TrendyVest · a standing record
The Decoy Chronicle
A running log of stale, contradictory or mistaken figures published as current by the financial press — each with its true value, the date it actually belongs to, and how this desk caught it.
1 withdrawn · 1 pending assessment · 10 displayed dates · 7 failure categories. Earlier lifetime records are outside this package.
Evidence and date notes
- The supplied entries name comparison sources, but the original package contains no direct source links. Publication does not independently verify the underlying market claims. A supplied source link is not itself a verification guarantee.
- Entry dates are retained as source display labels and may refer to an issue, session or observation. Entries 124–129 display 21 SEP while describing observations on Tuesday 22 September. No precise observation timestamps have been inferred.
- The supplied lifetime tally includes earlier records not included in this package and is not independently verified. Withdrawn entries remain visible but are excluded from the active count. Pending assessments are identified separately.
- Forecast outcomes and scoring notes are preserved as supplied, including the latest group’s Friday-night scoring reference. Open and unresolved forecasts have not been rescored or assigned inferred outcomes.
The following method and historical notes reproduce the supplied desk record. The evidence limitations above apply throughout.
Correction to this register
The count was published as 94 on 15 September. It is 92. The revised edition of that evening's recap re-counted two entries already logged in its first edition. Tracing the increments: 68 at the start, 79 after the Midday of 14 September, 85 after the Closing Edge, 89 and then 90 across the two editions of the 15 September Brief, 92 after the Recap, 96 after the first Errata. A register that logs other people's arithmetic has to survive its own.
16 September. Entry 94 is withdrawn: a headline giving the day's high beside a table giving the current level is not a contradiction, and Tuesday's intraday peak was 5.04 percent on three feeds, not the 5.02 this desk swore in. The count falls to 95, rises to 96 with entry 97, to 97 with entry 98 at midday, to 98 with entry 99 after the Fed Day piece's fact-check, and to 99 with entry 100 in the Closing Edge. The withdrawn entry stays on the page, numbered and marked, and is not counted.
16 September, evening. The Errata, No. 1, said the Summary of Economic Projections publishes nineteen individual projections. It publishes eighteen: there are nineteen participants and Chair Warsh does not submit one, as he said in June. The forecast built on that count (item 8 of No. 1) is scored on its other two forms. Corrected at full size at the top of No. 2. The count rises from 99 to 104 with entries 101 to 105.
18 September, midday. The Whisper, No. 2, was written against a count of 111, read from this page before the day's later entries were posted, and its sources section carries that figure. The page is right and the issue is behind it: the count stood at 115 when No. 2 went out and rises to 116 with entry 117 below. The issue was corrected before delivery; the discrepancy is recorded here because the register is the authority on its own arithmetic.
17 September. The Whisper, No. 1, said this register stood at 104. It stood at 106: the morning's Brief had logged entries 106 and 107 before The Whisper was written, and the desk carried a stale count into print. Corrected at full size at the top of the Closing Edge, which was built after Thursday's close and published on Friday, after the Friday Brief had added entries 108 to 112; its four entries are therefore numbered 113 to 116 and the count rises from 111 to 115.
22 September, morning. The Errata, No. 3, is Monday's issue and was delivered Tuesday morning: the scheduled Monday-evening run stalled after reading the runbook and resumed at 9:21 a.m. Tuesday, so its observations are stamped Tuesday between 9:21 and 9:57 a.m. It was drafted against a count of 120; the Tuesday Brief logged entries 122 and 123 while it was being built, and its first entry is the same page as entry 122, so it adds six, not seven. The page is right: the count stood at 122 when No. 3 went to press and rises to 128 with entries 124 to 129. No Errata ran on Thursday 17 or Friday 18 September; No. 2's forecast is scored in No. 3, four days late.
22 September, evening. The Errata, No. 3, said Tuesday's two-year auction was $78 billion. It was $69 billion, the same size as every two-year auction this year, per investingLive, Newsquawk and Helious. The figure sat in the forecast register (item 3 of No. 3) and in entry 128; both are corrected in place and the correction runs at full size at the top of No. 4. Separately, No. 3 printed November WTI's Monday settlement at $92.47 from Reuters; Tuesday's changes on Yahoo Finance and Avalon Capital both imply $92.37, and the ten cents is left unresolved until the exchange settlement can be read. No. 4 was delivered late: its run stalled between reads, and its observations are stamped to windows between 7:22 and 9:05 p.m. rather than to minutes. The count rises from 128 to 133 with entries 130 to 134.
23 September, morning. The Brief logged entries 135 and 136 and the count rises from 133 to 135. The Brief also notes that the flash PMI figures printed as consensus in No. 4's forecast (manufacturing 53.9, composite 56) are the August priors on Trading Economics; Investing.com's consensus is 53.6 for manufacturing and 55.8 for services, with none for the composite. Recorded here for tonight's scoring.
24 September, morning. The Brief logged entries 137 to 139 and the count rises from 135 to 138. The Brief was delivered after its 8:15 a.m. target; its observations are stamped between 8:58 and 9:05 a.m. rather than to the earlier window. A second edition, fact-checked between 9:10 and 9:20 a.m., corrects three of the desk's own errors at full size: August new home sales and the Kansas City Fed survey were printed with their Trading Economics priors (607,000 and 10) as consensus, the same failure logged against The Errata, No. 4; Muse was said to have been unveiled at Meta's Connect conference with no fee disclosed, when it was announced on 8 September and carries $20 and $100 monthly tiers, the transaction fee being what is undisclosed; and McDonald's fall was attributed to remarks on traffic without the $8.5 billion franchisee plan the market was pricing. The count is unchanged at 138.
24 September, evening. The Errata, No. 5, scores No. 4 a day late: no Errata ran on Wednesday 23 September. No. 4’s forecast item 3 carried the August flash PMI readings (53.9, 56) as consensus, as the Wednesday Brief recorded above; corrected at full size at the top of No. 5. No. 5 was delivered after its 8:15 p.m. target, with observations stamped between 8:21 and 8:25 p.m. The count rises from 138 to 143 with entries 140 to 144. A fact-check after delivery found that No. 5 had read a Yahoo Finance update's UTC page stamp as Eastern and printed 2:27 p.m. for 10:27 a.m.; corrected in entries 143 and 144, in the issue's second edition, its figure and its thread. The finding is unchanged.
25 September, morning. The Brief logged entries 145 to 147 and the count rises from 143 to 146. The Brief was delivered after its 8:15 a.m. target: the run stalled between reads, its observations are stamped between 7:20 and 7:40 a.m., and the 8:30 durable goods release is previewed in it, not reported.
25 September, midday. The Whisper, No. 8, adds entry 148 and the count rises from 146 to 147. The issue was delivered after its 12:30 p.m. slot, with observations stamped between 11:10 a.m. and 12:20 p.m. Its entry is a headline that restates its own body as a different quantity, which is the failure mode of entries 108 and 143 in a new setting: a forecast threshold published as a forecast.
What qualifies
An entry is a figure published as current that was not. A screen quote differing from an exchange settlement is a difference, not an error, and is not logged. A field that contradicts itself, a component carried forward from a month that has already been superseded, a benchmark quoted below the benchmark it always trades above, a record announced before the record occurred — those are errors, and they are.
Every entry names what was printed, what was true, and how the discrepancy surfaced. Publications are named because the point is not that any one outlet is careless; it is that the supply chain these numbers travel through has no error-correction layer at all, and almost nobody checks.
Corrections and updates to this record
· Entry 111
Removed the unsupported claim that the quoted magnitude was right. The supplied level of 157.871 and increase of 1.2965 imply a prior level of 156.5745 and approximately 0.828 percent, not 1.22 percent. This arithmetic does not independently establish the observed market move.
Previous assessment for entry 111A rising dollar-yen is a weaker yen. The pair rose after the Bank of Japan's increase to 1.25 percent; Invezz had the yen weakening toward 157 at 1:28 a.m. The magnitude was right and the direction was inverted.
