The material changes behind the product and its published research: launches, methodology revisions, performance restatements, and corrections that can change what a reader sees.
Last updated: August 22, 2026
What belongs here
We record changes that affect a product's availability, a strategy's reported history, the interpretation of a reading, or a subscriber's use of the service. Routine copy edits, visual refinements, and internal experiments do not belong in this log.
Research transparency does not require publishing the strategy recipe. Entries describe the category, reason, effective date, and visible effect of a material change. Exact formulas, factor weights, windows, thresholds, ranking rules, and implementation code remain proprietary.
Launch
Market Questions: dated answers to six questions traders ask
A new section answers six plain-English questions every trading day: is the rally broad, is this a bull or bear market, is market risk elevated, are investors too bullish, are small caps confirming the S&P 500, and is a recession coming. Each answer counts several of the site's tools, each read against a rule printed on the page.
The verdict is a count of gauges, never a blended score, so every part of it links to its own tool.
Each page shows how often today's reading has happened before and what followed every verdict, against all days in the same window. Several results run against intuition, and the pages report them as found.
Economic gauges are dated by the day they were published, so the recession history only uses what was known at the time.
Launch
Negative Beta Tracker: how many stocks move against the index
A new page publishes the share of S&P 500 members whose beta to the index is below zero, over three months and a year, every session since 2010, beside the same measure on 49 US industries since 1926. On October 6, 2026, 25.5% of members had a negative three-month beta, against a typical 1.2% and a high of 17.3% before 2026; the share peaked at 43.1% on August 10.
The industry record, which has no survivorship gap, reached 28.6% in July 2026, above its February 2001 high of 18.4%.
A negative beta means a stock has tended to move against the index from day to day, not that it has fallen. Over the same three months the median negative-beta member returned −3.0% and the median positive-beta member −2.8%.
The four earlier times the industry share first reached 10% (1927, 1999, 2000 and 2024) were followed by a median gain of 17.5% over the next year, against 13.5% for all days. Four cases are too few to call a signal.
Launch
High-Conviction Stocks: what the tracked funds are pressing, buying and cutting
A new page lists the stocks the tracked fund managers hold with conviction, in three sections: large positions still being added to, aggressive buying, and high-conviction trimming. Each stock carries a weekly price chart with every fund's quarter-end share count beneath it, green where the fund added and red where it sold.
A stock's first quarter after it lists is never counted as buying, because it makes a long-held private stake reportable for the first time.
Funds whose latest 13F is more than a quarter behind, and the Gates Foundation trust, are left out and listed on the page.
13F filings cover US-listed long equity only and arrive up to 45 days after the quarter ends.
Correction
Fund Managers: 13F amendments are now applied
The fund pages used only each quarter's original 13F filing and ignored amendments. Managers file an amendment to restate a quarter or to add positions disclosed late, often ones first kept confidential. Every tracked fund has been rebuilt from 2013 with its amendments, which apply to 44 quarters across 14 funds; some restated the original without changing it.
Positions disclosed late now appear in their quarter, such as Berkshire Hathaway's additions in seven quarters and Duquesne's in early 2021.
Restatements replace the original filing. Two corrected visible errors: Trian's December 2022 filing valued Invesco at $1 million instead of $1.0 billion, and Altimeter's March 2022 filing had misaligned rows.
An amendment that repeats positions already filed adds only the new ones.
Launch
PEG Ratio: what the market pays per point of earnings growth, from SEC filings
A new page publishes the PEG ratio of the largest US companies: the trailing P/E from the same filings as Market P/E Ratio, divided by the growth in earnings per share each company has reported over three and five years. It leads with the typical company, because the market-wide figure swings with a few very large, very fast growers.
Growth is reported, not forecast. Most published PEG ratios use analysts' expected growth, which we do not hold, so readings here can differ widely from a brokerage figure.
Companies with losses or falling earnings per share have no PEG; how many are left out is shown beside every reading.
The 40 largest companies are listed with their P/E, growth and PEG on both windows.
Methodology
Momentum Scanner v2: funds excluded, true 52-week highs, and the real factor weights
The Momentum Scanner now ranks stocks only and shows the factor weights the model actually uses. The page had listed weights of 25%, 15%, 15%, 15%, 10%, 10% and 10%; the model has always used 21.1% price performance, 20.4% volume, 16.9% moving-average alignment, 16.3% candles, 11.9% volatility, 10.5% trend consistency and 3% relative strength. The weights now come from the model's own configuration.
