Scenarios After 200MA Breadth Turns Pink

Data through 2026-10-02 · Current episode began 2026-09-18 · Session 11 · Completed historical episodes: 53

Market breadth dashboard · Pink Onsets Near Price Highs

Rapid switching excluded

Exclude all original segments in a completed chain of at least three pink episodes separated by at most five non-pink sessions, with a total first-onset-to-final-exit span of at most 40 sessions. Retained episodes keep their original dates and outcomes; episodes are not merged.

Original completed episodes: 56; retained: 53. Excluded onset dates: 2015-03-11, 2015-03-26, 2015-04-17.

Fresh long-history data and universe sensitivity

Fresh retrieval covered 1256 current/former symbols; usable adjusted price history was obtained for 983, including all 504 current constituents. SPY covers 1993-01-29 through 2026-10-02. Prices use FMP adjClose; the provider describes it as adjusted for splits and dividends. The archived inputs and acquisition manifest record individual coverage and hashes.

Historical membership was also reconstructed from the provider’s changes, but 43 reverse-history inconsistencies remain. Missing delisted histories prevent complete historical coverage. Its results are a sensitivity model, not the primary forecast evidence.

Different denominator and membership models change episode dates and counts. They should not be pooled into one probability estimate. The reference preserves the visible September 18 onset; the MA-eligible current-cohort variant produces a September 16 onset.

Full-episode results conditioned on surviving to the current age

Only 44 reference episodes were still pink on session 11. For those completed episodes, onset-to-low time averaged 28.7 sessions (median 12.0); low-to-end time averaged 37.8 (median 24.5); the low return averaged -6.9% (median -4.4%). The all-episode table below includes shorter warnings too.

Current conditions

Smoothed 200-day breadth: 61.7% (EMA200); short-term breadth: 46.5% (EMA8); unsmoothed percentage above the stock-level 200-day moving average: 45.4%.

The SPY-derived price series is 769.64, +1.0% since the pink onset and -0.8% below its highest close in the latest 63 sessions.

Full episode: onset, price low, and warning end

The price low is the first minimum close within the pink interval: onset included, first non-pink session excluded. Warning end is the first non-pink session and its close. Onset is day 0. The current unfinished episode is excluded from completed-episode statistics. These are retrospective measurements, not dates knowable in advance.

13 episodes never closed below their onset close during the pink interval. They contribute zero days and zero return to the price-low measurements. Means of the two time segments add to the mean total; medians generally do not.

Among only the 40 episodes that did decline below onset, time to the price low averaged 31.6 sessions (median 16.0), and the price-low return averaged -7.6% (median -4.6%). This outcome-conditioned subset is a diagnostic, not an unconditional forecast.

Each path ends at the first non-pink close. Episodes have different lengths; this chart is retrospective and is not a forecast band.

Sensitivity to the definition of a bottom

Using the minimum close in a fixed 126-session window from onset instead, n=53: time to the low averages 36.2 sessions / 52.6 calendar days, with medians of 21.0 / 29.0. Return averages -7.0%; median -3.8%. A fixed window can miss a later low or end during an ongoing decline.

Current episode: the lowest close observed so far occurred on 2026-09-18, 0 sessions after onset, with an onset-relative return of +0.0%. This is not a confirmed final bottom.

Price low versus breadth-bottom detection

All price returns use the pink-onset close as the denominator. A plotted bottom date and the date that bottom can first be detected are different. Compare only matching episodes; do not subtract averages from different samples or subtract marginal medians to estimate the median paired difference.

200MA breadth bottom: detected in 20 episodes; absent in 33

Detection preceded the eventual within-episode price low in 0/20 matched episodes. A ratio between detection price and the eventual low is not necessarily a realizable forward return.

8MA breadth bottom below 40%: detected in 25 episodes; absent in 28

Detection preceded the eventual within-episode price low in 8/25 matched episodes. A ratio between detection price and the eventual low is not necessarily a realizable forward return.