· Entry 115
Corrected the equation: 52,064 + 316 = 52,380. Dividing 316 by 0.0060 yields approximately 52,666.67, not 52,380. A displayed percentage may be rounded and does not reconstruct an exact prior close.
Previous assessment for entry 115A quote equal to its own previous close cannot carry a change of 316 points; 316 ÷ 0.0060 implies a base near 52,380, which the page does not show. The range contains neither Wednesday's close of 51,461.90 nor Thursday's of 51,778.04 (Google Finance; +316 points also per Trading Economics). The same page carried Wednesday's close correctly at 4:11 p.m. and an older session eleven minutes later.
· Entry 118
Removed the invalid assertion that identical benchmark prices with unequal percentage changes cannot both be correct. A Monday comparison does not establish Friday settlements. The entry is retained with source verification pending; no contemporaneous same-contract settlement evidence or new outcome has been supplied.
Previous assessment for entry 118Brent trades above WTI, and did all week: this morning the pair was $101.91 against $98.54, a spread of $3.37 on one page at one minute (Trading Economics, 6:40 a.m.). Identical levels with unequal percentage changes cannot both be right in any case, and $100.70 is also the same table's Thursday WTI level.
The register · newest entry numbers first
A linked entry stays visible when a filter is selected. Filters affect only the documented entries.
Showing all 70 documented entries.
-
14825 SEP
Source links not supplied
Entry 148 (25 SEP): threshold printed as forecast
Published claimA technology-news site published a UBS finding under the headline “UBS Warns Silicon Wafer Prices Could Jump 40% by 2027.”
Desk assessmentThe body of the same page says something else: wafer manufacturers “may need price increases of 40% to 50% just to justify building the capacity the AI industry says it needs,” and adds that suppliers “won’t commit to new fabs unless the economics justify it.” That is a threshold required to unlock investment, not a forecast of what prices will do — the page states the distinction explicitly two paragraphs below its own headline. The error matters this week because a real forty-percent figure is circulating beside it: the Commercial Times of Taipei put new 2027 long-term contract prices for polished twelve-inch wafers above forty percent on 23 September, a different quantity from a different source, and the two are already being merged downstream.
-
14725 SEP
Source links not supplied
Entry 147 (25 SEP): wrong sign
Published claimThe same provider's currency page, read at 7:35 a.m. Friday, carried dollar-yen at 157.726 with a change of +1.1380 and a percentage change of −0.72.
Desk assessment1.138 on an implied base of 156.588 is 0.727 percent, so the magnitude is right and one of the two signs is wrong; a field cannot rise by 1.138 and fall by 0.72 percent at once. The same page's dollar index line (100.987, −0.2977, −0.29 percent) carried consistent signs at the same minute. Which sign is wrong is left open; the pair's direction is not printed in the Brief.
-
14625 SEP
Source links not supplied
Entry 146 (25 SEP): quote contradicts narrative
Published claimThe same provider's Japan page, read at 7:30 a.m. Friday, quoted the Nikkei 225 at 66,572 beside a change of +850.21, +1.30 percent, while its own narrative on the same page said the index “rose 1.3% to close at 66,364 on Friday.”
Desk assessmentThe narrative reconciles: 66,364 less 850.21 is 65,513.79, Thursday's close of 65,514 (Business Standard; the same page's Thursday narrative). The quote does not: 66,572 less 850.21 is 65,721.79, a base no feed carried. Quote and narrative, one page, 208 points apart, the same failure as entry 138 on the same page a day earlier.
-
14525 SEP
Source links not supplied
Entry 145 (25 SEP): level contradicts change
Published claimA data provider's United States index page, read at 7:28 a.m. Friday, showed the S&P 500 at 7,722.99 beside a change of −1.90, −0.02 percent, and the Dow at 51,443.06 beside −161.61, −0.31 percent, both dated 25 September.
Desk assessmentThe changes are Thursday's cash session to the cent — the S&P 500 closed at 7,704.13, down 1.90, and the Dow at 51,349.98, down 161.61 (Associated Press 4:18 p.m.; The Motley Fool 5:33 p.m.) — and the levels are Friday morning's contracts. 7,722.99 plus 1.90 implies a Wednesday close of 7,724.89, and 51,443.06 plus 161.61 implies 51,604.67; no feed carried either. The ninth time in eleven sessions on this page (entries 107, 110, 120, 125, 130, 135, 137, 140), and the form The Errata, No. 5, named the night before (item 4).
-
14424 SEP
Source links not supplied
Entry 144 (24 SEP): level contradicts description
Published claimA market live blog, in an update stamped 2:27 on a page clock that reads as UTC (10:27 a.m. Eastern) Thursday, said “the 10-year yield remained at its highest level since 2007, at 5.10%.”
Desk assessmentWednesday closed at 5.11 on the Treasury's Daily Par Yield Curve and touched 5.14 (Investrade; Vantage Markets); Schwab's 9:15 a.m. table had Thursday at 5.12. A level below the previous day's close and high cannot be at its highest since 2007. Either the level or the description was stale; TheStreet had 5.192 at 3:46 p.m., where the description would have held.
-
14324 SEP
Source links not supplied
Entry 143 (24 SEP): wrong maturity
Published claimA market site's morning preview said the “30-year yield hits highest levels since 2007.”
Desk assessmentThe thirty-year's high is the highest since 2004: TheStreet at 3:46 p.m. (5.476 percent, “highest since 2004”), Yahoo Finance's morning update, Schwab's “22-year highs.” Since 2007 is the ten-year's comparison, and the same preview uses it correctly for the ten-year two lines earlier. One record, moved to the wrong maturity.
-
14223 SEP
Source links not supplied
Entry 142 (23 SEP): mislabelled size
Published claimA bond blog's Wednesday piece said “it took a yield of 4.787% to sell all $79 billion of 2-year notes” and gave Wednesday's five-year auction as “$80 billion.”
Desk assessmentThe two-year auction was $69 billion, the same size as every two-year sale this year (investingLive, Newsquawk, Helious); the five-year was $70 billion (investingLive 1:07 p.m. Wednesday; Helious). The yields are right; each size is a round ten billion high. The form The Errata, No. 4, named (item 1), which had feared its own $78 billion would propagate; scored there in a neighbouring form.
-
14124 SEP
Source links not supplied
Entry 141 (24 SEP): wrong series
Published claimA brokerage's 9:15 a.m. note said “weekly initial jobless claims fell to 197,000 from 202,000 the prior week.”
Desk assessmentThe Labor Department's release has the prior week at 198,000, revised up from 196,000, and the decline at 1,000; the 202,000 is the four-week moving average, 202,250, rounded. Kiplinger's close story carries 198,000 to 197,000. Right direction, wrong series, and a decline of 5,000 printed where the document says 1,000.
-
14024 SEP
Source links not supplied
Entry 140 (24 SEP): level contradicts change
Published claimA data provider's United States index page, read at 8:21 p.m. Thursday, showed the S&P 500 at 7,688.28 beside a change of −1.90, −0.02 percent, and the Dow at 51,260.37 beside −161.61, −0.31 percent, both dated 25 September.
Desk assessmentThe changes are Thursday's cash session — the S&P 500 closed at about 7,704, down 0.02 percent, and the Dow at about 51,350, down 0.3 (Kiplinger 4:12 p.m.; 7,706.03 less 1.90 is 7,704.13 and 51,511.59 less 161.61 is 51,349.98) — and the levels are Friday morning's contracts. 7,688.28 plus 1.90 implies a Wednesday close of 7,690.18, which no feed carried. The eighth time in ten sessions on this page (entries 107, 110, 120, 125, 130, 135, 137).
-
13924 SEP
Source links not supplied
Entry 139 (24 SEP): stale resurfacing
Published claimA search for Thursday's Asian session returned Investing.com's “Asian stocks mostly rise as South Korea leads rebound; Nikkei set for weekly loss,” reading as this morning's, with the Kospi up 0.90 percent — the same figure as Wednesday's Seoul close.
