Technicals & Seasonality · Follow‑up to No. 001

Five Weeks On: What the MACD Did, What September Did, and What October Now Asks

On 17 August we said the trend deserved the benefit of the doubt but the next two weeks did not deserve fresh leverage; that the signal line was the early warning and the 200‑day was the line that actually mattered; and that September would be rough in path but not in endpoint. An ephemeris is only useful if you go back and check the positions. Here is the same table, five weeks later, with each call scored.

Ideas and concepts by Ward McKinley · written by AI · data through Thursday, 24 September 2026

The short version: the momentum sequence No. 001 laid out played out in order and faster than expected; the seasonal read held; the level that would have changed the assessment was never approached; and the month that was supposed to be the resolution has not yet had anything to resolve. That last point is the one that matters for the next three weeks.

01The momentum sequence, as it happened

No. 001 described a specific pattern: a MACD still rising but rising more slowly than its own average, with the histogram narrowing four straight sessions at the highs. We said that was "not a sell signal" but "the specific pattern that precedes a signal‑line crossdown if price merely goes sideways for another week or two."

Price did not go sideways. It fell 1.3% over the next three sessions, and the bearish signal‑line cross arrived on 20 August — three trading days after publication, not one or two weeks. The histogram then stayed negative for 21 consecutive sessions, through 18 September. The early warning worked; it was simply earlier than the sideways case implied.

S&P 500 daily price and MACD (12, 26, 9), 27 July – 24 September 2026

The dotted verticals mark publication of No. 001, the bearish signal cross three sessions later, the closing low on the day of the Fed's hike, and the bullish re‑cross on 21 September.

MACD line / index priceSignal line (9‑period EMA)Histogram, positiveHistogram, negative

View data table (13 Aug – 24 Sep)

Date	SPX close	MACD	Signal	Histogram
13 Aug	7,798.99	+78.5	+51.3	+27.1
14 Aug	7,785.76	+81.3	+57.3	+24.0
17 Aug	7,745.06	+79.4	+61.7	+17.6
18 Aug	7,691.76	+72.7	+63.9	+8.8
19 Aug	7,707.98	+67.9	+64.7	+3.2
20 Aug	7,641.16	+58.1	+63.4	-5.3
21 Aug	7,674.37	+52.3	+61.2	-8.8
24 Aug	7,652.86	+45.6	+58.1	-12.5
25 Aug	7,677.28	+41.7	+54.8	-13.1
26 Aug	7,675.70	+38.0	+51.4	-13.4
27 Aug	7,730.99	+39.1	+49.0	-9.8
28 Aug	7,711.76	+38.0	+46.8	-8.8
31 Aug	7,686.14	+34.7	+44.4	-9.7
01 Sep	7,631.47	+27.3	+40.9	-13.6
02 Sep	7,666.60	+24.0	+37.6	-13.5
03 Sep	7,747.71	+27.6	+35.6	-7.9
04 Sep	7,718.60	+27.8	+34.0	-6.2
08 Sep	7,673.52	+24.1	+32.0	-8.0
09 Sep	7,636.36	+17.9	+29.2	-11.3
10 Sep	7,591.70	+9.3	+25.2	-15.9
11 Sep	7,656.98	+7.6	+21.7	-14.1
14 Sep	7,619.98	+3.3	+18.0	-14.7
15 Sep	7,585.73	-2.9	+13.8	-16.7
16 Sep	7,551.81	-10.3	+9.0	-19.4
17 Sep	7,637.76	-9.3	+5.4	-14.6
18 Sep	7,650.50	-7.3	+2.8	-10.1
21 Sep	7,764.70	+3.5	+3.0	+0.5
22 Sep	7,764.64	+11.9	+4.7	+7.1
23 Sep	7,706.03	+13.6	+6.5	+7.1
24 Sep	7,704.13	+14.7	+8.1	+6.5

The deeper level was the zero line — "the point where the 12‑day exponential average falls back below the 26‑day," which we called "the first genuine trend‑damage signal of the second half." It fired. The MACD closed below zero on 15 September, the eve of the Federal Reserve meeting, and bottomed at roughly −10 index points on the 16th, the day the Fed raised rates by a quarter point to 3.75–4.00% — its first increase since 2023 — and the index closed at 7,551.81, its low for the window. It spent four sessions below zero.

Then it reversed. On 21 September the index rose 1.5%, its largest gain since early August, and the MACD crossed back above both its signal line and the zero line in the same session. It now stands near +15 against a signal near +8. Two cautions belong here. First, the histogram has already narrowed in each of the last two sessions — the same deceleration shape as mid‑August, from a far lower base. Second, price spent the middle of September below its 50‑day average and is back above it by less than 1%; the 20‑day sits 0.4% below. The cushion under this re‑cross is thin.

The 200‑day moving average — the level we said "actually matters" — was never in play. The index entered September roughly 8% above it and, at the 16 September low, was still about 5% above. The "violent three‑week reversal" required to move this tape into the bad seasonal bucket did not occur, and nothing since has come close.

