, notes, warrants, rights, units and exchange-traded notes still carry none, and neither does an ETF. This corrects the entry below titled "Insider filings counted for every security listed under the same SEC filer", which says "Alphabet's insiders file under GOOGL, so GOOG has none, and News Corp's file under NWS, so NWSA has none". Under that entry's rule a filing counted only for the ticker it named, so a company's second class of common stock had no Smart Money reading of its own while its sibling class did. From this date GOOG carries the same filings as GOOGL, and NWSA the same as NWS. Measured on 19 September 2026, 31 such second classes gain their company's filings at their next re-check, about two days for a stock; nothing is re-read early. Lists that combine several companies' filings show each transaction once, with every class it belongs to. Which listing is common stock is judged from its name and ticker, because no data field we hold states it; a short list of listings named exactly like their company's common stock is decided by symbol, such as Strategy's four preferred listings, which carry none, and Bradesco's preferred ADR (BBD). BBD, Bradesco's main traded share, now joins PBR.A and CIG as a preferred listing that stays eligible for the daily record and the default ranked view, as the entry below titled "Notes, preferred shares and warrants no longer qualify for the daily record" describes for those two. The rest of the 17 September entry is unchanged. No recorded entry was changed.

16. 2026-09-19scope#889

    Six surfaces that need no account no longer carry current prices; futures and duplicate Berkshire listings are no longer covered

    From 19 September 2026 these surfaces, which need no account and no key, no longer carry current market data: the public signals, ticker and sector-heatmap JSON endpoints, the MCP server, the embeddable score widget and the image shown when a ticker page is shared. The signals and ticker endpoints and the MCP server still give the score, the signal label, the six factor readings and the order tickers are ranked in; the heatmap endpoint gives each sector's name and ticker count; the widget and the shared image show the score and its label. Prices, daily moves, day and 52-week ranges, volume and market capitalisation are left out, and the JSON endpoints say so in their responses. The MCP server's record tool still returns the recorded closes each listed pick was measured on. Ticker pages on the site and everything in the app are unchanged. The licence note on the embed page no longer offers the widget for free commercial use; it now says the widget shows the score and label only. Separately, 27 continuous commodity futures, such as CL=F, and two duplicate Berkshire Hathaway listings spelled BRK-A and BRK-B are no longer covered. None of the 29 had a price, yet each still carried a score. The futures had none because our market data covers US stocks and ETFs, not futures, and two of the six factors cannot be measured for a future. The two Berkshire rows had none because they spell Berkshire's share classes with a hyphen, while our data vendor spells them BRK.A and BRK.B, and those two rows are priced. None of them had a daily move, so none appeared in the scanner, search or site map before this change either; that exclusion is now explicit, our system no longer requests prices for them, and their pages say they are not covered. Berkshire Hathaway stays covered as BRK.A and BRK.B, and commodities through the USO, GLD, SLV, CPER and CORN funds. No recorded entry was changed.

17. 2026-09-19methodology#878

    Notes, preferred shares and warrants no longer borrow their issuer's fundamentals

    The Fundamentals factor reads reported figures such as margin, return on equity and growth. For a listing that trades like a stock but is not the company's common shares (an exchange-listed note, a preferred or depositary share, a warrant, a right or a unit), the figures a data vendor returns are the issuer's, not the listing's, and until this change they were scored as if they were its own. From this date those listings take no Fundamentals reading: the factor is unavailable for them and the composite uses a mid-range value in its place, as it does for most ETFs. Their scores and labels were recalculated without it, and where the one-line summary had cited fundamentals, it no longer does. They are recognised by listing name and symbol, the same rule as the entry below titled "Notes, preferred shares and warrants no longer qualify for the daily record", which does not catch every one. No recorded entry was changed.

18. 2026-09-19scope#875

    Notes, preferred shares and warrants no longer qualify for the daily record

    Some listings trade on a stock exchange without being a company's common shares: exchange-listed notes, preferred and depositary shares, warrants, rights and SPAC units. They reach our universe the way stocks do, and each is scored on the same six factors. From this date they can no longer be listed on the daily record, and the ranked scanner and the daily picks our MCP server returns leave them out by default. They can still be browsed: the scanner has a control that includes them, every row says whether the listing is one of them, and each keeps its own page. A saved screen now leaves them out unless that control is on. The reason is what their price follows. A note's price is anchored to its face value and a preferred share's to its dividend; a warrant is an option on the common stock, and a unit is tied to a trust. A trend or momentum reading on one of them describes that structure, not a company's shares. Detection is by listing name and symbol, plus a short list of symbols whose names say nothing, such as Strategy's four preferred listings. It is written to under-claim, so some of these listings will not be caught and can still be listed. Exchange-traded notes, which we hold as funds, are not covered by this change, and neither are PBR.A and CIG, the Petrobras and CEMIG preferred shares that are each company's main traded share. These listings are still scored; what changes is that they can no longer be listed on the record or appear in the default ranked view. No recorded entry was changed, so a comparison spanning this date crosses two definitions of what could enter the record, as one spanning 6 September 2026 already does.