ETFs and other funds are excluded. Across the 82 archived sessions a median of three funds left the list, which went from a median of 96.5 names to 93; every stock that remains scores exactly as before.
The 52-week high column is now the 252-session high; it previously showed the 90-session high. Displayed returns use trading-session lengths (1M is 21 sessions, 1Y is 252), so 9M no longer reaches further back than 1Y. The scoring inputs are unchanged.
The seven factor calculations are unchanged. Every archived session was rescored under the new model from today's price files, so the participation history has no break. The weights and the 60 cutoff were set by hand and have not been fitted to returns, and the page now says so.
Launch
Corporate Efficiency Tracker: is AI making US companies more efficient?
A new page tracks overhead and operating margins at the typical large US company since ChatGPT, each company measured against its own Q4 2022 level, beside the profit growth of the companies building AI. It reads "Not visible" until overhead falls at least as fast over four quarters as it typically did before ChatGPT.
Built from the same SEC filings as Corporate Capex and Market P/E Ratio, for companies outside the AI build-out and the software industry, financials excluded.
Overhead uses only companies that report the full selling, general and administrative line, because a general and administrative figure alone leaves out sales and marketing.
The profit chart shows each group's total operating profit, so it locates growth since ChatGPT without attributing it to AI.
Update
Gold Is Down 23% From Its Record: the record since 1971 and the forces behind the decline
The gold Insight of October 3 now carries a dated update. It extends the record from GLD's 2004 launch back to 1971 with the World Bank's monthly gold price, picks every parabolic peak by one rule, and shows what real yields, the dollar, the Fed, ETF and futures investors, central banks and geopolitical risk did around each. The original analysis is unchanged and keeps its date.
Six peaks qualify: the record month of a decline of at least 20% in the monthly average, after a gain of at least 70% over the two years before. On monthly averages the 2026 decline has not yet crossed 20%; on daily closes it has, and the update states both.
Four new pages carry the live data behind it: Gold Price, Real Yields, Gold ETF Flows and Central Bank Gold Buying.
The lean is sharpened, not reversed: further downside first, with sustained ETF selling as the sign of a larger decline.
Correction
Is This Another 2000?: the technology leaders' profits restated on operating income
The Insight comparing today's market with the 2000 top measured the seven largest technology companies' profits with net income alone. In 2026 net income includes large gains on stakes in private companies, most of them at Alphabet and Amazon, so the leaders' earnings growth and profit margin overstated what their businesses earned. Operating-income figures now appear beside the net ones.
Over the year to September 2026 the leaders' net income rose 67% and their operating income 34%; their operating margin went from 30.6% to 33.3%, where the net margin had shown 26.4% to 35.7%.
The finding the article rests on is unchanged: outside the seven, earnings and operating profit both rose about a fifth, and most of those companies reported higher earnings.
Economy-wide BEA profits, which carry the profit-share arithmetic, exclude gains on holdings and were not affected.
Update
New Highs – New Lows: what happened after new lows piled up near a record
The New Highs – New Lows page has a fourth section that tests a widely shared warning: NYSE-listed new lows outnumbering new highs by 100 or more while the S&P 500 sits within 5% of its record. It shows whether today meets the rule, every earlier episode since 2010, and what SPY did over the following month and quarter against any session.
The all-time high comes from SPY's full price history, so years before the breadth record caught up with the 2007 peak are not mistaken for near-record years.
A second version of the rule, with the threshold as a share of NYSE issues rather than a fixed count, is reported beside the first, because the number of listed stocks changes over time.
The NYSE count is built from today's listed stocks, so companies that later delisted are missing and older readings understate new lows; the page says so under the table.
Methodology
Fear & Greed Index v3: flows into cash ETFs join as an eighth component
The Fear & Greed Index now averages eight sentiment families instead of seven. The new one is the 13-week net flow into cash-like ETFs from the new Cash ETF Flows page, inverted so that a rush into cash reads as fear, and scored against its trailing year like every other input.
The flow series is weekly, so each session carries the latest weekly reading; a reading more than ten calendar days old is dropped, as the put/call input is after seven.
The component scores from 2017, when a full year of weekly readings exists. Earlier history still averages the components available, and every past reading was recomputed under the new method, which moves the average day by a few points.
The zone study was rerun before the change shipped. The contrarian gradient held, with the fearful zones still preceding the best three-month returns and Extreme Greed the weakest. The basket audit on the page is now computed from the data rather than written by hand.