Desk assessmentIt is dated 21 August 2026, with the ten-year Treasury at 4.7 percent, the thirty-year at 5.3 and Brent settling at $93.12; Thursday's readings were 5.13, 5.42 and $105.13. Tokyo had been shut for five days and reopened Thursday up 0.76 percent, not the 1.36 in the piece. Correctly dated at source; nothing where it surfaced said so.
-
13824 SEP
Source links not supplied
Entry 138 (24 SEP): quote contradicts narrative
Published claimThe same provider's Japan page, read at about 9:00 a.m. Thursday, showed the Nikkei 225 at 65,385 beside a change of +495.04, +0.76 percent, dated to the day, while its own narrative on the same page said the index “rose 0.76% to close at 65,514 on Thursday.”
Desk assessmentThe narrative is right: 65,514 less 495.04 is 65,018.96, Friday's close on two feeds (Xinhua; TradingKey), and Business Standard carries 65,514 independently. The quote beside it is a contract level: 65,385 less 495.04 is 64,889.96, a base no feed carried. Quote and narrative, one page, 129 points apart. The same page did the same thing on 17 and 18 September (entries 107 and 110).
-
13724 SEP
Source links not supplied
Entry 137 (24 SEP): level contradicts change
Published claimA data provider's United States index page, read at about 8:59 a.m. Thursday, showed the S&P 500 at 7,667.21 beside a change of −58.61, −0.75 percent, and the Dow at 51,365.01 beside −352.10, −0.68 percent, both dated 24 September.
Desk assessmentThe changes are Wednesday's cash session to the cent — the S&P 500 closed at 7,706.03, down 58.61, and the Dow at 51,511.59, down 352.10 (Associated Press 4:12 p.m.; The Motley Fool) — and the levels are Thursday morning's contracts. 7,667.21 plus 58.61 implies a Tuesday close of 7,725.82, and 51,365.01 plus 352.10 implies 51,717.11; no feed carried either. The seventh time in nine sessions on this page (entries 107, 110, 120, 125, 130, 135).
-
13623 SEP
Source links not supplied
Entry 136 (23 SEP): field contradicts field
Published claimA finance portal's Hang Seng index page, read at 7:47 a.m. Wednesday, carried a level of 25,087.75, up 45.04, labelled “at close, 4:08 p.m.” Hong Kong time, beside a previous close of 24,750.80.
Desk assessmentLevel less change is 25,042.71; the previous-close field is 292 points from what the change beside it implies, and no feed carried 24,750.80 for Monday. Investing.com carries 25,087.75 as Tuesday's close and as the previous close for Wednesday's session, which stood at 24,850.00, down 237.75, when read at 7:44 a.m. The level and the change on the portal's page are Tuesday's and agree with a second feed; the field beside them agrees with neither.
-
13523 SEP
Source links not supplied
Entry 135 (23 SEP): level contradicts change
Published claimA data provider's United States index page, read at 7:24 a.m. Wednesday, showed the S&P 500 at 7,768.06 beside a change of −0.06 and the Dow at 51,873.85 beside −185.14, −0.36 percent, both dated 23 September.
Desk assessmentThe changes are Tuesday's cash session to the cent — the S&P 500 closed at 7,764.64, down 0.06, and the Dow at 51,863.69, down 185.14 (Yahoo Finance 4:09 p.m.; TheStreet's percentages agree) — and the levels are Wednesday morning's contracts. 7,768.06 plus 0.06 implies a Monday close of 7,768.12, and 51,873.85 plus 185.14 implies 52,058.99; no feed carried either. The sixth time in eight sessions on this page (entries 107, 110, 120, 125, 130), and the form The Errata, No. 4, named the night before (item 7).
-
13422 SEP
Source links not supplied
Entry 134 (22 SEP): stale resurfacing
Published claimA search for Monday's Chicago Fed National Activity Index returned a release page reading “Actual −0.08, prev −0.02, below prior,” reading as August's print.
Desk assessmentIt is the July release, published 24 August and dated in its own text. August's figure, published Monday, was −0.04 against a July reading revised up to +0.08 (Benzinga; Trading Economics): a decline, not the improvement from −0.08 that The Errata, No. 3, warned of. Correctly dated at source; nothing where it surfaced said so.
-
13322 SEP
Source links not supplied
Entry 133 (22 SEP): stale frame
Published claimA business magazine's piece stamped 1:06 p.m. Tuesday, syndicated by Yahoo Finance, was headlined “The 10-year Treasury yield just hit 5% for the first time since 2007.”
Desk assessmentThe ten-year hit five percent on 14 September and cleared the 2007 comparison on the fifteenth (entries 79, 80); it first closed above five on the Treasury's par curve on the sixteenth (entry 104). On Tuesday it sat at 4.96 percent all afternoon on two feeds. The body says “this month,” which is true; the headline says “just,” a week later.
-
13222 SEP
Source links not supplied
Entry 132 (22 SEP): propagated error
Published claimThe same 5:00 p.m. card said the ten-year was “easing to 4.96% after touching 5%,” down five basis points.
Desk assessmentThe ten-year closed at 4.96 percent Tuesday (Avalon Capital; BBN Times) and at 4.95 or 4.96 on Monday on four feeds: the day's move was zero to one basis point. Five basis points is the distance from 5.01, the level the same outlet's Monday card gave as a close, up seven, on a day the yield fell (entry 124). Monday's error became Tuesday's base.
-
13122 SEP
Source links not supplied
Entry 131 (22 SEP): mislabelled
Published claimA market site's card stamped 5:00 p.m. Tuesday gave the S&P 500 “at close” at 7,776.60 (+0.03%), the Dow at 51,947.30 (+0.01%), the Nasdaq 100 at 30,774.80 (+0.08%) and the Russell 2000 at 2,895.43 (−0.01%).
Desk assessmentNone is a close. The Dow fell 0.36 percent to 51,863.69 on three feeds; the Russell 2000 rose, 0.55 percent to 2,891.13 (Yahoo Finance) or 0.51 to 2,889.92 (Avalon); the S&P 500 closed at 7,764.64, down six cents. Two of four indices with the wrong direction, under a label that says the opposite of what the numbers are.
-
13022 SEP
Source links not supplied
Entry 130 (22 SEP): level contradicts change
Published claimA data provider's United States index page, read Tuesday evening, showed the S&P 500 at 7,765.34 beside a change of −0.06 and the Dow at 51,885.76 beside −185.14, −0.36 percent, both dated 23 September; its search snippet advertised “7,767 points” and a “0.03% gain.”
Desk assessmentThe changes are Tuesday's cash session to the cent — the S&P 500 closed at 7,764.64, down 0.06, and the Dow at 51,863.69, down 185.14 (Yahoo Finance 4:09 p.m.; Avalon Capital) — and the levels are Wednesday's contracts. 7,765.34 plus 0.06 implies a Monday close of 7,765.40, which no feed carried; Monday's was 7,764.70. The fifth time in seven sessions on this page (entries 107, 110, 120, 125).
-
12921 SEP
Source links not supplied
Entry 129 (21 SEP): stale resurfacing
Published claimSearches for Chicago Fed President Goolsbee's Monday remarks in London returned “Fed's Goolsbee: I want more proof of inflation easing to end rate hikes,” which in a hiking cycle reads as this week's.
Desk assessmentThe URL dates it 1 August 2023. Monday's remarks, per the Associated Press at 11:12 a.m. and Investing.com at 9:10 a.m., were that the rate response to demand-driven inflation “is more aggressive and more front-loaded” and that bringing inflation down “is going to be painful.” Correctly dated at source; nothing where it surfaced said so.
-
12821 SEP
Source links not supplied
Entry 128 (21 SEP): stale resurfacing
Published claimA search for Tuesday 22 September's calendar returned “Building permits, new home sales, and consumer confidence due Tuesday,” reading as this week's.