19. 2026-09-19correction#876

    Coins could not reach STRONG SETUP because one of the two Trend measurements was on the wrong scale

    Until 19 September 2026 the Trend reading for every crypto pair left out one of its two measurements. Trend averages a multi-month price change with how close the latest price sits to the top of its own 52-week range. For coins, the second measurement was calculated on a different scale from the one the score reads, so every coin counted as being as far from its 52-week high as the scale allows, however close it actually was. That held every coin's Trend reading at 50 or below and put STRONG SETUP out of reach: no coin could score 70. Measured on 19 September 2026, before the change, all 116 scored pairs had a Trend reading of 50 or less, and the highest score among them was 65.0. From 19 September 2026 a coin's 52-week measurement is calculated on the same scale as a stock's, over its last 365 daily closes, which is 52 weeks for a market that trades every day of the week. Every coin's Trend reading and score rise at the next daily crypto run, by more for coins closer to their 52-week high. Coins are still listed separately from stocks and ETFs and are scored on four of the six factors: fundamentals and insider filings have no equivalent for a coin and still count as neutral, so a coin still cannot reach HIGH CONVICTION. No coin has ever been listed on the published record, and from the same change the daily record leaves coins out explicitly, as the scanner's default view already does. So do the other top-pick lists: the daily Top 10 email, the top picks in the welcome and briefing emails, the weekly newsletter's movers and a new account's starter watchlist. No recorded entry was changed.

20. 2026-09-17correction#862, #849, #866

    Insider filings counted for every security listed under the same SEC filer

    Until 17 September 2026 a company's Form 4 filings counted for every security the SEC lists under the same filer. One filer often covers several listed securities: a second share class, preferred or depositary shares, notes, warrants and exchange-traded notes. Measured against our database at 21:57 UTC on 17 September 2026, 2,609 of the 52,897 stored transaction lines were held this way, across 179 of our symbols in 74 groups: Strategy's four preferred listings carried Strategy's insider sales, two notes issued by JPMorgan carried JPMorgan's, and the notes listed as GREEL and TMUSZ carried their issuers'. From 17 September 2026 a filing's transactions count only for the ticker the filing itself names, or, when it names no ticker the SEC lists for that company, the company's first-listed ticker. Every ticker's stored filings are re-read on that rule, and a borrowed reading is removed at that ticker's next successful re-check, which is about two days for a stock and up to about a month for an ETF, exchange-traded note or futures contract. Two consequences are by design. Grouped by name, about 60 of those 179 symbols are preferred or depositary shares, notes, exchange-traded notes or warrants, which should never have carried a company's insider filings: each loses its Smart Money value, and its score is recomputed with that factor counted as neutral. Grouping by name under-claims rather than over-claims, because no data field we hold states what a listing is. The rest are common stock, second share classes of a company whose insiders file under one of them, where only the class the filing names keeps a reading: Alphabet's insiders file under GOOGL, so GOOG has none, and News Corp's file under NWS, so NWSA has none. The methodology page for the factor now states this. The 14 September 2026 entry below titled "Insider Form 4 filings now come from SEC EDGAR instead of a data vendor" says that a ticker with no Form 4 filer of its own on EDGAR has no reading; from 17 September 2026 that is not the only case, because a ticker whose company does file has no reading when the filings name another of that company's tickers. Two further problems are corrected in the same change. From 14 September 2026, when a filer amended one Form 4, every other Form 4 its owner filed that day was dropped instead of only the filing the amendment restates, which left out 13 filings carrying about 110 transaction lines for five companies: Microchip Technology, TransDigm, CoreWeave, Kymera Therapeutics and Similarweb. An amendment now replaces the one filing it restates, together with any identical re-filing of that filing lodged under a second accession. The same 14 September entry says an amended filing replaces the filing it amends, which was the intent and not what our code did. And one stored line was dated 3 September 2027, a filer's typo, which sorted it to the top of every most-recent list; from 17 September 2026 a line dated later than the filing that reports it is dropped. Separately, the "purchases only" filter on the insider lists now means a Form 4 purchase, code P, where it previously also included share grants and option exercises. None of this reached the published record. Of the symbols on the record that held shared filing lines, every one is its own company's common stock, and the only one listed since we began reading SEC EDGAR on 14 September 2026 is IMPP, whose filings are for its common shares. None of the five companies whose filings were dropped has ever appeared on the record, and nor has the company whose filing carried the 2027 date. No recorded entry was changed.