Launch
Cash ETF Flows: weekly money moving into T-bill and ultra-short ETFs
A new Sentiment tool tracks net flows into 23 T-bill, floating-rate Treasury and ultra-short bond ETFs, the funds investors use as cash that still pays, and scores the 13-week flow against the same funds over the prior three years.
Flows are measured from share counts, so price moves and reinvested interest do not count as money coming in.
Money market funds are left out: their weekly balances swing with tax dates and quarter-end corporate cash, and the retail figures arrive weeks late.
The page publishes what the S&P 500 did after every extreme reading against the average session, with the sample size stated.
Launch
US Federal Budget: income, spending, the balance sheet and the if-nothing-changes arithmetic
A new Government & Debt page reads the federal budget like a household statement: receipts by source and outlays by function from the Monthly Treasury Statement, the annual totals since 1901, the audited balance sheet and net cost by agency from the Financial Report, and a ten-year projection computed from the last ten years' growth rates.
The budget figures are cash and the balance sheet is accrual; the page shows both, labels each, and never nets one against the other.
The projection holds three trailing ten-year compound growth rates and the Treasury's latest average interest rate constant and adds up the consequences, with the rate moved one point each way. It is labelled arithmetic wherever it appears; the CBO's Long-Term Budget Outlook is cited as the reference forecast and its figures are not copied.
There is no forward-return study, for the reason the national-debt page gives: the levels are reached once, so conditioning on one reports the equity market's own history under a fiscal label.
Launch
Futures Open Interest: total ES and NQ open interest, read with price
A new Sentiment tool sums open interest across every live S&P 500 and Nasdaq-100 futures contract each day since 1997 and reads it with the 20-session price change as new longs, short covering, new shorts or liquidation, with what followed each state.
The continuous futures symbol reports only the front contract's open interest and collapses at every quarterly roll, so the series is built from each quarterly contract and summed, with a rolling average over one contract cycle turning the quarterly expiry hump into a constant offset.
Open interest measures size and never direction; the page says so in its first paragraph and claims nothing about which side of the book the marginal contract sits on.
Notional, contracts times price times the multiplier, is published beside the contract count because the index has more than doubled since 2019 and the micro contract took part of the retail book.
Launch
Stock Map: every US stock by sector, industry and size, coloured by return
A new Stocks tool places about 4,700 US-listed common stocks and ADRs on one grid, rows by sector and industry and columns by market-cap bucket, with each tile coloured by its return over one day to one year and every row carrying the median and the count.
Sector and industry come from the SEC industry code on each company's filings, supplied by the data provider with the SEC filing index as fallback, mapped through a published editorial table to 11 sectors and about 80 industries, with a short list of symbol overrides where the code misleads.
Market cap is the provider's current share count times the latest close, so a split can no longer pair a new price with an old count. The pre-split cap snapshot is not used here.
Names without an industry code from either source sit in an Unclassified row rather than being dropped.
Launch
Market Regime: a two-axis read of how tradable the tape is, folded from six intraday tools
A new Tape tool combines Futures Candle Size, Intraday Momentum, Range Efficiency, Index Sync, Bull vs Bear Power and Session Momentum into two axes. Movement asks how much the market travels in a session; quality asks how cleanly it travels. The pair is read as one of four regimes, Expansion, Whipsaw, Grind or Chop, since 2004.
Each input is ranked against its own history before it is averaged, so series that begin in 1997 and in 2009 sit on the same scale. Direction is deliberately excluded: a bear-swept session and a bull-swept one are equally tradable, and who is winning is the Tape Tracker's question.
The regime is read on 20 sessions and only switches when an axis crosses a threshold, so the label does not flutter. A separate 10-session read names the regime the market is shifting toward when the two disagree.
The page publishes regime durations, the transition table and what each regime meant for the next session's range and efficiency, beside the all-sessions baseline.
Update
Intraday Momentum now covers SPY as well as QQQ
The Intraday Momentum tool adds SPY beside QQQ: the same 50-day and 10-day averages of the open-to-close move, the volatility bands, the weekday table and the spike base rates, on a symbol toggle. QQQ remains the default view.
SPY's record runs from 1993, QQQ's from 1999. Each ETF's percentile is measured against its own history, so the two percentiles are not directly comparable.
Nothing changes for existing readers or links: the page opens on QQQ, and the published dataset keeps its QQQ fields where they were, with SPY added alongside.
Launch
Futures Candle Size: the average ES and NQ 1-minute candle, cash session and overnight separately
A new Tape tool measures the high-to-low range of every 1-minute candle on ES and NQ futures in percent of price, averages it per session, and bands the 20-session average from quiet to extreme against every prior reading. The cash session and the overnight are published as separate series.