Desk assessmentIt is Investing.com's preview for Tuesday, 25 August, published 24 August and dated as such in its own text, with consensus figures (1.443 million permits, 620,000 new home sales, confidence 90.3) superseded a month ago. None of the three is due this Tuesday; the calendar has the Richmond Fed survey at 10:00, a $69 billion two-year auction at 1:00 (printed as $78 billion until 22 September; see the audit block) and three Federal Reserve speakers. Correctly dated at source; nothing where it surfaced said so.
-
12721 SEP
Source links not supplied
Entry 127 (21 SEP): morning's move in a close story
Published claimA closing digest for Monday, timed 21:04, said “Meta Platforms, which jumped 7.09% following a Wells Fargo price target hike to $796,” beside index closes that were right to the cent.
Desk assessmentMeta closed up 11.34 percent per TheStreet, 11.4 per Kiplinger and Reuters. The seven is the morning's move: Yahoo Finance's live blog had Meta “jumped nearly 7%” in an update stamped 10:09 a.m. Eastern. A ten o'clock number in a five o'clock digest, on the stock that carried the day.
-
12621 SEP
Source links not supplied
Entry 126 (21 SEP): a dip labelled a session
Published claimA Nasdaq close story said “Brent dropped below $100 a barrel for a fourth straight session,” beside a WTI decline of “about 4% to 5%.”
Desk assessmentBrent's November contract settled at $100.34, down $3.53, per Reuters at 3:37 p.m., which called it the lowest settlement since 9 September; it dipped under $100 during the day and came back. A settlement that is the lowest since 9 September and still above $100 means every settlement in between was above $100 too: there has been no session below $100 at the close, let alone four. TheStreet's 4:00 p.m. table had Brent at $101.30 beside WTI at $97.79; WTI's October contract settled at $95.78 (Reuters; Kiplinger).
-
12521 SEP
Source links not supplied
Entry 125 (21 SEP): level contradicts change
Published claimA data provider's United States index page, read at 9:24 a.m. Tuesday, showed the S&P 500 at 7,771.88 beside a change of +114.20, +1.49 percent, and the Dow at 52,386.72 beside +366.19, +0.71 percent, both dated to the day.
Desk assessmentThe changes are Monday's cash session to the cent; the levels are Tuesday's pre-market contracts. 7,771.88 less 114.20 implies a prior close of 7,657.68, and 52,386.72 less 366.19 implies 52,020.53; no feed carried either. Monday's closes were 7,764.70 and 52,048.83. The fourth time in six sessions this page has paired one session's change with another session's level (entries 107, 110 and 120).
-
12421 SEP
Source links not supplied
Entry 124 (21 SEP): wrong direction
Published claimA closing-bell card said “the 10-year just closed above 5% for the first time in this cycle,” and carried the yield at “5.01% +7 bp” under an 8:00 p.m. stamp; the same card had the S&P 500 “closed at 7,768.43.”
Desk assessmentThe ten-year fell on Monday: 4.951 percent from 5.00 per TheStreet's closing table, 4.95 per Stock Market Watch, 4.96 per The Motley Fool, “fell below 5%” per Reuters. And the first close above five in this cycle was Wednesday, 16 September, at 5.01 on the Treasury's par curve (entry 104); Wolf Street has 5.01 again on Friday. Wrong direction on the day, and a record five days stale. The S&P 500's settled close was 7,764.70 on seven feeds.
-
12322 SEP
Source links not supplied
Entry 123 (22 SEP): misquoted range
Published claimA venture blog's post, headlined “Meta AI Capex 2026: Guidance Narrows to $135-145B After Q2,” carried that range as the company's current capital expenditure guidance.
Desk assessmentExhibit 99.1 to Meta's Form 8-K of 29 July 2026 gives $130 to $145 billion, narrowed from $125 to $145 billion. The low end is five billion dollars off in a headline about the range, on the day Meta's market value moved by more than the whole of it.
-
12222 SEP
Source links not supplied
Entry 122 (22 SEP): wrong session
Published claimA market site's close story headlined “Stock Market Today, September 21, 2026: S&P 500 Closes Mixed as Rates Stabilize,” timed 21:16 UTC, carried the S&P 500 at 5,847.92, the Nasdaq Composite at 18,234.16, the Dow at 42,156.39, the ten-year at 3.94 percent and WTI at $71.18.
Desk assessmentMonday's closes were 7,764.70, 27,122.09 and 52,048.83 (Associated Press 4:34 p.m.; Motley Fool 5:02 p.m.), the ten-year 4.95 or 4.96 percent and WTI above $92. Every figure on the page belongs to a market roughly two years older than the one it is dated to; the S&P 500 level alone is 1,900 points off. The page discloses that it uses AI systems to draft its analysis.
-
12121 SEP
Source links not supplied
Entry 121 (21 SEP): wrong direction
Published claimThe same closing table's opening line told readers that Friday's mixed session came “just two day after the Federal Reserve cut rates.”
Desk assessmentThe Federal Open Market Committee raised its target range by a quarter point on 16 September, to 3.75–4.00 percent — its first increase since July 2023, reported as such all week by this desk and by every major outlet. A policy decision printed in the opposite direction, two days after it happened, at the top of the page carrying the closing levels. Three errors on one page this morning, with entries 118 and 119.
-
12021 SEP
Source links not supplied
Entry 120 (21 SEP): level contradicts change
Published claimA data provider's United States index page, read at 6:31 a.m. Monday, showed the S&P 500 at 7,693.29 beside a change of +12.74, +0.17 percent, and the Dow at 51,986.13 beside −95.39, −0.18 percent, both dated to the day.
Desk assessmentThose are Friday's cash changes to the cent — the S&P 500 closed at 7,650.50, up 12.74, and the Dow at 51,682.64, down 95.40 (Yahoo Finance; TheStreet) — printed beside Monday morning's contract levels. 7,693.29 less 12.74 implies a prior close of 7,680.55, which no feed carried. The third time in five sessions this page has paired one session's change with another session's level (entries 107 and 110).
-
11921 SEP
Source links not supplied
Entry 119 (21 SEP): mismatched level and change
Published claimThe same closing table carried the Russell 2000 at 2,876.85, down 0.50 percent, for Friday's session.
Desk assessmentFriday's close was 2,860.40, down 14.23, on Yahoo Finance at 4:30:13 p.m., against a prior close of 2,874.63. The 2,876.85 is the same table's own Thursday level, printed there beside a gain of 0.63 percent. Thursday's level, Friday's change, second week running on the same page.
-
11821 SEP
Source verification pending
Entry 118 (21 SEP): impossible relationship
Published claimA market site's Friday closing table, read at 6:28 a.m. Monday, carried West Texas Intermediate at “$100.70, −1.19%” and Brent at “$100.70, −1.98%” — one level for two benchmarks, with two different changes.
Desk assessmentDifferent benchmarks can reach the same price from different previous prices and therefore have different percentage changes. The supplied Monday comparison ($101.91 Brent against $98.54 WTI at 6:40 a.m., attributed to Trading Economics) does not establish the quoted Friday settlements. The supplied text also notes that $100.70 is the same table's Thursday WTI level. Whether the Friday table was wrong requires contemporaneous, same-contract settlement evidence. Source verification pending.
source verification pending
This entry is retained with its assessment pending contemporaneous source evidence. It is not presented as an independently verified error.
-
11718 SEP
Source links not supplied
Entry 117 (18 SEP): stale resurfacing
Published claimA search for this week's unusual options activity in a large-cap semiconductor name returned an options-flow scan reading as current: thirty-two trades, calls outnumbering puts twenty-three to nine, strikes “from $140.0 to $280.0 over the last three months,” and the stock quoted at $215.89, up 0.26 percent.
Desk assessmentThe stock traded above $543 on Friday morning and closed at 545.09 on Thursday (stockanalysis.com; Investing.com). The scan's strike ladder and its expiries — January, February and May 2026 — place it in late 2025, and its own text dates it to thirty-four days before an earnings date it does not name. Correctly dated at source; nothing where it surfaced said so, and an options scan carries no headline date to warn a reader.