21. 2026-09-17correction#842

    Release notes further down overstated how fresh the data is

    Added on 17 September 2026. Five release notes further down this page describe the data as fresher than it is, and we have left their wording in place. The note dated 30 August 2026, titled "The card moved off the front door", used to say, and still says, that the Free plan "opens the live scanner" ... "on live data, no delay". The note dated 27 April 2026, in the words added when it was superseded, used to say, and still says, "Free is live scores on the top-10 scanner". The note dated 4 May 2026 used to say, and still says, that the public ticker pages "render the live score". The note dated 14 May 2026 used to say, and still says, that each /blog/ticker page "renders the live Tapeline composite". The methodology entry dated 18 July 2026 used to say, and still says, that Free shows "the top 10 live scanner rows". None of those was true. Measured on 14 September 2026: prices on every plan, Free included, are delayed about 15 minutes by our price vendor's plan; since a fix that afternoon, every covered stock and ETF is re-read about every 60 seconds during US market hours; a score is recalculated on each pass, but its inputs are daily readings, so a score usually changes about once a day; and the public ticker pages are saved snapshots that can be an hour old or more. What is true of Free is the top ten scored rows of any scan, on the same delayed prices every plan reads. The change merged on 17 September 2026 (#842) moved the rest of the site's freshness wording to these measured figures. No recorded entry was changed.

22. 2026-09-15correction#851

    The app's Insider tab now lists the filings we read from SEC EDGAR

    Added on 17 September 2026. The entry titled "Corrections to the entry on the switch to SEC EDGAR" says the Insider tab on a ticker's page in the app still lists filings from the data vendor. That stopped being true on 15 September 2026 (#851): the tab now lists the Form 4 filings we store from SEC EDGAR, the same filings behind the Smart Money factor and the insider pages, instead of asking the data vendor when it opens. When we checked on 17 September 2026, every insider filing line we store came from SEC EDGAR. No recorded entry was changed.

23. 2026-09-14correction#848

    Corrections to the entry on the switch to SEC EDGAR

    Added on 15 September 2026, the day after the entry below titled "Insider Form 4 filings now come from SEC EDGAR instead of a data vendor". That entry needs four corrections. First, it overstates the scope: the Smart Money factor and the insider filings we store are read from SEC EDGAR, but the Insider tab on a ticker's page in the app still lists filings from the data vendor. Second, the whole non-derivative table of each filing is counted (Table I, which is mostly common stock), not only common stock; the table of options and other derivative securities is still left out. Third, the switch made every ticker due for a re-check at once, and a ticker's stored filings and Smart Money value change over at its next successful re-check. A ticker whose re-check fails is tried again on its usual schedule, about every two days for a stock and about monthly for an ETF or futures contract; the 95 stored filing lines from the vendor that remained at 21:48 UTC, for 6 tickers, were all for stocks. Fourth, on 14 September 2026 the Smart Money methodology page, the data sources page and the insider pages still said these filings come through a data vendor that can run weeks behind SEC EDGAR; that wording predates the change. No recorded entry was changed.

24. 2026-09-14correction#824, #833

    Some tickers held a Smart Money value with no insider filing behind it, and daily lists were ranked with it

    At 13:30 UTC on 14 September 2026, 856 tickers held a Smart Money value although no SEC Form 4 filing for them was on file: 629 ETFs, 222 stocks and 5 commodity futures contracts. 42 of those values are below 10 or above 90, and the Form 4 calculation only produces values from 10 to 90. Where the values came from has not been established. For these tickers two statements were not true: the Smart Money methodology page says a ticker with no disclosed filings in the window has no reading at all, and the 2026-08-23 entry for #620 says a missing reading stays blank. From 14 September 2026 (#824), when the source we read Form 4 filings from reports none for a ticker in the window, its Smart Money value is removed, its stored filings are deleted, and its score is recomputed with that factor counted as neutral, unless a filing we already hold from that source records a transaction in the last 80 days. A reading whose filings have all left the window is removed at the ticker's next re-check, not at once. A ticker holding a value with no filing on file becomes due for a re-check at the next daily run, ahead of other Smart Money re-checks, instead of on its usual schedule, which for ETFs and futures contracts is about monthly (#833). No recorded entry was changed.