To determine detection dates, the chart algorithm was replayed on each daily data prefix. For 200MA breadth: find_peaks on the negative series, distance=50 and prominence=0.015. For short breadth: retain EMA8 observations below 40%, then find_peaks with prominence=0.02. A qualifying pivot must be inside the pink interval; the first detection is sought before the next pink onset. Later-revised or disappearing candidates still count as first detections. These are chart detections, not the separate trading engine’s entry signals.

Supplement: fixed-horizon outcomes from today’s equivalent age

Historical episodes must still be pink at the same elapsed age as the current episode. Outcomes are measured from that equivalent date, including after pink ends. Only fully observed horizons that end before the current episode begins are used. The 63-session horizon covers roughly three months ahead—about 73 elapsed sessions from onset in this report—and is not selected for demonstrated predictive power.

5 / 21 / 63 / 126 / 252 sessions are approximately one week / one month / three months / six months / twelve months. The quartiles describe observed dispersion, not confidence intervals. Longer windows have fewer fully observed cases.

These mutually exclusive scenarios use the lowest close relative to the equivalent-age anchor, not peak-to-trough drawdown or intraday lows. An episode can fall substantially and still end the horizon with a positive return.

Repeated-signal sensitivity

Keeping onsets more than 126 sessions apart reduces the same-age sample to 28 for the 63-session horizon: Less than 5% decline: 15/28 (53.6%), 5% to less than 10% decline: 5/28 (17.9%), 10% or larger decline: 8/28 (28.6%). Selection depends on onset dates, not future price outcomes. The sample still contains related market episodes.

What to monitor

• Recovery: short breadth turns up and its gap to long breadth narrows. Distinguish a price-only rally from broader participation.
• Further weakness: short breadth falls again while price makes new lows. The 40% and 30% short-breadth levels are monitoring references; conditional probabilities at those thresholds were not estimated here.
• Warning end: pink ends when either defining condition fails. That change does not guarantee a final price bottom.

Methods and limitations

• Pink means Breadth_200MA_Trend == -1 and Breadth_Index_8MA < Breadth_Index_200MA. The trend uses the code’s hysteresis rule.
• Stock-level 200-day breadth is smoothed using EMA200; its short line is EMA8. This is different from an index price crossing its own 200-day moving average.
• The study uses freshly retrieved price histories with a 1990 warmup, aligned to SPY from 1993; signal onsets in the first 200 SPY observations are excluded to reduce initialization effects. Rising-price episodes are retained. The old report’s hand-selected 2007–2025 sample is not merged into this study.
• Constituent history is not verified as point-in-time; survivorship and missing-data denominator biases may affect the underlying breadth. Constituents and raw provider prices were not independently reconciled.
• S&P500_Price is the project’s SPY-derived series, not the index level. FMP describes adjClose as split- and dividend-adjusted; these price-return calculations are not a transaction-level dividend cash-flow simulation.
• The sample is small and related episodes are not independent. Wilson intervals are reference intervals under a binomial independence assumption, not calibrated forecast intervals. Similarity weighting and extra outcome filters are not applied.
• Price lows and full pink intervals are identified retrospectively. Reported returns exclude costs and do not constitute simulated trades. No trading profitability or optimal entry rule is established.
• Price data use closing observations, not intraday lows. Future news and macroeconomic conditions after the data cutoff are not incorporated.

Changes from the previous publication

The previous page described the March 12, 2026 episode as of March 26, excluded rising-price divergence cases, and assigned similarity-weighted bottom forecasts. This report studies the September episode with reproducible criteria, retains rising-price cases, treats March as a completed historical episode, and does not carry forward the old probabilities, dates, or price targets.

Data-source documentation: FMP historical price API; daily chart. History begins 1993-01-29. Snapshot SHA-256: 80e1c0848645543854236661f7bf7f80123d017379392e625eee8fff17410e31. This report is frozen at 2026-10-02; daily chart updates do not automatically refresh its statistics.