Sector breadth today

SPY after each state

Averages over every day in that state; the periods overlap, so treat them as a description, not a forecast.

Every decline of 10% or more

“Bearish by the high”: bearish at some point from a month before SPY’s high to a week after it.

Variants tested

The same measures for other common settings, so the result can be checked for luck: a rule that only works with one exact number is fragile. The highlighted row is the published rule.

“SPY or T-bills”: SPY, switching to T-bills (BIL) at the next close after the rule turns bearish and back after it stops, 0.10% per switch, interest and dividends included. Hypothetical: no costs beyond that, no taxes.

Overlap with the other indicators

It agreed most often with Stocks above their 200-day average (87% of days in the same state). An indicator that almost always agrees with another adds little to the score.

Research: what worked and what did not

The same test run many ways: the published rule year by year, every variant year by year, and grids over its settings. Return numbers are SPY-or-T-bills against SPY over the same days, in points; green means better than holding SPY, red worse.

Year by year

* Part of a year: Sep 18 to the end of 2017, and 2026 to Oct 16.

Every variant, year by year

Does a setting win steadily, or in one lucky year? Each square is that year’s difference from SPY.

The settings, tested on a grid

Bearish cut-off and waiting: return against SPY

Each square: how much more (green) or less (red) a year SPY-or-T-bills returned than SPY. Rows: bearish when that many sectors or fewer have most stocks above the 200-day average. Columns: how many trading days in a row a new state must hold before it counts (the usual way to cut false signals; the site's rule does not wait).

Bearish cut-off and waiting: how much smaller the worst fall was

Each square: SPY's worst fall minus the strategy's, in points (green: it fell less). Rows: bearish when that many sectors or fewer have most stocks above the 200-day average. Columns: how many trading days in a row a new state must hold before it counts (the usual way to cut false signals; the site's rule does not wait).

Bullish cut-off: SPY after the bullish days

Each square: SPY's average return over the next 63 trading days from the bullish days, minus its average from every day, in points (green: bullish days were followed by better than usual returns). Rows: bullish when that many sectors or more are healthy. Columns: how many trading days in a row a new state must hold before it counts (the usual way to cut false signals; the site's rule does not wait).

The two halves of the history

The best settings found over the whole period, and how each did in 2017–2021 and in 2022–2026 on its own. A setting that only works in one half is fitted to the past.

Summary: the settings that helped the most

• Best single setting tested: the site's rule, but bearish at 1 or fewer, waiting 2 days: −3.8 points a year over SPY (−4.6 in 2017–2021, −3.2 in 2022–2026), against −8.0 for the site’s rule. Found by looking at the results, so it is the most flattering number here, not a forecast.
• Bearish at: 1 or fewer did best on average (−4.2), 2 or fewer worst (−6.3).
• Wait: 3 days did best on average (−4.7), No wait worst (−6.8).
• Bullish at (for the bullish days): 6 or more did best on average (−0.6), 10 or more worst (−1.3).
• Wait (for the bullish days): made little difference (averages within half a point).
• Choosing the best setting in advance: the setting with the best result in 2017–2021 (the site's rule, but bearish at 2 or fewer, waiting 2 days, −1.6) scored −7.4 in 2022–2026, against −7.7 for the site’s rule; 37 of the 45 settings tested beat the rule in those later years.

The site keeps the rule it published, so its live record stays honest. The heat maps show which ingredients mattered across many settings; any one best square is partly luck, because the history holds only a handful of real declines.