25. 2026-09-14methodology#835, #837

    Insider Form 4 filings now come from SEC EDGAR instead of a data vendor

    From 14 September 2026 the Smart Money factor reads corporate-insider Form 4 filings from SEC EDGAR, the SEC's own filing system, instead of through a data vendor. What is read also changed. Only the common-stock table of each filing is counted, not the table of options and other derivative securities. An amended filing (Form 4/A) replaces the filing it amends instead of being counted as well. A ticker with no Form 4 filer of its own on EDGAR, which includes most ETFs and funds, has no reading. A transaction line whose share count is too large for our database is left out rather than failing the whole ticker (#837). A ticker's stored filings and its Smart Money value change over at its next successful re-check, about every two days for a stock and about monthly for an ETF or futures contract; until then its vendor-sourced filings and value remain. The switch made every ticker due for a re-check at once, and at 21:48 UTC on 14 September 2026, 95 stored filing lines from the vendor remained, for 6 tickers. No recorded entry was changed.

26. 2026-09-14correction#821

    Release notes described congressional-trade and squeeze features that did not work as described

    Three release notes below describe congressional-trade and squeeze features that did not work as described. The 2026-08-30 note says a card turns on congressional-trade filings: all 338,015 rows in our congressional-trades table were test output, not real filings. The 2026-05-17 note says the Smart Money factor explanation was updated to cover congressional disclosures. Starting in May 2026 the factor also read a count from a spreadsheet tab that listed congressional trades among other signals. We cannot now check what that tab held, that input is not connected today (checked 13 September 2026), and the factor is now computed from SEC Form 4 insider filings only. The 2026-08-12 note says squeeze alerts go out over email and browser push. The squeeze data behind squeeze alerts was not real market data: it was test output, last written on 18 July 2026. We have left the wording of those release notes in place and are correcting them here.

27. 2026-09-10correction#798, #800

    Three of the six factors were not refreshed from 6 to 10 September

    Added on 14 September 2026. From 6 September until the fixes merged on 10 September 2026, the refresh of the trend, relative-strength and momentum inputs did not complete, so those factors kept using price data fetched on 6 September, and score updates written by the main loop were being discarded. Nothing marked the scores as stale. No recorded entry was changed.

28. 2026-09-06correction#766

    Renamed spreadsheet columns made many scores too high until 7 September

    Added on 14 September 2026. The scoring spreadsheet renamed three columns (3-month, 6-month and 1-year return). Our reader did not recognise the new names, so those returns were treated as missing and scored as neutral, and nothing raised an alarm. When the reader was fixed, a dry run against the live spreadsheet changed 4,088 of 4,112 spreadsheet-based scores: 3,233 went down and 855 went up. The corrected scores took effect on 7 September 2026. Our stored score history begins on 24 August 2026 and shows the problem from that day, so it lasted from at least 24 August; when it started is not known.

29. 2026-09-06scope#761

    Leveraged and inverse funds no longer qualify for the daily record

    The scorecard freezes each day's ten highest-scoring names and back-checks them against SPY. From this date a geared fund — a 2x, 3x, Ultra, inverse or otherwise multiplied ETF — is no longer eligible to be frozen, and is excluded by default from the ranked scanner. It can still be browsed: the scanner carries a control that includes them, and every row states whether the fund is geared. The reason is that a geared fund's one-day move is a multiple of something else's, so comparing it to SPY measures the multiplier rather than the ranking, which is what the record exists to measure. Detection is by fund name, because no vendor field states gearing; the rule is written to under-claim rather than over-claim, so a fund whose name does not disclose its gearing will not be caught. A comparison spanning this date crosses two definitions of what could enter the record, and should say so.

30. 2026-09-06methodology#762

    Most tickers were being scored on four of the six factors, and now are not

    Two of the six factors — fundamentals and recent insider activity — are gathered by a daily pass over the tracked list. Where a factor has no reading for a ticker, the composite treats that factor as neutral, so it neither helps nor hurts. That fallback is deliberate and is unchanged. What was wrong was how often it applied. The daily pass kept its progress only in memory, so every deployment of the site restarted it from the beginning; it also chose which tickers to fetch by ranking the whole list rather than by asking which tickers were still missing, so each restart re-fetched the same names instead of continuing on. Measured before this change, 5,697 of 7,417 scored tickers — 77 per cent — had no reading for either of those two factors, including 1,001 of the 1,035 scored tickers valued above ten billion dollars. Those two factors together carry a fixed share of the composite, so a ticker missing both had a ceiling on the score it could reach. That was an artefact of how the data was collected. It is not a statement about the companies that were missing, and it was not a judgement the method intended to make. Three things changed. The pass now records which tickers it has already attempted, so a restart resumes instead of repeating, and a ticker the data provider simply has nothing for stops being asked again forever. It now selects the tickers still missing a reading, most-traded and largest first, instead of re-ranking the whole list. And it runs ahead of the passes that fill in display-only fields, because those can wait and a missing factor cannot. Separately, both readings are now restored from stored data when a process starts, rather than starting empty. Expect scores across a large part of the list to move as coverage fills in over the following days, and expect which tickers qualify for the top ten to change with them.

