Each daily top 10 on the record, recorded when it printed and checked against SPY. Losing days included.

Restatement notes — 15 June 2026 and 25 August 2026

15 June 2026: scores from 18 May to 12 June capped at 100

A bug let raw factor values, which are not on the 0–100 scale, into the stored score. On 15 June 2026 a database change set every recorded score above 100 to 100 and kept no copy of the originals. All 190 entries recorded from 18 May to 12 June 2026 now read 100. Scores of 120 to 137 had been verified in entries from 22 May to 5 June, and until a fix on 9 June 2026 the daily top 10 could be ranked on such scores. Because the original values were not kept, we cannot tell which of the 190 entries were changed or by how much. The lists were not re-ranked and no entry was removed, so any list that was ranked on scores above 100 still shows the names chosen that way. Prices and next-session results were not touched by this change, and the summary figures do not use the score column. We did not disclose this until 14 September 2026.

25 August 2026: prices restated

Every scored row for sessions up to 21 August 2026 was recomputed on 25 August 2026, bar four the data vendor could no longer price. The “price at flag” column had been recorded from the last trade of the day including extended-hours trading, rather than the official closing price. The SPY column was always taken from official closes, so the two columns were measured on different bases.

For the restated rows both columns are now the official close. 684 of the 688 scored rows were re-derived from the vendor’s unadjusted daily closes for the same two sessions; the remaining 4 are unchanged, because the vendor no longer returns daily bars for those symbols and we would rather leave a row alone than estimate it. 197 rows — 29% — had a flag price more than 2% away from the close, the largest 18%.

The list for 24 August 2026 was recorded shortly before this fix and was not part of the restatement. Its flag prices are still the last trade including after-hours trading; for 7 of its 10 entries that price differs from the official close by 0.08% to 1.36%. It has not been corrected.

Individual rows moved in both directions, by up to 8.8 points of the SPY-relative column. The summary figures moved in both directions too: across the nine windows the correction was applied in, the share of entries that moved further than SPY went up in four and down in four. We are not claiming the correction flattered or hurt the record overall, because we cannot show that, and the current figures are on this page either way.

On 25 August 2026 no entry was added, removed, re-ranked or re-scored. The dates, the symbols, the ranks and the scores stayed as they were stored; only the recorded prices, and the two figures derived from them, changed. The raw exports linked below carry the corrected values and list both restatements, so a check you run against an independent price source reconciles for the restated rows. Expect differences for the 4 rows left as first recorded and for 7 entries on the 24 August 2026 list.

Gaps and known limitations

Entries are not re-ranked or deleted. We have corrected recorded values twice, and said so: prices on 25 August 2026, and scores from 18 May to 12 June capped on 15 June 2026. Below are the gaps and problems we have verified, with dates.

Trading days with no top 10

No list was recorded for these US trading days, and none was filled in afterwards.

• 2026-08-31 (31 August 2026) — Cause not established.
• 2026-09-02 (2 September 2026) — Cause not established.
• 2026-09-04 (4 September 2026) — Cause not established.
• 2026-09-09 (9 September 2026) — Our system stopped writing data at 15:36 UTC that day and did not recover before the daily list was due.

Corrections to recorded values

• 15 June 2026: A database change set every recorded score above 100 to 100 and kept no copy of the originals. All 190 entries from 18 May to 12 June 2026 now read 100. Scores of 120 to 137 had been verified in entries from 22 May to 5 June, and until a fix on 9 June 2026 the daily top 10 could be ranked on such scores. Because the originals were not kept, we cannot tell which of the 190 entries were changed or by how much. The lists were not re-ranked. This was not disclosed until 14 September 2026.
• 25 August 2026: Recorded prices were restated on 684 of the 688 entries for sessions up to 21 August 2026, because they had been taken from after-hours trades instead of the official close. The other 4 were left as first recorded. The list for 24 August 2026 was not part of this restatement and still uses the after-hours prices (see below). See the restatement note above.

Other known limitations

None of these changed a stored entry. The lists from those days stand as they were recorded, including any effect these problems had on them. Some are still open.

• Lists recorded before 24 August 2026: After each restart of our system, tickers not covered by our scoring spreadsheet were scored from random placeholder factor values until the daily data pass ran. A list from this period could include names scored that way, and our stored data cannot show whether any was. Fixed on 23 August 2026.
• The list for 24 August 2026: This list was recorded shortly before we switched to official closing prices, and it was not included in the 25 August 2026 restatement. Its prices at flag still come from the last trade including after-hours trading. For 7 of its 10 entries that price differs from the official close, by 0.08% to 1.36%, so their results are not on the official-close basis either. Not corrected.
• All lists to date: When measured on 6 September 2026, 5,697 of 7,417 scored tickers (77%) had no reading for two of the six factors (fundamentals and insider activity), which then counted as neutral, so the listed names were drawn mostly from the tickers that did have readings. Changes merged on 6 September 2026 (#762) and 7 September 2026 (#775). Coverage is still incomplete: on 14 September 2026, 6,092 of 11,649 scored tickers had neither reading.
• Lists recorded before 7 September 2026, from at least 24 August 2026: Three renamed columns in our scoring spreadsheet were read as missing and scored as neutral, which made many spreadsheet-based scores too high. Corrected on 7 September 2026. Our stored score history starts on 24 August 2026, so the start of the problem is not known.
• 6 to 10 September 2026: The refresh of the trend, relative-strength and momentum inputs stalled, so those factors kept using price data fetched on 6 September. The list for 8 September 2026 was ranked while this was happening. Whether it had caught up before the 10 September list was recorded was not verified. Fixed on 10 September 2026.
• Entries from 22 May to 10 July 2026: As of 14 September 2026, 32 entries from this period have no next-session back-check. Their result columns are blank rather than estimated, and they are not in the summary figures.
• Entries from 24 August to 11 September 2026 (16 entries): 16 entries were ranked while the ticker held a Smart Money value with no SEC Form 4 filing on file: BBH on 24, 25, 26 and 28 August and 1, 3 and 8 September; BBP on 25, 26 and 28 August and 1 September; BIB on 26 August and 1 September; PLX on 8, 10 and 11 September. The same tickers appear in 7 earlier entries (BBH on 19 and 21 August, BBP on 20 August, BIB on 7, 12, 13 and 20 August), where whether they held such a value cannot be checked. At 13:30 UTC on 14 September 2026, 856 tickers held such a value: 629 ETFs, 222 stocks and 5 commodity futures contracts. 42 of those values were below 10 or above 90, which the Form 4 calculation cannot produce. Where the values came from has not been established. From 14 September 2026 (#824) an empty Form 4 answer removes the value, unless a filing we already hold from that source records a transaction in the last 80 days, and such tickers become due for a re-check at the next daily run, ahead of other Smart Money re-checks (#833). Lists not changed.

Most of these are also dated entries on the changelog, and the downloads carry them as restatements, missing_sessions and known_limitations.