Market topsand bottoms.

How stretched is the S&P 500? Fear, momentum, breadth and liquidity in one place.

Signal board

Where fear, credit, breadth and liquidity stand right now.

Loading market data…

Historical research, not investment advice · Updated daily after the close · Methodology

Bottom and top scores

Loading…

Each factor is ranked 0 to 100 against every earlier day it has data for (100 = looks most like the turn), and a score is the plain average of its factors. A bottom score in its zone came at or near all 15 SPY lows that ended a 15%+ slide since 1998, but it also flags slides that kept going. Low fear and overbought readings alone are not a top warning: they mark calm uptrends, which were followed by fewer drops than usual.

How this is calculated

SPY and VIX factors are ranked from the mid-1990s, the Baa spread from 1986, and equal-weight vs cap-weight only from 2003. The top score looks instead for what the cycle peaks of 2007, 2018, 2020 and 2022 shared: price near its high while the Baa credit spread rose off its low and the equal-weight S&P 500 lagged. A day gets a top score only when all three factors exist, so the top score starts in 2004 and the 2000 peak has none.

SPY weekly close, bottom score and top score, last 5 years

What followed past readings in the zone

History, not a forecast. Compare each zone row with the "All days" row under it. The factors and cut-offs were chosen by hand on this same history, so these odds flatter the scores; nothing here is tested out of sample. A bottom score in the zone has come with bigger rebounds and also far more 10%+ further drops: it says "washed out", not "the low is in".

Scores at past 15%+ slides

What each score read at every SPY slide of 15% or more since each score begins (the bottom score in the mid-1990s, the top score in 2004, once all three of its factors exist), newest first, with the current pullback on top when SPY is below its last swing high. Scores are whole numbers as they would have shown that day; days are trading days. The rows at the bottom summarize the finished slides only; the current pullback is not counted.

In-sample, not a forecast. The factors and cut-offs were picked by hand while looking at these same turns, so this table flatters the scores, the top score most of all. The slide lows are marked with hindsight, which nobody had on the day, and many of the peaks are rallies inside a longer bear market, not cycle highs.

How this is calculated

A slide is a fall of at least 15% on SPY closes from a swing high to a low that was then confirmed by a 15% rise off it (a 15% zigzag, fixed in advance). An episode is a run of days in the zone, with runs less than 21 trading days apart counted as one. A bottom episode is a miss when it touches no slide from its peak to 21 trading days after its low, or a fall of 15%+ still under way; a top episode is a miss when no slide peak came from its first day to 126 trading days after its last. A top episode whose 126-day window is still open, or whose window holds a high not yet followed by a 15% fall, is unresolved, not a miss.

Weekly Williams %R

Loading…

Williams %R (21 weeks) measures where the weekly close sits in its 21-week range: 0 is the top of the range, −100 the bottom; the teal line is its 13-week average. Red dots on the price mark weeks that closed within 5 points of −100. Washouts like that have often come near market lows, but they can repeat while a market keeps falling, so read one as stretched, not as a buy signal. Not tested here.

VIX fear spikes and weekly momentum

Loading…

The VIX spikes when investors rush for protection, and the largest spikes have tended to come near stock-market lows; the oscillator (VIX weekly Williams %R, same settings as above) near 0 flags a VIX at the top of its recent range. Amber diamonds mark weeks within 10 points of 0. Near −100 the VIX sits at the bottom of its range, which is complacency and has often come near market highs. The dashed line at 25 is the usual fear threshold. Not every spike marks a lasting low.

Weekly MACD

MACD (12, 26, 9) on weekly closes. Bars are the histogram, the gap between the MACD line and its signal line. Deep negative histogram readings show momentum stretched to the downside; the marked lows are the deepest histogram readings in each of the last three downswings. Context, not a tested signal.

Historical odds

History, not a forecast: nothing here has been tested for an edge. Compare each "today's regime" row with the "all days" row under it; if they look alike, the regime tells you nothing. Days overlap, so "non-overlapping" counts the windows that share no days; fewer than 30 is flagged. "Fell 5%+" is how often the lowest close along the way was at least 5% under the starting close, even if the price recovered by the end.

SPY daily streak

Past cases

Date is the day each earlier streak reached this length; the move so far is measured from the close before it began, and the later columns are SPY's change from that day's close. A dash means that outcome is not known yet.

History, not a forecast: nothing here has been tested for an edge. Fewer than 10 cases is flagged. Compare each streak row with the "All days" row under it, the result after any ordinary day; if they look alike, the streak tells you nothing. The "Lasted" column is hindsight, shown for context and used in none of the figures.

A case is every past streak that reached this many days, whether or not it ran longer, because that is all anyone knew at that day's close; counting only streaks that ended there would be hindsight. Cases are not filtered by how far SPY had moved, since picking a cut-off after the fact flatters the result.

Gap-up reversals

Most recent cases

Gap is the open against the prior close and Day is that day's close against the prior close. The rest is the change from the close of the reversal day; "Low within 1 month" is the lowest close in the following month. A dash means that outcome is not known yet.

History, not a forecast: nothing here has been tested for an edge. Fewer than 10 cases is flagged. Compare each reversal row with the "All days" row under it, the result after any ordinary day; if they look alike, the reversal tells you nothing. The QQQ history leans on the 2000–2003 crash, when these days were common.