31. 2026-08-23correction#620

    Before this fix, some tickers could be scored from random placeholder numbers

    Added on 14 September 2026. Tickers not covered by our scoring spreadsheet get their factor readings from caches that are empty after every restart of our system and were refilled by a once-daily pass. Until this fix, a missing reading kept a random placeholder value instead of being left blank, so after each restart those tickers were scored from random numbers until the daily pass ran. Since this fix a missing reading stays blank and no score is produced from it. No recorded entry was changed.

32. 2026-08-23scope#568

    Per-ticker pick history now applies the same 7-day publication delay

    Three public surfaces serve the per-ticker scorecard rows. Two of them, including the scorecard page, withhold the most recent seven days from readers not on a paying tier, and say so. A third endpoint — the one backing the score sparkline on ticker pages — served the same rows with no delay at all. It now applies the identical seven-day delay for the same readers, so every public surface publishes the record on one schedule. No historical entry was changed.

33. 2026-07-18scope#360

    Stopped collecting investing experience and portfolio size

    Onboarding previously asked for investing experience level and portfolio size. Both questions were removed and are no longer collected anywhere on the site. Tapeline does not ask for capital, holdings, risk tolerance, experience or investment goals, which is what keeps its output general information rather than something tailored to an individual.

34. 2026-07-18scope#345, #348, #349

    Free tier restated to match what the product actually enforces

    Pricing and comparison surfaces were corrected to describe the free tier as it is enforced in the backend: 12 ticker look-ups per day, a 5-slot watchlist, the top 10 live scanner rows, a top-3 squeeze preview, 2 web-push alert rules, and the full public scorecard. Several pages had been describing older, tighter limits that no longer applied. The enforcement itself was widened the previous day, on 2026-07-17.

35. 2026-07-12disclosure#342

    Exact factor weights, the scoring equation and indicator lists removed from the site

    The public site previously published the numeric weight of each factor, the composite equation, and per-factor indicator and parameter lists. Those were removed. What is still published: the six factor names, the relative ordering of the weights, each factor's contribution on every ticker, and the full public scorecard. The reason is competitive, not evasive — the numeric recipe was the part a competitor could copy directly. The scoring behaviour itself did not change on this date; only what the site discloses about it.

36. 2026-07-11correction#337, #359

    Prescriptive advice and performance claims removed from live pages

    A review found copy on live pages that told readers what to do with a score, and copy that implied the product produces returns. Both were removed. Signal labels describe the state of the factor data; they do not prescribe an action. A follow-up on 2026-07-18 added an automated check that fails the build if that language reappears.

37. 2026-07-09methodology#329, #330

    Liquidity floor applied to the ranked scanner and the scorecard

    High-scoring names that barely trade could top the ranked list and be frozen onto the scorecard. A liquidity floor now filters them out of both. This changed which names QUALIFY going forward only. Names with no volume reading are left in, and the floor can be switched off on the scanner to browse the full scored universe.

38. 2026-05-17correctionv0.1.12

    Smart Money was described as 13F holdings; it reads SEC Form 4

    Marketing copy across pricing, comparison and per-ticker pages described the Smart Money factor as using 'elite 13F holdings'. That was wrong. The factor reads corporate-insider transactions disclosed on SEC Form 4, which is what the underlying code had been doing all along. The copy was corrected to match the implementation rather than the other way round.

What the score means is on how it works. What the product as a whole is weak at is on limitations. Found something wrong? Email support@tapeline.io and the correction gets logged here with a date.

Releases

What we shipped, when

1. The card moved off the front door — sign up with an email and a password

   improvement

   v0.1.21 · 2026-08-30

   • ·Signing up takes an email and a password. The account it makes lands on the Free plan and opens the live scanner straight away: the top ten scored rows of any scan, on live data, no delay. Free also carries one saved screen, a 5-symbol watchlist, and 12 ticker deep-pages a day.
   • ·A card is what starts the 30-day Premium trial, and it is what turns on every matching row instead of the first ten, a second saved screen, alerts on email and browser push, the 200-symbol watchlist, and congressional-trade and insider filings. Free now carries no alerts on any channel, browser push included.
   • ·The trial terms are unchanged: $0 charged that day, the exact first-charge date shown before you confirm, an email three days before the charge, and one click to cancel.
   • ·This supersedes the arrangement in the 22 August 2026 entry below, where a new account added a card at first sign-in before the logged-in product opened. That entry stands as a record of what shipped then. The published record — daily top 10, scorecard, per-ticker pages — is unchanged and still needs no account.