02September, scored against the two populations

The seasonal argument in No. 001 was that September's headline average of −0.7% is a blend of two populations: entering the month above the 200‑day (+1.3% average, 60% positive) and entering below it (−4.2%, 15% positive). This year was firmly in the first group. Through 24 September the index is +0.2% for the month, with four sessions remaining. That is consistent with the favourable bucket and inconsistent with the calendar‑only view; it is not, with four sessions to go, settled.

The other half of the seasonal call was about shape rather than sign: "dispersion, not drift, is the risk." That is the part that showed up most clearly. Measured from the 13 August record, the drawdown reached 3.2% on a closing basis and 4.0% intraday, all of it recovered by the 21st. Within September alone the index was down 1.7% at the low and up 1.0% at the high. The VIX, which closed at a 2026 low of 14.25 on 14 August, rose 25% to a closing peak of 17.84 on 10 September — the day crude went through $100 on fighting around the Strait of Hormuz and producer prices printed 5.4% year on year — and printed 18.94 intraday on Fed day before settling back to 15.67.

The one part of the seasonal read that has not matched is timing. We noted the VIX's historical tendency to peak in late September and early October. So far the peaks have been mid‑month, attached to specific events rather than to the calendar. That window is still open, and nothing about the macro backdrop argues it will close quietly: the 10‑year Treasury yield is above 5% for the first time since 2007, the 30‑year at 5.48% is the highest since 2004, the Fed's own projections now point to a further hike, and Brent is above $100. That the index is 1.2% from its record against that backdrop is itself the most notable fact of the five weeks.

03What October now asks

No. 001 spent its longest section on October: the month that holds six of the forty worst monthly declines on record and has also ended more post‑war bear markets than any other. The practical read was that weakness beginning in September has historically found its low in October rather than run through it, and that the far side of the window has been a better place to add than to reduce.

The honest complication is that the set‑up described there is not, yet, the set‑up we have. The September low came mid‑month, on a known event, and has been fully recovered. There are two readings of that, and the data does not yet choose between them:

• The low is in. The 16 September close of 7,551.81 was the seasonal low, the Fed's hike was the capitulation event, and the bear killer has nothing to kill. On this reading the 21 September re‑cross is the resolution, and the first half of October is a range rather than an air pocket.
• The low is not in. The recovery is narrow, rates are at cycle highs, the histogram is already narrowing again, and the classic first‑half‑of‑October weakness has not yet been tested. On this reading the 21 September move was a relief rally inside a still‑unresolved window, and the levels below matter more than the levels above.

The levels that would decide it are specific, and they are closer together than they were in August. On the downside: the signal line, near +8, where a close would mark the second bearish cross in six weeks; the 20‑day and 50‑day averages at roughly 7,672 and 7,632, less than 1% below price; and the 16 September closing low at 7,551.81 (7,507.77 intraday), which is the level that defines whether September's low holds. A close below it puts the first‑half‑of‑October script back in play, and at that point the 200‑day near 7,200 — still 7% below — becomes the reference again. On the upside: the 13 August record at 7,798.99, 1.2% away, which the index has now failed to retake on two approaches (27 August and 21–22 September). A new high with the histogram widening and breadth extending beyond semiconductors would settle the question the easy way.

04The scorecard

What No. 001 said	What happened	Score
The trend deserves the benefit of the doubt; the next two weeks do not deserve fresh leverage.	Index −1.3% over the next three sessions and −2.5% from the publication close by 16 Sep; trend structure (200‑day) never threatened.	Right.
The signal line is the early warning; a crossdown follows "if price merely goes sideways for another week or two."	Bearish cross on 20 Aug, three sessions after publication; histogram negative 21 straight sessions.	Right — faster than the sideways case implied.
Losing the zero line is "the first genuine trend‑damage signal of the second half."	Lost 15 Sep, four sessions below, regained 21 Sep. The damage was real but shallow.	Fired — the signal was right, the consequence was small.
A 200‑day break is the one that actually matters; it would take a violent reversal to get there.	Never closer than ~5%; entered September ~8% above.	Right.
This tape is in September's favourable bucket: +1.3% average, 60% positive.	+0.2% month‑to‑date with four sessions left.	On track — not settled.
Expect a rougher path than endpoint; VIX peaks late Sep / early Oct.	−4.0% intraday drawdown fully recovered; VIX 14.25 → 18.94 intraday. Peaks so far were mid‑month, on events.	Right on dispersion, early on timing.
Hedge rather than de‑risk through the window.	De‑risking on 17 Aug would have avoided a 2.5% closing drawdown and missed the +1.5% day of 21 Sep; net −0.5% since publication either way. The distribution, not the drift, is what showed up.	Defensible — neither choice was expensive.
October is the resolution, not the extension; September weakness bottoms in October.	Not yet testable. The low so far was 16 Sep and has been recovered.	Open.

Two cautions on the scorecard itself. Five weeks is a short window, and one September is not a test of a 76‑year base rate; the seasonal calls are scored on consistency, not on proof. And a note like this one is a record of positions, not a claim of foresight — the point of writing down the levels in August was so that the reader, not the author, could check them now.