* COVID · 27 February 2020: the outlined yellow dot marks the COVID warning reported in the original study. This period is a historical NYSE data blackspot, so the warning cannot be independently confirmed from our inputs. It is shown for context and excluded from calculated warning statistics.
| Five-condition candidate | Highest known spread | Spread date | Known sessions | Added filter |
|---|
The five-session window uses the S&P 500 trading calendar: the candidate day plus its previous four sessions. Missing observations are unknown. A known value of at least 4.5× proves a pass; a fail requires all five known values to be below 4.5×. The Nasdaq AUM history starts on 9 February 2010, so earlier candidates cannot pass this added condition.
The original five-condition reconstruction, including its December 2021, February 2025 and October 2026 examples, is documented in the original reconstruction evidence. Its 14-session expiry, threshold timing and reset policy remain inferred. The February 2020, August 2023 and December 2024 reference warnings still lack sufficient NYSE inputs; this remains a partial historical record.
The setup and confirmation are evaluated using known observations at each close. Missing data is unknown, not a failed condition. No future observations are used to create or backdate a warning.
Click a signal date to inspect it on the chart. Daily horizons count trading sessions; weeks and months use calendar dates, moving forward to the next trading session when necessary. Pending horizons are excluded separately from each summary. Three-month loss and drawdown use daily closes. * Dates inferred from the original monthly labels and matching daily returns.
The study’s five signal conditions are listed above. McClellan describes a preliminary Titanic setup and additional confirmation. The Bluekurtic reference does not publish its exact setup expiry, threshold timing, reset policy or underlying data.
Our calculated candidate arms on a new 252-session SPX closing high, paired with a known NYSE lows-above-highs observation within seven sessions either side of the high. Confirmation requires lows above highs in at least four of the latest five sessions, SPX down closes in at least four of those sessions, and NYSE new highs below 1.5% of issues on at least one of those five sessions on or after the high. The setup expires 14 trading sessions after the high and can fire once per high cycle. These additional timing and reset choices remain inferred.
Added experimental filter: use the exact daily observations from the Nasdaq leveraged ETF AUM spread chart: (TQQQ + QLD assets under management) / (SQQQ + QID assets under management). At least one of the current and previous four trading sessions must have an unrounded ratio ≥4.5. All six conditions must pass before a warning consumes its one-per-high-cycle allowance. A warning also starts a three-calendar-month cooldown shared across all source segments. A fresh high cannot bypass it. The earliest next warning is the first fully qualifying trading session on or after the three-month anniversary; blocked days do not extend the cooldown or consume a new setup. For month-end dates, the anniversary uses the last day of the target month when needed. No future values or carried-forward observations are used. The study’s reported warning series uses its original conditions.
Three source segments are calculated independently: the historical NYSE all-issues archive through 10 February 2020, and the WSJ / Dow Jones NYSE Market Diary for recovered recent windows and daily updates. February 2025 inputs come from original printed WSJ tables and Dow Jones Newswires. October 2026 observations are cross-checked against explicitly dated WSJ tables reproduced by MrTopStep. The December 2021 segment uses five original NYSE StocksRecap tables in the Citrus County Chronicle, 14–18 December 2021; its vendor universe is kept separate from WSJ. The download preserves each observation’s sources.
High/low counts and denominators use the same NYSE universe. Where a report omits unchanged issues, its exact total is left unknown. A threshold can nevertheless be proved when new highs are already below 1.5% of advances plus declines, since adding unchanged issues can only lower that percentage. Such results are labelled as an upper bound, not an exact percentage. Other missing observations are never filled with zeros.
The scheduled server refresh collects explicitly dated completed WSJ sessions and recomputes the filtered experimental warnings and forward outcomes three times daily, after the Nasdaq AUM spread refresh. The embedded saved snapshot is refreshed too. Current-day partial prices are excluded. A failed source fetch preserves prior observations, leaves new breadth unknown and displays a coverage notice. The page’s date is the latest completed price session, which can be newer than its last known NYSE observation.
Daily return horizons count trading sessions. Weeks and months use calendar dates and move to the first trading session on or after the target. Unfinished horizons remain pending and are excluded separately from each summary. Three-month maximum loss is measured from the signal close to the lowest later close; drawdown measures the largest peak-to-subsequent-trough loss, including the initial close.
The first seven reported dates are inferred from the study’s monthly event labels and matching subsequent daily returns; 13 October 2026 is explicitly dated in the study. All 84 completed historical forward-performance values reproduce its published results at the reported precision. This supports the event dates and outcomes; the reconstructed signal rules remain subject to the coverage and timing limitations described above.
Download prices, calculated events, conditions and source-labelled recent inputs · Reconstruction audit