2. Premium trial takes card details up front — $0 today, reminder email, one-click cancel

   improvement

   v0.1.20 · 2026-08-22

   • ·From 22 August 2026, a new account adds a card at first sign-in and starts the 14-day Premium trial. $0 is charged on day 0; the first charge lands on day 14 at the plan picked at sign-up.
   • ·A reminder email goes out three days before the first charge, and cancelling is one click on the first screen of the cancel flow — cancel before day 14 and no charge is ever made.
   • ·Existing accounts are unaffected, and the published record — daily top 10, scorecard — stays free to read with no account.

3. Ticker pages rebuilt as a decision aid

   shipped

   v0.1.19 · 2026-08-22

   • ·Key statistics on every ticker page — relative volume, 52-week range, market cap, next earnings date, dividend yield, beta, P/E and EPS — populated from the same feeds that drive the score. Fields we hold no peer ranking for are labelled as such rather than presented as judgeable.
   • ·The composite score and all six factors are now ranked against sector peers, with the peer group and its size printed next to every percentile. Rankings with fewer than 30 peers say so instead of printing a number a single peer could swing.
   • ·A six-factor radar shows the shape of the score at a glance — balanced or lopsided — using the same factor order as /how-it-works.
   • ·The ticker's own published record is on the page: every time it appeared in a daily top 10 and how each entry resolved, losses included. Same data as the public scorecard, scoped to the ticker you're looking at.

4. Open-access month: 1,000 ranked scanner rows on Free until 8 September

   improvement

   v0.1.18 · 2026-08-20

   • ·Until 8 September 2026, signed-in Free accounts see 1,000 ranked scanner rows instead of the top 10 — the same row cap the paid plans use. On 8 September the cap returns to the top 10 automatically.
   • ·Nothing else about the Free plan changes: same daily look-ups, watchlist, alerts and public scorecard.

5. Public MCP server — connect Tapeline to Claude, ChatGPT or any MCP client

   shipped

   v0.1.17 · 2026-08-20

   • ·Tapeline now runs a Model Context Protocol server at api.tapeline.io/mcp. Any MCP-compatible AI assistant can read the record directly: five public tools cover a ticker's six-factor score, today's published top 10, the full track record with its sample-size qualifier, a single ticker's pick history with losses included, and name-to-symbol search.
   • ·Free, no account, no API key. Setup instructions — including the three-line config for Claude and ChatGPT — are at tapeline.io/mcp.
   • ·The server states the same caveats as the site: an assistant that calls the tools and a reader of the scorecard page get the same record with the same qualifiers attached.

6. Telegram alerts retired; alerts consolidated on email + browser push

   improvement

   v0.1.16 · 2026-08-12

   • ·The Telegram alert channel has been retired. Score-change, squeeze and regime alerts now go out over email and browser push, which cover the same triggers. Existing rules keep firing on those channels — nothing about your watchlist, billing or the feeds changes.

7. Rankings + scorecard now filter out untradeable names

   improvement

   v0.1.15 · 2026-07-10

   • ·The scanner and the public scorecard now apply a liquidity floor. High-scoring names that barely trade — a bond or single-strategy ETF changing hands a few hundred dollars a day — no longer top the ranked list or get frozen onto the scorecard. A score you can't realistically act on isn't useful, so the tradeable names come first now.
   • ·Methodology note, in keeping with our published practice: this only changes which names qualify going forward.
   • ·Names with no volume reading are left in, so the filter only ever removes rows that are genuinely untradeable. On the scanner you can still turn the liquidity floor off to browse the entire scored universe.

8. Founding pricing: Pro $9.99/mo, Premium $19.99/mo · 30-day money back

   improvement

   v0.1.14 · 2026-07-03

   • ·New founding pricing across every plan: Pro is $9.99/mo or $99/yr ($8.25/mo effective), Premium is $19.99/mo or $199/yr ($16.58/mo effective). Tapeline is new and the price now says so — subscribers keep their rate for as long as the subscription stays active. No countdowns, no fake scarcity.
   • ·Money-back guarantee extended from 7 days to 30 days on every paid plan — email support@tapeline.io within your first 30 days for a full refund, no forms.
   • ·Feature gating unchanged: same tiers, same limits, same 14-day no-card Premium trial.