-
11617 SEP
Source links not supplied
Entry 116 (17 SEP): narrative vs. table
Published claimA data provider's index page at 4:27 p.m. said in its narrative that “the S&P 500 rose 1.1%, the Dow added 316 points, and the Nasdaq gained 1.7%,” and in its table that the S&P 500 rose 90.69 points or 1.20 percent and the Dow 335.22 points.
Desk assessmentThe narrative matches the index closes (S&P 500 +1.14 percent, Dow +316.14, Nasdaq Composite +1.69, per Google Finance); the table is the site's own contract, quoted against a different base, and the page presents both under one heading. The same page and the same failure as entries 95 and 100, in the other direction: this time the prose is right and the table is not.
-
11517 SEP
Source links not supplied
Entry 115 (17 SEP): level contradicts change
Published claimA data site's Dow Jones page, read at 4:22, 4:31 and 4:48 p.m. Thursday, showed a quote of 52,064 beside a change of −316 (−0.60%), a previous close of 52,064, a day's range of 51,963–52,292 and the label “Closed·15:59:59.”
Desk assessmentA quote equal to its own previous close cannot carry a change of 316 points; 52,064 + 316 implies a previous close of 52,380, which the page does not show. The range contains neither Wednesday's close of 51,461.90 nor Thursday's of 51,778.04 (Google Finance; +316 points also per Trading Economics). The same page carried Wednesday's close correctly at 4:11 p.m. and an older session eleven minutes later.
-
11417 SEP
Source links not supplied
Entry 114 (17 SEP): mislabelled week
Published claimA market-news site's claims story, stamped 12:34 GMT, dated continuing claims of 1.730 million to “the week ending August 5.”
Desk assessmentThe Labor Department's release dates insured unemployment of 1,730,000 to the week ending September 5. Right number, wrong month. Forecast in The Errata, No. 2 (item 3, in its “claims week mislabelled” form); scored there.
-
11317 SEP
Source links not supplied
Entry 113 (17 SEP): impossible relationship
Published claimA market live blog, in an update stamped 10:25 a.m. Thursday, printed: “The S&P 500 climbed 1.19%, the Dow Jones Industrial Average gained 8.81%, and the Nasdaq slipped 0.01%. The Russell 2000 dipped 0.40%.”
Desk assessmentA Dow gain of 8.81 percent would be about 4,500 points; the Dow was up 0.47 percent at 9:45 on Yahoo Finance's board and closed up 0.61, or 316 points, on Google Finance and Trading Economics. A Nasdaq slipping 0.01 percent beside an S&P 500 up 1.19 describes no session this week; the Nasdaq Composite closed up 1.69 percent on two feeds. Two impossible numbers in one sentence, the second of which is Wednesday's Nasdaq close.
-
11218 SEP
Source links not supplied
Entry 112 (18 SEP): stale resurfacing
Published claimA search for Friday's Tokyo close after the Bank of Japan's decision returned “Japanese Shares Trim Gains on Hawkish BOJ,” reading as this morning's session.
Desk assessmentA note dated 26 February 2026, with the Nikkei at 58,753 and the Topix at 3,880, seven thousand points below Friday's close. Correctly dated at source; nothing where it surfaced said so.
-
11118 SEP
Source links not supplied
Entry 111 (18 SEP): wrong direction
Published claimThe same provider's currency page at 7:38 a.m. said “the yen strengthened notably, climbing over 1.2%” beside a dollar-yen quote of 157.871, up 1.2965, +1.22 percent.
Desk assessmentA rising dollar-yen is a weaker yen. The pair rose after the Bank of Japan's increase to 1.25 percent; Invezz had the yen weakening toward 157 at 1:28 a.m. The quoted level and increase imply a prior level of 156.5745 and a gain of approximately 0.828 percent, not 1.22 percent. The quoted percentage is internally inconsistent; the actual observed move still requires contemporaneous source verification.
-
11018 SEP
Source links not supplied
Entry 110 (18 SEP): level contradicts change
Published claimA data provider's Japan page, read at 7:25 a.m. Friday, showed the Nikkei 225 at 65,635 beside a change of +882.70, +1.38 percent, dated to the day.
Desk assessmentThe change is Friday's cash session to the yen; the level is not. The Nikkei closed at 65,018.95 (Xinhua; TradingKey 65,018.73), and 65,635 − 882.70 is 64,752.30, a prior close no feed carried; Thursday's close was 64,136.25. The same page did the same thing on Thursday morning (entry 107).
-
10918 SEP
Source links not supplied
Entry 109 (18 SEP): component as total
Published claimA market site put Generac's “data center backlog” before the Amazon deal at “approximately $1.35 billion.”
Desk assessmentThe backlog was approximately $1.6 billion on 29 July per the company's second-quarter release, excluding the second hyperscale customer; $1.35 billion is the portion of it scheduled for 2027 delivery, per the chief executive on the call. A part of the number printed as the whole.
-
10818 SEP
Source links not supplied
Entry 108 (18 SEP): headline vs. body
Published claimA crypto news wire headlined Generac's Amazon agreement “Generac Signs $24 Billion Amazon Data Center Generator Deal,” timed 4:24 a.m. on its own clock.
Desk assessmentIts own body gave $2.4 billion of initial deliveries across 2027 and 2028 and an $8 billion cap on warrant vesting, which is what Generac's 8-K of 16 September says as summarised by TipRanks and StockInvest. Headline and body, one page, a factor of ten apart.
-
10717 SEP
Source links not supplied
Entry 107 (17 SEP): level contradicts change
Published claimA data provider's United States index page, read at 7:24 a.m. Thursday, showed the S&P 500 at 7,618.82 beside a change of −33.93, −0.45 percent, and the Dow at 51,889.06 beside −631.21, −1.21 percent, dated to the day. Its Japan page showed the Nikkei at 64,700 beside +213.25, +0.33 percent.
Desk assessmentThe changes are Wednesday's cash session to the cent; the levels are Thursday's pre-market quotes. 7,618.82 less 33.93 implies a prior close of 7,652.75, which no feed carried, and 33.93 ÷ 7,652.75 is 0.443 percent, which does not round to the 0.45 printed. On the Japan page, 64,700 less Wednesday's close of 63,922.78 is +777, not +213.25; the cash close was 64,136. One failure mode, two pages, one morning.
-
10617 SEP
Source links not supplied
Entry 106 (17 SEP): mislabelled pair
Published claimA morning preview for Thursday carried “Housing Starts: Est. 1.41M (prior 1.433M)” and “Building Permits: Est. 1.309M (prior 1.2339M).”
Desk assessmentThose are each other's figures. Trading Economics and Investing.com both have August starts at 1.31 million from 1.239 and permits at 1.41 million from 1.433. Starts up and permits down became starts down and permits up. The exact form The Errata, No. 2 named the night before; scored there.
-
10516 SEP
Source links not supplied
Entry 105 (16 SEP): median called an average
Published claimA brokerage's decision note said “the average projection for rates at the end of this year rose to 4.1%.”
Desk assessmentThe Fed publishes a median, and 4.1 is the median. The average of the eighteen published projections — twelve at 4.125, four at 4.375, two at 3.875, per Yahoo Finance's account of the table — is 4.15, which rounds to 4.2. Right number, wrong word. Forecast in The Errata, No. 1 (item 8, in one of its three forms).
-
10416 SEP
Source links not supplied
Entry 104 (16 SEP): two o'clock's tape in a close story
Published claimA close story with the right S&P 500 close (7,551.81) said “two-year yields eased roughly one basis point, while 10-year and 30-year yields dropped four and five basis points, respectively.”
Desk assessmentThose were the moves at 2:05 p.m. At the close the Treasury's par curve had the two-year at 4.74, up seven basis points, the ten-year at 5.01, up one, and the thirty-year at 5.35, down one. Wrong sign on two of three, and the ten-year's first close above five percent on the government's own series went unreported in a story about the close. Forecast in The Errata, No. 1 (item 5).