How this is calculated

The rule (a gap up of at least 1% at the open, then a close below the prior close) was fixed in advance with round numbers, not tuned to fit a recent day. "Fell 5%+" is how often the lowest close along the way was at least 5% under the reversal day's close.

Triple witching

Most recent cases

SPY's change from the close of each triple-witching day. A dash means that outcome is not known yet.

History, not a forecast: nothing here has been tested for an edge. Compare each triple-witching row with the "All days" row under it, the result after any ordinary day; if they look alike, the date tells you nothing. The week after has been weaker than an ordinary week, but the effect is small and it is gone after a month.

How this is calculated

Triple witching is the quarterly expiration of stock options, index options and index futures. The rule was fixed in advance by the calendar: the third Friday of March, June, September and December, or the session before it when that Friday was a market holiday. "Fell 5%+" is how often the lowest close along the way was at least 5% under the triple-witching close.

After 5%, 10% and 20% drops

History, not a forecast: nothing here has been tested for an edge. S&P 500 index, price return, excluding dividends. A case is the first close at least 5%, 10% or 20% under the record close so far, counted once per slide; a slide ends only when the index closes back at its record. It is known at that close: the eventual low is never used. Compare each row with the "All days" row under it, the result after any ordinary day. "Non-overlapping" counts the windows that share no days; fewer than 10 is flagged. "Fell 10%+" is how often the lowest close along the way was a further 10% or more under the close that marked the drop. The data starts in December 1949, before the index passed its 1929 peak in 1954, so the first few years measure from the highest close since 1949.

Sentiment and positioning

Context, not a signal: each gauge shows its value and where that sits in its own history. None of these has been tested here as a predictor of direction. Extremes are often read as contrarian, but they can stay extreme for a long time.

Are bonds cushioning stocks?

Whether Treasuries have been rising when stocks fall, from the daily returns (dividends included) of SPY and IEF, the iShares 7–10 year Treasury ETF, over the last 63 trading days. Context, not a signal: each figure shows where it sits among all past 63-day windows. IEF covers one maturity range, and overlapping windows are not independent evidence, so the percentiles overstate how much history there is.

Breadth and sector rotation

How widely the market is participating, from the 11 SPDR sector ETFs. Narrow leadership is a classic late-cycle warning; broad washouts show up near lows. Returns include dividends; "vs SPY" is the sector's return minus SPY's over the same dates. Sector participation is the share of sector funds above their 200-day average each day since 1999, counting only funds with 200 days of history (real estate joins in 2015, communication services in 2018); it measures sectors, not individual stocks. EFA and EEM compare developed and emerging markets outside the US with SPY. Percentiles compare each figure with its own history. Context: none of it has been tested here as a forecast.

Sectors, ranked by 3-month return vs SPY

Macro and liquidity

Methodology

About Ticker Otter. A research dashboard that tracks how stretched the S&P 500 is, using signals that have marked past market tops and bottoms. Data comes from free public sources: Yahoo Finance, FRED (St. Louis Fed) and CBOE. It is historical research, not investment advice.

Weekly charts. SPY, QQQ and VIX daily bars are rolled into ISO weeks (the latest week is still in progress). Williams %R = −100 × (21-week high − close) / (21-week high − 21-week low), with a 13-week exponential average. MACD is the standard 12, 26 and 9 on weekly closes. All are drawn from the same data as the other gauges, on every update. None has been tested here as a signal.

Live gauges. Sentiment, breadth, macro and liquidity are recomputed on every scheduled update from public data. Sentiment, breadth and the 10-year yield split are shown with the percentile of their own history; the other macro and liquidity tiles show the latest reading, its recent change and the date of the observation, marked stale when a series has not updated on its usual schedule. They are context: extremes can persist.

Bottom and top scores. Bottom factors: SPY close vs its 52-week high, VIX level, VIX vs its 50-day average, 14-day RSI, and close vs its 200-day average. Top factors: SPY close vs its 52-week high, the rise of the Baa corporate spread (FRED BAA10Y, taken a day late) above its 52-week low, and the 3-month change of equal-weight (RSP) vs cap-weight SPY. Each factor is ranked against strictly earlier days only, so no past score used later data; a score is the equal-weighted average of its ranks (a day is scored only when every factor of its side is ranked, so the top score starts in 2004, once RSP and the credit spread both have a year of ranks). The factor sets and cut-offs (bottom 80, top 70, elevated 60) were picked by hand after comparing about ten variants on the same history, so the odds table is in-sample. All-time-high figures use the S&P 500 index close since 1950.

Regime analogs and drawdown odds. Every past SPY day is sorted by trend (close above or below its 200-day average) and by volatility (low, middle or high third of VIX, with the cut-offs taken only from VIX closes before that day). For today's regime the table counts what followed on past days in the same regime, next to all days: how often the price ended higher, the median and 10th to 90th percentile return, and how often the lowest close inside the window was 5% or 10% under the starting close. Only outcomes that are already known are counted. Neighbouring days share most of their window, so the non-overlapping count (windows that share no days) is the honest sample size, and a regime with fewer than 30 is flagged. These are plain historical frequencies; they are not fitted, not tested out of sample and not a forecast.

For information and research only. Not investment advice. Markets are unpredictable and past patterns do not guarantee future results.

Data sources