Percent is of the prior session's un-adjusted front-month close. The back-adjusted continuous contract the candles come from has a fake historical level, so dividing by it would understate early volatility by several times.
Averages are counted in sessions rather than candles, because a fixed candle count on a 23-hour series would drift through the intraday volatility curve every day.
The record was extended the same day to September 1997 for ES and July 1999 for NQ, the start of the 1-minute history. The bands rank each session against every prior one, so the longer record re-ranked the launch-day readings: ES moved from the 2nd to the 4th percentile and NQ from the 10th to the 22nd.
Correction
Advance–Decline history now uses dated listings and stricter price checks
The A–D charts previously included some funds, assigned historical votes using current exchange labels, and could compare prices across missing sessions. The rebuilt record excludes those votes, retains available delisted histories, and assigns stocks to their dated listing venues.
The corrected record starts November 1, 2016. Earlier Nasdaq listing metadata is too incomplete for a reliable comparison; historical coverage remains incomplete after that date too.
NYSE and Nasdaq now open on a shared percentage scale, with 21- and 63-session advance shares. Their cumulative lines remain available as a second view.
The McClellan and Zweig tools and the Hindenburg Omen’s breadth-momentum input use the rebuilt series. Historical readings, signal lists and forward-outcome studies have been recalculated.
Methodology
Realized Correlation adds the six-month series across the 500 largest stocks and a Nasdaq-100 basket
The page measured average pairwise correlation over 21 sessions across the 100 largest stocks. It now also publishes the 126-session measure across the 500 largest, the series usually quoted as S&P 500 realized correlation, and the same measure for the 100 largest Nasdaq-listed stocks at both windows, each with its own index reference and forward-return table.
Both new baskets are point-in-time, re-picked monthly from trailing data, and start in mid-2010, the earliest a six-month window fits inside our stock history.
The Nasdaq basket is a proxy for the Nasdaq-100 and is labelled as one: the index excludes financials, and exchange is each name's current listing.
The existing 21-session series, its regime bands and its studies are unchanged.
Methodology
Unusual Options Activity scores open-interest growth against each name’s own history
The scanner’s open-interest growth factor was a dollar change compared across the whole universe, so the largest option books led the rankings by size alone. It now measures each symbol’s change against its own recent sessions, the way the implied-volatility factor already did, before the cross-universe comparison.
Rankings change from the September 3 session onward; the archive records which method produced each day’s lists, and the new forward-return study reports the two separately.
Index and sector ETFs remain in the rankings and are now labelled as such rather than removed.
Coverage counts now reconcile: every scanned chain is either ranked or carries a stated exclusion reason, including chains with no out-of-the-money open interest.
Product
No paid tier: every tool is free, and the systematic strategies are private
The Premium tier, its account pages and the checkout are removed. The Unusual Volume, Unusual Options Activity and Momentum scanners are ordinary public tools. The Citrine and Sunstone strategy pages are withdrawn from the site and run privately.
Nothing about the strategies is published any more: no scores, holdings, targets or performance.
Tool and indicator pages carry a clearly labelled placement for systemtrader.co, a separate site by the same author.
No subscriber was affected; the tier had none when it was removed.
Launch
Systematic strategies join The Trading Tools
Jade, Citrine, and Sunstone had public Performance and Methodology pages, with the complete trade record, current portfolio, next-session target and ranked candidates behind the paid tier. Withdrawn on August 31, 2026; see the later entry.
The Strategy Lab compares every published system on the same cost-adjusted basis.
Public methodology explains the thesis, portfolio behavior, testing assumptions, execution timing, and principal risks without exposing the proprietary formula.
The local Trader screen remains an operating tool and is not part of the public product.
Methodology
Published strategy records move after execution costs
Strategy pages now lead with the historical result after modeled execution costs at a $100,000 account. A strong frictionless backtest can no longer qualify a strategy for publication by itself.
The cost estimate responds to order size, liquidity, and volatility; actual fills can still differ.
Every catalog member must have a calibrated cost record and clear the same post-cost publication gate on both the complete history and the holdout period.
Gross research remains in the internal audit record; subscriber-facing performance uses the post-cost result.
Corrections
If a data or implementation error materially changes a published conclusion or performance record, we correct the affected page and document the effect here. Questions about a record can be sent to info@thetrading.tools. Read the full methodology and strategy disclosure.