9. Public /data-sources page; squeeze list back to populating

   shipped

   v0.1.13 · 2026-05-17

   • ·New /data-sources page lists every feed that powers a Tapeline score — what it's used for, where it shows up in the product, how often it refreshes, and whether it's public-domain or licensed. Six feeds covered end-to-end: prices, fundamentals, SEC filings, macro, news, and the composite scoring workbook. Linked from the footer's transparency strip.
   • ·Squeeze list on /app/squeeze now populates again. An upstream schema change meant the page was rendering an empty table for ~12 hours; the parser now reads the new format and the spike scores are back live.
   • ·Licensing posture written into the data-sources page as a Public domain vs Licensed badge per feed — same standard as the public scorecard and signal-methodology disclosures.

10. Honest about what powers Premium — insider Form 4, not elite 13F

    improvement

    v0.1.12 · 2026-05-17

    • ·Premium feature line updated across pricing, comparison pages, and per-ticker share pages: 'Elite 13F holdings' replaced with 'Recent insider buys (SEC Form 4)'. What's on the /app/holdings page hasn't changed — same SEC Form 4 transactions that have been live for weeks — the marketing copy just now matches what's actually delivered.
    • ·Smart Money factor explanation updated everywhere: Form 4 insider transactions + Congressional disclosures. No mention of 13F.
    • ·Open API roadmap, comparison tables, and the LLM-readable site facts all reconciled. No promise the product can't deliver.

11. Live signal-system universe, iOS-feel design, light + dark mode

    shipped

    v0.1.11 · 2026-05-17

    • ·Tapeline now reads the full ranked ticker universe live from its source of truth — refreshed every 5 minutes. The scanner went from 112 mock tickers to 7,162 tickers tracked, 4,399 of them actively scored, including names like HYLN that were missing before.
    • ·Light, Dark, and System appearance modes — toggle in your account menu (top-right). System mode follows your Mac/Windows theme automatically as you change it.
    • ·Heatmap rebuilt around 14 clean GICS sectors (Information Technology, Health Care, Financials, Industrials, Consumer Discretionary, Consumer Staples, Communication Services, Energy, Materials, Utilities, Real Estate, Commodities, Funds & ETFs, Uncategorized). Was 51 fragmented labels; the dropdown filter now actually matches what's on the page.
    • ·Search box on both the scanner and the heatmap — type any ticker substring, results filter live.
    • ·Material event filings from SEC EDGAR now appear in the breaking-news bar ~5-30 minutes before they're re-reported by the news wires. 8-Ks, item codes, the lot. Free, regulatory-quality, no marketing fluff.
    • ·Recent insider buys page (Premium) populates correctly now — was showing empty regardless of the data behind it.
    • ·Per-ticker page shows the proper 'No fundamentals coverage for ETFs and funds' message instead of six empty cards when a ticker is an ETF.
    • ·New design language across the dashboard: translucent surfaces, frosted-glass cards, system-font stack, pill buttons. No more boxed-template borders.
    • ·Fluid type — the whole UI compresses cleanly as you narrow the window. Cell padding tightens on the scanner table so the numeric grid stays the focus on laptops and small screens.
    • ·News bar now shows 3 headlines at once (was 1 rotating) and pulls 20 per refresh (was 8) so the feed actually feels active.
    • ·Authenticated actions like Add to Watchlist and Notify on News no longer return errors — a cross-origin cookie issue that affected most dashboard buttons.
    • ·Four Pro/Premium endpoints (Congress, Heatmap, Squeeze, Regime) that were accidentally accessible without authentication despite the UI gating them — now properly gated end-to-end.
    • ·Top-right search shortcut chip now reads 'Ctrl K' on Windows and Linux. Was the Mac Command symbol everywhere.
    • ·New /app/account hub — your username menu links here for a single home for billing, watchlist, alerts, email preferences, and the new appearance toggle.

12. New tagline, clearer privacy policy, payment-failed email

    improvement

    v0.1.10 · 2026-05-15

    • ·New product tagline across the site: Read the tape.
    • ·Privacy policy rewritten to match what Tapeline actually does today — no stale references to tools or sub-processors no longer in use, and a clear note that request IP addresses live in memory only, not in any persistent log.
    • ·New email when a renewal charge fails — a calm note with a one-click link to update your card, the moment the failure happens. Stripe's automatic retry behaviour is unchanged; the difference is you find out immediately rather than when access drops.

