Technicals & Seasonality

The S&P 500 at the Highs: What the Daily MACD and the Autumn Calendar Actually Say

The index sits within half a percent of its record. The daily MACD is decisively positive but decelerating — the histogram has narrowed four straight sessions. And the calendar turns hostile in three weeks, though the seasonal statistic everyone quotes is not the one that applies to this tape, and the month that follows has a better record than its reputation.

Ideas and concepts by Ward McKinley · written by AI · data through Monday, 17 August 2026

Every year around this point, the same two questions arrive together: is the tape still healthy, and does September actually matter. This year the honest answers are yes, but less than it was and only conditionally — and the second answer is far more nuanced than the seasonality chart that gets circulated every August, particularly once October is put back alongside it. Together they argue for a specific posture over the next six weeks rather than a directional call.

01The tape: strong, but the second derivative has turned

The S&P 500 closed Friday at 7,785.76, roughly 0.4% below the 7,816.70 intraday high set earlier this month, and above every major moving average — about 2.6% above the 20‑day, 3.6% above the 50‑day, and 10.1% above the 200‑day. The 14‑day RSI sits in the mid‑60s: firm, not stretched. Directional strength is real, with a 9‑day ADX near 35 and +DI comfortably above −DI. This is a trending market, and nothing in the price structure argues otherwise.

The interesting detail is in the momentum oscillator. The daily MACD (12, 26, 9) generated a bullish signal‑line crossover on 3 August and pushed above the zero line on 4 August, the session the index gapped 1.8% higher. That is a textbook confirmation sequence: momentum turning before the zero‑line break, then the break itself validating the turn. The MACD line has continued to rise every session since and now stands at roughly +82 index points against a signal line near +62.

S&P 500 daily price and MACD (12, 26, 9)

The histogram — the gap between the MACD line and its signal — peaked on 7 August and has narrowed in each of the four sessions since, even as price held near its high.

MACD line / index price Signal line (9‑period EMA) Histogram, positive Histogram, negative

View data table (last 25 sessions)

Date	S&P 500	MACD	Signal	Histogram
2026-07-14	7,540	+9.5	-7.1	+16.6
2026-07-15	7,570	+14.3	-2.8	+17.1
2026-07-16	7,529	+14.7	+0.7	+14.0
2026-07-17	7,454	+8.8	+2.3	+6.5
2026-07-20	7,442	+3.2	+2.5	+0.7
2026-07-21	7,504	+3.7	+2.7	+0.9
2026-07-22	7,496	+3.3	+2.8	+0.5
2026-07-23	7,403	-4.4	+1.4	-5.8
2026-07-24	7,411	-9.8	-0.8	-8.9
2026-07-27	7,412	-13.7	-3.4	-10.3
2026-07-28	7,430	-15.3	-5.8	-9.5
2026-07-29	7,316	-25.4	-9.7	-15.7
2026-07-30	7,438	-23.3	-12.4	-10.9
2026-07-31	7,492	-17.1	-13.4	-3.7
2026-08-03	7,599	-3.5	-11.4	+7.9
2026-08-04	7,736	+18.1	-5.5	+23.6
2026-08-05	7,720	+33.5	+2.3	+31.2
2026-08-06	7,708	+44.3	+10.7	+33.6
2026-08-07	7,755	+56.0	+19.8	+36.2
2026-08-10	7,753	+64.3	+28.7	+35.6
2026-08-11	7,728	+68.1	+36.6	+31.6
2026-08-12	7,747	+71.9	+43.6	+28.3
2026-08-13	7,801	+78.3	+50.6	+27.8
2026-08-14	7,786	+81.3	+56.7	+24.6
2026-08-17	7,776	+81.9	+61.7	+20.1

But the histogram tells a different story from the line. It peaked at roughly +36 index points on 7 August and has compressed in every session since, to about +20 now. The MACD is still rising; it is simply rising more slowly than its own average. That is momentum deceleration at the highs — not a sell signal, and on its own a common feature of any sustained advance, but the specific pattern that precedes a signal‑line crossdown if price merely goes sideways for another week or two.

02September and October: the only reliably bad stretch — conditionally

September's reputation is earned. Since 1928 the S&P 500 has averaged a 1.2% decline in September, and it has finished negative 55% of the time versus 39% for every other month combined. Nine of the forty worst monthly declines on record landed in September, more than any other month; October is second with six. On the post‑1950 sample the average is −0.7% with a 44% win rate — the only month of the twelve that fails to clear a coin flip.