-
10316 SEP
Source links not supplied
Entry 103 (16 SEP): page never turned over
Published claimA data site's S&P 500 page showed “7,585.68, −34.30 (−0.45%), Closed 15:59:59” as the latest close at 8:57 p.m. on Wednesday, with no 16 September row in its history table.
Desk assessmentThat is Tuesday's close, five hours after Wednesday's, which was 7,551.81, down 33.92 (BBN Times; Kiplinger). The same site's Dow page, read a minute earlier, had rolled to Wednesday and carried the right close, 51,461.90.
-
10216 SEP
Source links not supplied
Entry 102 (16 SEP): table dated a day late
Published claimA finance site's S&P 500 history table, read at 8:58 p.m., carried a row dated 16 September with an open of 7,612.30, a high of 7,617.26, a low of 7,572.69 and a close of 7,585.73; the row dated 15 September carried a close of 7,619.98.
Desk assessmentThose are Tuesday's and Monday's sessions. Investing.com's dated table carries the identical figures, to the cent, for 15 and 14 September, and the same page's own header said “At close: September 15” beside 7,585.73. Every row is one day late. Wednesday's close was 7,551.81.
-
10116 SEP
Source links not supplied
Entry 101 (16 SEP): headline vs. board
Published claimA market live blog was headlined “Dow sinks 600 points” at 4:50 p.m. while its own quote board, on the same page, showed the Dow at 51,363.01, down 730.10 — and still showed it at 8:55 p.m. The same board had the S&P 500 at 7,527.80, down 57.93, and the Nasdaq Composite at 25,875.21, down 106.36.
Desk assessmentThe Dow closed at 51,461.90, down 631.21 (Investing.com; Kiplinger 51,461, down 631; Trading Economics' point change agreeing). The S&P 500 closed at 7,551.81, down 33.92 (BBN Times; Kiplinger 7,551). The Nasdaq Composite closed at 25,978.42, nearly flat (BBN Times; Kiplinger). Three indices, misses of 24, 99 and 103 points; a headline 130 Dow points from its own board. The board's implied base, 52,093.11, was Tuesday's close: right base, wrong level.
-
10016 SEP
Source links not supplied
Entry 100 (16 SEP): narrative vs. table
Published claimA data provider's index page, an hour after the Fed-day close, said in its narrative that the S&P 500 “fell 1%,” the Dow “1.5%” and the Nasdaq 100 “0.5%.”
Desk assessmentThe table on the same page, at the same minute, showed the S&P 500 at 7,554.59, down 0.41 percent, the Dow at 51,488.15, down 1.15, and the Nasdaq 100 up 0.10. Prose and table, one page, in opposite directions on one index. The second time this week the same page has disagreed with itself about the Dow (entry 95).
-
9916 SEP
Source links not supplied
Entry 99 (16 SEP): stale under a fresh date
Published claimA market-news page dated 8 September put the odds of a September hike at “nearly 56%” after Chair Warsh's Jackson Hole remarks, with June's unemployment rate and June's PCE as the latest data.
Desk assessmentThe speech was on 28 August; the July jobs report of 7 August and the August report of 4 September had both been published by the page's date, and the odds were 66 percent on 31 August and 70 on 10 September. The page carries a late-August reading under a September date. It reached the first edition of this desk's Fed Day piece, which placed the 56 on 8 September; corrected at full size.
-
9816 SEP
Source links not supplied
Entry 98 (16 SEP): stale resurfacing
Published claimSearches at midday for what the market has priced for the September 2026 meeting returned “Fed Traders Now Fully Pricing In Seven Standard Hikes for 2022” and “Traders Cast Doubt on More Fed Hikes as Even December Odds Ease,” both reading as current.
Desk assessmentThey are Bloomberg pieces from March 2022 and December 2018, dated in their own URLs. What is priced for September 2026 is a quarter point at 92.7 percent and about 100 basis points over the following year, per Yahoo Finance and Bank of America via Benzinga.
-
9716 SEP
Source links not supplied
Entry 97 (16 SEP): stale resurfacing
Published claimSearches for a September 2026 Fed preview on the morning of the decision returned “Fed preview: forecasts from 15 major banks, a pause but the end of rate hikes” and “Fed September preview: terminal rate projection is key,” both reading as current.
Desk assessmentThey are FXStreet's previews of the September 2023 and September 2022 meetings, dated in their own URLs. The 2026 meeting is a hike, not a pause, and the committee's target range is 3-1/2 to 3-3/4 percent per the July minutes.
-
9615 SEP
Source links not supplied
Entry 96 (15 SEP): wrong sign
Published claimA secondary economic calendar published the August retail sales consensus as minus 0.9 percent.
Desk assessmentThe consensus is plus 0.9 percent, per Trading Economics, against a July reading of minus 0.6. The inverted sign reached an earlier edition of this desk's 15 September recap, where it supported a section that was withdrawn and rewritten at full size. Logged against the calendar; the failure to check it before building on it was this desk's.
-
9515 SEP
Source links not supplied
Entry 95 (15 SEP): narrative vs. table
Published claimA data provider's post-close narrative said the Dow “lost 1%.”
Desk assessmentThe table on the same page showed the Dow at 52,059.96, down 361.24 points, 0.69 percent. Prose and table on one page, a third of a point apart.
-
9415 SEP
Withdrawn — retained in the record, not counted
Entry 94 (15 SEP): this desk's error
Published claimWithdrawn 16 Sep. The same note's headline said the ten-year Treasury yield had “hit 5.04%,” and its table said 5.00; this register logged that as a contradiction.
Desk assessmentIt was not one. Kiplinger's live blog carries Tuesday's intraday high at 5.041 percent and Saxo's Asia note carries 5.04: the headline reported the day's high and the table the level at that moment. Two true things. The 5.02 this desk had sworn in as the peak was the level Bloomberg saw at 1:20 a.m. and TheStreet at 9:35 a.m. The entry stays numbered and is not counted.
-
9315 SEP
Source links not supplied
Entry 93 (15 SEP): level contradicts change
Published claimA brokerage's 9:12 a.m. market table carried the S&P 500 at 7,619.98, up 0.48 percent.
Desk assessmentMonday's close was 7,621.54. A level below the prior close cannot be a gain. For +0.48 percent to hold, Monday would have had to close at 7,583.58, a figure no feed carried.
-
9215 SEP
Source links not supplied
Entry 92 (15 SEP): mismatched level and change
Published claimA live blog showed the S&P 500 at 7,619, down about half a percent, an hour after the close.
Desk assessment7,619 is within three points of the previous day's close. Monday's level beside Tuesday's percentage change. The magnitude was right and the level was a day old.
-
9115 SEP
Source links not supplied
Entry 91 (15 SEP): wrong session
Published claimA market page carried a session summary dated to the day, describing index moves, single stocks and Federal Reserve expectations.
Desk assessmentIt describes an entirely different session: job openings at 7.67 million “in October,” an expected rate cut rather than a hike, JPMorgan's 2026 expense guidance and a Warner Bros bid. None of it happened that day.
-
9015 SEP
Source links not supplied
Entry 90 (15 SEP): stale resurfacing
Published claimA search for the September 2026 Empire State survey returned a report headlined “index jumps to 19 in September vs 10 expected.”
Desk assessmentThe September 2023 survey — the same calendar date, three years apart, which is precisely why it surfaced and precisely why it reads as today's.
-
8915 SEP
Source links not supplied
Entry 89 (15 SEP): propagated error
Published claimA major network carried the “19-year high” description of the previous day's 5.01 percent print.
Desk assessmentEntry 80's error, now propagating. The figure had not cleared October 2023's 5.02 when the description ran. It cleared it later the same day, which is not the same thing as having been right.
-
8815 SEP
Source links not supplied
Entry 88 (15 SEP): impossible relationship
Published claimA live blog quoted Brent crude at about $102 while quoting West Texas Intermediate at $103.55, at the same minute.
Desk assessmentThe benchmarks are inverted. Brent trades above WTI and did that morning — a separate feed had Brent at $107.90 against WTI at $104.10. An oil price without its companion benchmark hides this; printed as a pair, it cannot.