13. Referrals, signup hardening, 50 per-ticker SEO pages

    shipped

    v0.1.9 · 2026-05-14

    • ·Referral mechanic live. Share your code from /app/referrals — both you and the friend you refer get one free month of Premium, credited at the next paid checkout.
    • ·Signup form is more forgiving with intermittent Turnstile loads, so fewer real users get bounced. Signing out now reliably clears the session across every Tapeline subdomain.
    • ·50 new per-ticker landing pages at /blog/ticker/{SYMBOL} — AAPL, NVDA, MSFT and 47 others. Each renders the live Tapeline composite, the 6-factor breakdown, and a structured FAQ so the page is useful in search results.
    • ·Referrals page now shows 0 unused credits cleanly if you've never referred anyone, instead of a momentary blank.

14. Sharper Pro vs Premium presentation, faster app pages

    improvement

    v0.1.8 · 2026-05-04

    • ·Premium plan card now leads with what it adds on top of Pro instead of duplicating the Pro list.
    • ·Loading states across the app now show shimmer placeholders in the shape of the eventual content — the product feels noticeably snappier on first visit.
    • ·Per-ticker public share pages (/t/AAPL, /t/NVDA etc.) render the live score and 6-factor breakdown without needing a sign-in. Share-on-X button included.
    • ·Public scorecard, status page, and blog all reachable from every page footer.

15. Real earnings + IPO calendars; live macro indicators

    shipped

    v0.1.7 · 2026-05-03

    • ·Earnings calendar now reflects 1,500 upcoming reports across the next two weeks, not a sample.
    • ·IPO calendar shows actual upcoming listings (Rare Earths Americas, HawkEye 360, etc.) sorted by expected date.
    • ·Macro tile (DXY, 10-year, VIX) on the regime page now pulls live macro indicators instead of static defaults.
    • ·Per-ticker confidence pill on the scanner — mega-caps with full coverage land 90+; less-followed names sit lower so you can deprioritise signals built on thin data.

16. Browser push alerts + end-of-day digest

    shipped

    v0.1.6 · 2026-05-02

    • ·Browser push notifications — lock-screen alerts on desktop and Android, one click to enable. Free at any volume on Pro+.
    • ·End-of-day watchlist email digest fires every weekday after market close for every Pro+ user with watchlist items.
    • ·Continue-with-Google sign-in option now appears on the signin and signup pages.

17. Elite 13F holdings, harder Free tier, three competitor comparisons

    shipped

    v0.1.4 · 2026-04-27

    • ·Elite-fund 13F holdings live for Premium: latest positions from Buffett, Burry, Tepper, Ackman, Druckenmiller, Laffont, Coleman, Singer.
    • ·Free tier reworked at the time so the trial-end transition was meaningful. (Since superseded: Free is live scores on the top-10 scanner with 12 look-ups/day, and where the card is asked for changed on 22 August 2026 and again on 30 August 2026 — see those entries, and /pricing.)
    • ·Plain-English Why on every scanner row rewritten with sector-aware language across ~100 phrase variants.
    • ·Side-by-side comparison pages live for Finviz, Zacks, and WallStreetZen.

18. Telegram alerts (Premium), commodity ETF universe, three-tier pricing

    shipped

    v0.1.3 · 2026-04-26

    • ·Telegram alert channel for Premium subscribers — paste a chat ID on the billing page, send a test, get hourly market-regime + watchlist digests plus per-rule alerts. Retired since: see the 12 August 2026 entry — alerts now run on email and browser push.
    • ·32 commodity ETFs added to the scanner universe (gold, silver, oil, gas, agriculture, copper, uranium, miners) with a dedicated Commodities sector filter.
    • ·Pricing simplified to three tiers: Free, Pro $29/mo, Premium $49/mo (or $24.99 / $39.99 billed annually). These 2026 figures are HISTORICAL and no longer apply — Tapeline now costs $9.99/mo for Pro ($99/yr) and $19.99/mo for Premium ($199/yr).

19. Public scorecard live + descriptive signal labels

    improvement

    v0.1.1 · 2026-04-24

    • ·Public scorecard goes live: every top-10 we publish is back-checked against the next-day move alongside SPY for the alpha column. No cherry-picking.
    • ·Signal labels (HIGH CONVICTION, STRONG SETUP, CONSTRUCTIVE, NEUTRAL, CAUTION, WEAK) tightened to descriptive language so the score communicates the data rather than prescribing action.

20. Tapeline launches

    shipped

    v0.1.0 · 2026-04-22

    • ·Six-factor scoring methodology live with all six factors published on /how-it-works.
    • ·Scanner with one composite score and one plain-English Why on every US-listed ticker.
    • ·Watchlist with smart score-change alerts.
    • ·30-day Premium trial available with no credit card. (Since superseded: the trial is card-required. Where the card is asked for changed on 22 August 2026 and again on 30 August 2026 — see those entries, and /pricing.)