S&P 500 average monthly return, 1950–2025

Bars show the average return; the percentage beneath each month is the share of years that finished positive. September is the outlier in both.

Positive averageNegative average

View data table

Month	Avg return	Positive
January	+1.0%	59%
February	-0.1%	54%
March	+1.1%	64%
April	+1.5%	71%
May	+0.3%	59%
June	+0.1%	54%
July	+1.2%	59%
August	-0.1%	55%
September	-0.7%	44%
October	+0.9%	61%
November	+1.5%	68%
December	+1.4%	74%

The distribution is worth more than the average. When September is down it is badly down — averaging −3.8% — and when it is up it averages +3.2%. This is a fat‑tailed month, not a slow‑bleed month, which is exactly why it rewards hedging over de‑risking. Recent history has been worse than the long run: four of the past five Septembers averaged roughly −4.2%, including the −9.3% inflation shock of 2022.

Here is the part that usually gets left out of the seasonality chart. September's average is not a single number — it is two very different numbers depending on the trend the market carries into it.

September returns, split by trend going into the month

The same month, conditioned on whether the index entered September above or below its 200‑day moving average.

Entering September above the 200‑day, the index has averaged +1.3% with a 60% win rate. Entering below it, −4.2% with a 15% win rate. Almost the entire September effect is a downtrend effect. The index is currently about 10% above its 200‑day and would need a violent three‑week reversal to enter September in the bad bucket. On the base rates that matters far more than the calendar.

One structural caveat holds regardless: the VIX has historically peaked in late September and early October. Even in the favourable bucket, the path is rougher than the endpoint.

October: the bear killer

Which leads to the month that actually closes the window. October has a split reputation, and both halves of it are earned. It owns the two most infamous single days in market history — 1929 and 1987 — plus the worst of the 2008 collapse, and it holds six of the forty deepest monthly declines on record, second only to September. That is the source of what the Stock Trader's Almanac calls "Octoberphobia."

The other half is the more useful one. The Almanac's name for October is the bear killer, because more bear markets have ended in October than in any other month: twelve since the Second World War — 1946, 1957, 1960, 1962, 1966, 1974, 1987, 1990, 1998, 2001, 2002 and 2011. Declines that begin in the September window have a strong historical tendency to find their low in October rather than to extend through it.

The practical read is that the risk window running from roughly the second week of September to the middle of October should be treated as one continuous stretch rather than two months. It is the widest‑dispersion patch on the calendar. It is also, historically, where the better entries have been found — provided the trend structure survives it.

03What the two readings add up to

They do not conflict — they sequence, and they point at the same six‑week window.

• Momentum is positive but decelerating. The MACD is above its signal line and above zero; the histogram has narrowed four straight sessions. Price is above every major moving average, RSI sits in the mid‑60s, and directional strength is intact. The trend deserves the benefit of the doubt. The next two weeks do not deserve fresh leverage.
• September's bark is conditional. Entering the month roughly 10% above the 200‑day puts this year in the bucket that has averaged +1.3% with a 60% win rate, not the one that has averaged −4.2% with a 15% win rate. The headline −0.7% average is a blend of two populations, and this tape is in the better one.
• Expect a rougher path than endpoint. Even in the favourable bucket, September is fat‑tailed — averaging +3.2% when it works and −3.8% when it does not — and the VIX has historically peaked in late September into early October. Dispersion, not drift, is the risk.
• October is the resolution, not the extension. The month holds six of the forty worst monthly declines on record, and it has also ended twelve post‑war bear markets — more than any other month. Weakness that starts in September has historically bottomed in October rather than run through it.

The practical implication is about instrument choice rather than direction. A month with a favourable base rate but a wide distribution is a month to hedge rather than de‑risk: reducing exposure gives up the 60% case to avoid the 40% case, while carrying protection through the volatility peak keeps both. The two levels that would change the assessment are specific. A close that pulls the MACD back below its signal line — the first bearish cross since 23 July — is the early warning. A break of the 200‑day moving average is the one that actually matters, because it is the line that moves this market from the +1.3% September population into the −4.2% one. Until that happens, the seasonal case against being long is weaker than the calendar makes it look — and the bear‑killer record argues that the far side of this window has been a better place to be adding than reducing.