-
8715 SEP
Source links not supplied
Entry 87 (15 SEP): stale resurfacing
Published claimSearches for the September 2026 Empire State print returned two analyses with headline numbers reading as current.
Desk assessmentOne was the September 2025 survey. One was May 2026. Both correctly dated at source; neither carried a warning where it surfaced.
-
8615 SEP
Source links not supplied
Entry 86 (15 SEP): carried-forward components
Published claimA provider published a September Empire State manufacturing headline of 7.6 with its component detail.
Desk assessmentThe September components — new orders 17.3, shipments 11.7 — are identical to the values the New York Fed published for August, which this desk retrieved and read directly. Two other calendars showed no September actual at all. The whole row is unusable.
-
8514 SEP
Source links not supplied
Entry 85 (14 SEP): internal contradiction
Published claimA provider published two different closing levels for the same index, on the same page, at the same moment.
Desk assessmentThe quote widget read 7,619.90, down 0.48 percent. The prose beneath it read 7,621, down 0.46. One page, two closes.
-
8414 SEP
Source links not supplied
Entry 84 (14 SEP): stale resurfacing
Published claimA search for the day's enterprise software moves returned an article reporting Salesforce up 7 percent, ServiceNow 8, Workday 10.
Desk assessmentDated 27 July 2026 — seven weeks old — and carrying S&P futures at 7,824.75, a level nowhere near that day's. Nothing on the page marked it as historic.
-
8314 SEP
Source links not supplied
Entry 83 (14 SEP): mislabelled
Published claimA market widget carried index levels under the label “at close,” on a piece published at 10:11 a.m. Eastern.
Desk assessmentSix hours before any close. The level shown was live and moving. The words on the box were not describing the number in it.
-
8214 SEP
Source links not supplied
Entry 82 (14 SEP): wrong series
Published claimThe same wire described core inflation as running at 3.4 percent, above the Federal Reserve's 2 percent target.
Desk assessment3.4 percent is the headline CPI rate. Core CPI was 2.4 percent — a five-year low — and core PCE 3.3. Naming the wrong series inverts what the number argues for.
-
8114 SEP
Source links not supplied
Entry 81 (14 SEP): internal contradiction
Published claimA wire wrote that the yield “hadn't reached 5% since 2007 before last touching it in July 2023.”
Desk assessmentThe sentence contradicts itself — both cannot be the last time — and the 2023 touch was October, not July. Two errors in one clause.
-
8014 SEP
Source links not supplied
Entry 80 (14 SEP): contradicted by prior peak
Published claimA data provider described the day's 5.01 percent intraday high as a “19-year high.”
Desk assessmentOctober 2023 printed 5.02 percent. A reading of 5.01 does not clear it, so the 19-year framing was unavailable that day. The same provider's sister framing — a 52-week high — was correct.
-
7914 SEP
Source links not supplied
Entry 79 (14 SEP): stale comparison
Published claimA widely circulated market update described the ten-year Treasury yield as approaching five percent “for the first time since 2007.”
Desk assessmentThree independent wires put the last five percent print in October 2023, less than three years earlier. The 2007 comparison belongs to a different episode. Twenty-four hours later the claim became true — the yield reached 5.02 percent on 15 September and took the 2007 record properly. It was still wrong when it was published.
The forecast register
The forecast register
Each issue of The Errata ends by naming the errors it expects in the next day's coverage. They are listed here the night they are made and scored the following night against what was actually printed — occurred, did not occur, or unresolved — as reporting, not as a grade. A forecast that occurred also becomes a numbered entry above.
All 40 supplied forecast records are shown. Outcomes remain the desk’s reported observations; open or unresolved records are not inferred successes or failures.
The Errata, No. 1 · made Tue 15 Sep for Wed 16 Sep, Fed day
scored Wednesday night: four occurred, four did not, one unresolved
- did not occur
A "second monthly decline" in retail sales that is not in the data: the August consensus is +0.9%, after −0.6% in July; watch also for a mechanical Prime Day rebound reported as clean strength.
Census: August +1.2% (±0.4), July revised to −0.5%. The print removed the premise.
- did not occur
"The last meeting before the midterms." The FOMC meets again 27–28 October; the election is 3 November.
Searched Wednesday’s coverage; not found.
- occurred
A sixth published value for the CME FedWatch hike probability, after 85, 88, 90, "over 92" and 92, with no outlet saying when it looked.
Nine values by 11:19 a.m.: Yahoo, Kiplinger, Investing.com, Benzinga, FXStreet, UPI, Schwab (two), Kiplinger’s weekend note. Documented in the noon piece.
- did not occur
"First hike since" with the wrong year: it would be the first since July 2023. Watch for "since 2022," "in two years," and "first of the Warsh era" presented as the same claim.
“Since 2023” or “three years” at CNBC, Fox Business, CNN, Yahoo Finance and Schwab; “since 2022” not found.
- occurred
The ten-year error, flipped: "highest since October 2023" is now the stale claim; and a close above 5% reported when the yield only touched it, or under-reported if it is real.
Under-reported: BBN Times’ close story carried the 2:05 yield moves on a day the Treasury’s curve closed the ten-year at 5.01, up one. Entry 104. “Highest since October 2023” not found.
- unresolved
A live blog showing the wrong index level beside the right percentage change between 2:00 and 3:00 p.m., in either direction.
The live blogs served caches between 2:00 and 3:00 (Yahoo’s page 2:38 all afternoon). The same failure occurred after the close on Yahoo’s board: entry 101.
- occurred
"The Dow fell X points" with no base level, inheriting Tuesday's thirty-two-point disagreement between feeds.
Four values: 600 (Yahoo’s and CNBC’s headlines), 631 (Kiplinger, Investing.com), 700 (CNBC as indexed at 4:50), 730 (Yahoo’s board). Reconciled in No. 2 at 631.21 to 51,461.90; base 52,093.11.
- occurred, one of three forms
The dots misread three ways: a median of nineteen reported as "the Fed sees"; the 2026 and 2027 medians swapped; projection materials posted at 2:00 quoted as said at 2:30.
“The average projection” for a median: Schwab, entry 105. Swapped medians and 2:00 materials quoted as 2:30 remarks: not found. The forecast’s own count (nineteen) was wrong; it is eighteen — see the audit block.
- did not occur
Deutsche Bank's conditional — a hold would be the biggest dovish surprise at a scheduled meeting since 1994 — quoted without the "if," in either direction.
Searched; not found. The hike made the conditional moot by 2:01.
The Errata, No. 2 · made Wed 16 Sep for Thu 17 Sep, claims, starts and the Philly Fed at 8:30
scored in No. 3, four days late: two occurred, four did not, one unresolved
- did not occur
The fed funds rate quoted as a single number that is neither bound of 3.75–4.00: 3.88 (the midpoint), “4%” alone, or 3.63 (Wednesday's effective rate carried into Thursday).
Searched Thursday’s coverage; not found. The Associated Press printed “approximately 3.9 percent” on Monday 21 September, four days on. Reported, not scored.
- did not occur
The Dow's Wednesday drop quoted as 600, 700 or 730 in Thursday's recaps. It was 631.21, to 51,461.90.
Thursday’s recaps found carried 631 (Mitrade, KuCoin) or “over 600” (Benzinga), which is true of 631.21.
- occurred
The claims week mislabelled: initial claims cover the week ended Saturday 12 September, continuing claims the week ended 5 September; watch for “week ended September 13” and the prior reported after revision as the print. Consensus 208K and 1.780M.
Continuing claims dated to “the week ending August 5” by a market-news site; the Labor Department says September 5. Entry 114.
- occurred
Starts and permits with the signs swapped: consensus has August starts up (1.31M from 1.239M) and permits down (1.41M from 1.433M); expect “both fell” or “both rose,” and permits reported as starts.
A morning preview carried each series’ figures under the other’s name. Entry 106.
