Loading…
Pre-openThe market has not opened yet. Showing Wed 23 Sept's close, 16:00 ET. These numbers start moving again when it opens today at 09:30 ET.
Loading…
Every time this setup happened before on one ticker, and what came next — including the times it went badly.
8,470 trading days · yahoo · close 2026-09-23
This page finds every time in the past that today's setup happened, and shows what came next. The grey row underneath each result is the same question asked about every day in the history. If the two rows look alike, the pattern didn't tell you anything. Most patterns don't. This is a record of what already happened, not a forecast.
It happened 450 times, each marked on the price history below.
History used · 8,470 trading days for SPY, 1993-01-29 to 2026-09-23 (33.6 years). Prices are adjusted for stock splits and exclude dividends.
1 break of more than five days between sessions: 2001-09-10 → 2001-09-17. A market closure and a hole in the data look the same from here.
74 separate stretches left · clear
Covers the whole history bar the first 50 days, so from 1993-04-12.
Covers the whole history bar the first 200 days, so from 1993-11-11.
VIX starts in 1990, so this reaches back to 1993-01-29 — the whole of this history. Thirds are cut on the VIX days that overlap this symbol, not on all of VIX.
Unavailable. Nothing on this site stores what the dealer book looked like on a past date — the session series records breadth and sectors only, it began weeks rather than years ago, and it cannot be backfilled. This filter would cover almost none of the history, so it is switched off rather than shown against a handful of days.
The number beside each option is how many separate stretches would be left if you picked it. Below 10 stretches no results are shown at all.
Third straight close below the previous close.
Nothing special happened after this.
The market went up 65% of the time — about the same as it usually does.
The grey line under each row is what normally happens. If the two lines look alike, this pattern did not matter.
| How long after | Separate stretches | Typical result | Best time | Worst time | Went up | Typical dip | Worst dip |
|---|---|---|---|---|---|---|---|
| 1 day1 session | 74144 days | +0.1% | +2.0%2008-06-04 | -3.0%2015-08-20 | 58% | 0.0% | -3.0% |
| What normally happensall 8,470 days, not just these | 2,823days | +0.1% | — | — | 53% | 0.0% | — |
| 1 week5 sessions | 74144 days | +0.3% | +4.3%1997-04-11 | -4.4%2016-01-04 | 58% | -0.6% | -8.2% |
| What normally happensall 8,470 days, not just these | 2,822days | +0.4% | — | — | 59% | -0.6% | — |
| 2 weeks10 sessions | 74143 days | +0.9% | +7.2%2023-01-19 | -6.4%2016-01-04 | 60% | -1.1% | -8.2% |
| What normally happensall 8,470 days, not just these | 2,817days | +0.7% | — | — | 61% | -1.0% | — |
| 1 month21 sessions | 74141 days | +1.5% | +14.5%1997-04-11 | -14.5%2011-07-12 | 65% | -1.8% | -14.6% |
| What normally happensall 8,470 days, not just these | 2,811days | +1.3% | — | — | 64% | -1.7% | — |
| 2 months42 sessions | 73140 days | +1.9% | +19.0%1997-04-11 | -16.3%2002-05-29 | 65% | -2.5% | -25.5% |
| What normally happensall 8,470 days, not just these | 2,795days | +2.2% | — | — | 66% | -2.6% | — |
This happened 144 times, but they came in clumps — about 74 separate stretches. So it is really 74 stories, not 144.
After 3 consecutive down closes, the typical result over 2 months was +1.9%, across 74 separate stretches. On a normal day it came to +2.2%. Little difference between the two.
History used: 1993-01-29 to 2026-09-23 (33.6 years, 8,470 trading days) · it happened between 1993-02-17 and 2026-09-16
"Better" and "worse" in the headline mean the median return was at least 1.5 percentage points from the baseline over 2 months. When the median and the positive rate disagree in direction, the headline reports the disagreement instead of picking one. These are presentation cutoffs, not a significance test.
Price returns, excludes dividends — close to close from the session that completed the condition, on split-adjusted prices.
"Typical dip" and "worst dip" are the maximum drawdown inside the window, measured on closes rather than intraday lows: the deepest close against the entry close. An intraday figure would be deeper.
Periods showing fewer times than the total have matches whose forward window has not finished yet; those are truncated out of that row rather than counted early.
The what normally happens row is the unconditional baseline: every window of that length in the same lookback, condition or not, measured identically — all 8,470 sessions, not a sample.
It includes the early sessions where this condition could not yet fire — a 200-day average and a 52-week high need a year of history before they exist. It is therefore identical under every table on this page, which is what makes the conditions comparable with each other.
Episodes are separate stretches of market, not single days: several times close together count as one.
The comparison worth making is the gap between the two rows, not the level of either. No significance test is applied anywhere on this page.
Both rows of the table are filtered: the pattern days and the 2,824 comparison days both passed the same filters, so the two describe the same kind of day.
Greyed dates fall within two months of the one before, so they belong to the same stretch of market rather than a separate one.
74 separate stretches drawn · the highest finish was 1997-06-04 and the lowest 2002-05-29, the two amber lines. Hover any line for its date.