- did not occur
A Philadelphia Fed reading in the thirties described as a collapse (consensus 30.5 from 47.4; thirty is expansion), and the Empire State's September figure quoted as Philadelphia's.
37.8 from 47.4 against 30.5; headlines said “falls less than expected” (Seeking Alpha) and “kept climbing” (Dow Jones via MarketScreener), which describes activity rather than the index. No collapse, no Empire State.
- did not occur
“The Fed sees rates at 4.1 percent in 2027” attributed to the chairman, who submitted no projection.
Searched; not found.
- unresolved
The chairman's line on financial conditions rendered a fourth way; the Board's transcript settles it. Wednesday: “I would be hard pressed to describe broad financial conditions as restrictive” (Reuters), “I am hard-pressed to call financial conditions restrictive” (FXStreet), “hard to describe overall financial conditions as restrictive” (404k Research).
The Board’s preliminary transcript: “I would be hard-pressed to describe broad financial conditions as restrictive.” Reuters’ version is the verbatim one; the other renderings found were paraphrases, not quotations, so none is a fourth.
The Errata, No. 3 · made Mon 21 Sep, delivered Tue morning, for Tue 22 Sep, WTI expiry, three Fed speakers, the Richmond Fed and a two-year auction
scored Tuesday night: none occurred in the form named, eight did not, one detail unresolved
- did not occur
A “$3 drop” in oil that is the contract roll: front-month WTI goes from October (settled $95.78) to November (settled $92.47); a decline of about 3.5 percent that did not happen, or two feeds quoting two contracts without saying which.
TheStreet and Yahoo Finance measured November against November; Avalon printed both contracts, labelled. Two feeds’ changes imply November settled Monday at $92.37 against the $92.47 printed here; unresolved, see the audit block.
- did not occur
The ten-year's first close above 5 percent dated to Monday or “last week” without the date. It was Wednesday 16 September; a Tuesday close above five is a return, not a first.
Not found in that form. Fortune’s “just hit 5%” is the neighbouring error: entry 133.
- did not occur
The $69 billion two-year auction's result with the wrong sign (a tail called a stop-through or the reverse), or its high yield printed as the two-year's closing yield.
investingLive and Helious both reported a 0.2 bp tail at 4.787%; FX.co’s “yield jumps to 4.787%” describes the auction. The size here was wrong: $69 billion, corrected in the audit block.
- did not occur
Goolsbee's Monday London remarks re-dated to Tuesday, or the “painful” line attributed to Williams (10:05), Jefferson (10:20) or Barkin (1:00).
Quartz and Yahoo Finance, 11:48 a.m. Tuesday, dated them to Monday.
- did not occur
The fed funds rate as one number in Fed-speaker coverage: “4%” alone, 3.88, or the AP's “approximately 3.9 percent.” The range is 3.75–4.00.
Not found in the coverage read (Reuters on Williams, Newsquawk, Avalon on Barkin); BBN Times printed the range.
- did not occur
AutoZone's sixteen-week fiscal fourth quarter compared with the twelve-week third: a sequential revenue “jump” that is the calendar. Made after the fact: AutoZone reported at 6:55 a.m. Tuesday ($56.05 a share against Kiplinger's $54.30), before No. 3 went out; the coverage is what is scored.
The release and GuruFocus compared the quarter with the year-earlier sixteen weeks, up 5.6 percent.
- did not occur
The Chicago Fed index reported as “improved to −0.04 from −0.08,” using July's unrevised figure; July was revised to +0.08 and August's −0.04 is a decline.
Benzinga and Trading Economics used the revised +0.08. The unrevised July page surfaced anyway: entry 134.
- did not occur
The Nasdaq's record with the wrong predecessor (27,093.90 on 2 June per Reuters): “first since May,” or the Nasdaq 100's record conflated with the Composite's; and an S&P 500 record called on the intraday high before the close confirms it.
BBN Times named 27,093.90 on 2 June; Quartz had Monday’s as the first since June; the S&P 500 set no record.
The Errata, No. 4 · made Tue 22 Sep for Wed 23 Sep, the flash PMIs, Governor Barr and a five-year auction
scored in No. 5, a day late: two occurred (one in a neighbouring form), five did not, one unresolved
- occurred, neighbouring form
The two-year's 0.2 basis point tail (4.787 against 4.785 when-issued) carried into Wednesday as a stop-through or “strong demand”; or the $78 billion size, this desk's own error, surfacing elsewhere.
Wolf Street’s Wednesday piece: “$79 billion of 2-year notes” and an “$80 billion” five-year; $69 and $70 billion on three feeds. Entry 142. No stop-through found; Phil Newton called demand “merely adequate.”
- did not occur
The five-year auction's high yield printed as the five-year's closing yield, or its result with the wrong sign.
$70 billion at 5.033 against 5.002 when-issued, a 3.1 bp tail (investingLive 1:07 p.m.). No stop-through found; TheStreet printed the closing yield as 4.996.
- did not occur
A flash PMI in expansion described as contraction (manufacturing consensus 53.9, composite 56; fifty is the line), or the ISM's 1 October index quoted as tomorrow's S&P Global flash.
Manufacturing 57.0, services 58.7, composite 58.4 (FXStreet); every outlet read said expansion; the ISM not quoted. The consensus figures here were the August priors; see the audit block.
- did not occur
A third straight Nasdaq record called on the intraday high (Tuesday's was 27,288.79; the record close is 27,244.28), or the Nasdaq 100 conflated with the Composite.
The Composite fell 1.13 percent to 26,936.04 (Minichart; Investrade); Vantage called Tuesday’s the second consecutive record close.
- did not occur
The ten-year's Wednesday move measured from 5.01 rather than 4.96, the mechanism of entry 132; change plus prior should equal 4.96.
Kiplinger up 13.7 bp to 5.104, TheStreet up 13.9 to 5.106: both from about 4.967.
- did not occur
The S&P 500's flat Tuesday (7,764.64, down 0.06) quoted as “edged higher” or as 7,776.60 in Wednesday's recaps.
Not found; Vantage’s “easing 0.04%” is a neighbouring slip (six cents on 7,764 is not four basis points).
- occurred
Tuesday's changes (−0.06, −185.14) beside Wednesday's contract levels on the same data provider's index page, a sixth time; level less change should equal 7,764.64 and 51,863.69.
Trading Economics at 7:24 a.m. Wednesday; the Brief’s entry 135.
- unresolved
November WTI's Monday settlement as two numbers, $92.47 and $92.37, with Tuesday's change quoted from either.
No feed read carried the November contract’s Monday settlement; October figures found (DTN $94.59, Rallies $94.99 for Tuesday) disagree with each other.
The Errata, No. 5 · made Thu 24 Sep for Fri 25 Sep, durable goods, the Michigan final, Williams and Hammack
scores Friday night
- open
PCE reported as Friday's release: August personal income and outlays is due Wednesday 30 September; or July's actuals (0.2 monthly, 3.7 headline, 3.3 core) quoted as the August consensus.
- open
The Michigan final (consensus 47.6) compared with August's 51.7 rather than the preliminary 47.8, and reported as a plunge.
- open
Durable goods (consensus −0.4 percent after +1.1; ex-transport +0.6) reported as a manufacturing collapse on the headline alone, or the ex-transport figure printed as the headline.
- open
Thursday's changes (−1.90, −161.61) beside Friday's contract levels on the same data provider's index page, a ninth time; level less change should equal 7,704.13 and 51,349.98.
- open
The ten-year's Thursday close as three numbers (5.19, 5.206, “5.2”), with Friday's change measured from any of them; the Treasury's Thursday par close settles the base.
- open
The thirty-year's record given the ten-year's date (2007 for 2004), or the seven-year auction's “highest since 1993” attached to a benchmark yield.
- open
A “fourth straight weekly loss” for the Dow printed before Friday's close; last Friday's close was 51,682.64.
- open
Schwab's 202,000 propagating: last week's claims as 202,000, or “claims fell 5,000” in a weekly recap; the Labor Department says 198,000 and 1,000.