In short
- Long history is favourable, the last ten years are not. Over 19 to 41 Octobers (depending on the index) the month has been positive 62–71% of the time, with median returns between +1.9% and +3.3%. Over the last ten years the average runs from −0.6% to +0.9% and the share of positive Octobers falls as low as 40% (ASX 200). The long history also carries deep falls: −21.8% for the S&P 500 and −27.0% for the Nasdaq 100 in 1987, −14.7% for the Euro Stoxx 50 in 2008.
- The day-of-month pattern is weak. The 12th trading day was positive in 8 of 10 years for the Nasdaq 100, the FTSE MIB and the ASX 200, but in only 3 of 9 for the DAX. Days 18–20 are negative most often in Australia (7 years of 10). With ten observations these are leads to check, not signals.
- Stocks are moving little together. Average pairwise correlation over 30 sessions is between 0.06 (ASX 200) and 0.22 (FTSE MIB), below the five-year average everywhere. October is not a structurally different month for correlation: the gap to the rest of the year is at most 0.04.
- The calendar is packed at both ends. First week: US jobs on 2 October. Last week: the Fed on the 28th, the ECB on the 29th, the Bank of Japan on 29–30 and US labour costs on the 30th. In between, US CPI on the 14th and the start of earnings season with JPMorgan on the 13th.
- Three scenarios, with historical ranges. Down, middle and up, each tied to the average of the matching fifth of past Octobers. These are reference ranges, not forecasts: matching a scenario to today is a judgment, not a calculation.
Where we start from
Macro elements gathered by web search on 26 September 2026. They should be re-read at the source before use: not every page was opened in full (see the section on AI).
Central banks
- Fed: 25 bp hike on 16 September, 12–0 vote, funds rate at 3.75–4.00%. Futures price about 4.3% by December.
- ECB: hike on 10 September, the second since the US–Iran war began, deposit rate at 2.50%.
- Bank of England: held at 3.75% on 17 September, with 3 of 9 votes for a hike.
- RBA: decision on 29 September, with the four major Australian banks expecting a hike to 4.60%. The outcome was not known at the time of the analysis.
Markets and macro
- The 10-year Treasury yield reached 5.23% in the week of 21 September, the highest since June 2007.
- Oil around $100 in the context of the US–Iran war, lower late in the week on hopes of diplomatic progress.
- September jobs expected to slow sharply: about 85,000 against 162,000 in August, unemployment 4.1%.
- In our 18 September crash check none of the seven markets was inside one of the studied crash phases; the ASX 200 was the only one below its 200-day average.
October seasonality
Index return from the September close to the October close. Full available history of each index on the left, the last ten years (2016–2025) on the right.
| Index | Octobers | Average | Median | % up | Worst | 10-year avg | % up, 10 years |
|---|---|---|---|---|---|---|---|
| S&P 500SPY | 41 (1985–2025) | +1.0% | +1.9% | 63% | −21.8% (1987) | +0.7% | 50% |
| Nasdaq 100QQQ | 40 (1986–2025) | +1.7% | +3.3% | 62% | −27.0% (1987) | +0.9% | 50% |
| Euro Stoxx 50EZU | 19 (2007–2025) | +0.7% | +2.0% | 63% | −14.7% (2008) | +0.2% | 60% |
| DAX 40EWG | 38 (1988–2025) | +2.0% | +2.4% | 68% | −14.5% (2008) | −0.0% | 60% |
| FTSE MIBEWI | 28 (1998–2025) | +1.6% | +2.9% | 71% | −16.3% (2008) | +0.8% | 70% |
| FTSE 100EWU | 41 (1985–2025) | +0.5% | +2.1% | 68% | −26.0% (1987) | −0.6% | 50% |
| ASX 200EWA | 33 (1993–2025) | +0.8% | +1.9% | 64% | −12.7% (2008) | −0.2% | 40% |
Indices in local currency: price for the S&P 500, Nasdaq 100, Euro Stoxx 50, FTSE MIB, FTSE 100 and ASX 200; the DAX is a total-return index. History available: Euro Stoxx 50 from 2007, FTSE MIB from 1998, ASX from 1993, Nasdaq 100 from 1986, DAX from 1988. Averages are those of the capitalisation-weighted index, so they differ from the equal-weight stock averages shown in the other R&D notes.
Average October: full history against the last ten years
The pattern inside the month
Equal-weighted market of the stocks, trading days numbered from the start of October, 2016–2025.
| Market | 12th day up | 12th day, avg | Days 18–20 down | Days 18–20, avg |
|---|---|---|---|---|
| S&P 500SPY | 6 of 10 | +0.45% | 5 of 10 | −0.75% |
| Nasdaq 100QQQ | 8 of 10 | +0.48% | 4 of 10 | −0.63% |
| Euro Stoxx 50EZU | 6 of 10 | +0.52% | 4 of 10 | −0.46% |
| DAX 40EWG | 3 of 9 | +0.06% | 2 of 9 | −0.46% |
| FTSE MIBEWI | 8 of 10 | +0.73% | 5 of 10 | −0.63% |
| FTSE 100EWU | 7 of 10 | +0.44% | 5 of 10 | −0.43% |
| ASX 200EWA | 8 of 10 | +0.52% | 7 of 10 | −0.71% |
The 12th day is the best on average in five of seven markets in the R&D note "October, day by day", but consistency across years is mixed. Looking at 21 days in 7 markets, some peaks appear by chance.
Pairwise correlation between stocks
Average of the correlations between all pairs of stocks in each market, on daily returns. Zero means independent stocks, one means a single block.
| Market | Today (30 sessions) | 5-year average | Percentile | Average in Octobers | Average, other months | Correl. with October return |
|---|---|---|---|---|---|---|
| S&P 500SPY | 0.10 | 0.26 | 5th | 0.27 | 0.27 | +0.09 |
| Nasdaq 100QQQ | 0.09 | 0.27 | 6th | 0.30 | 0.28 | −0.24 |
| Euro Stoxx 50EZU | 0.18 | 0.30 | 17th | 0.29 | 0.33 | −0.20 |
| DAX 40EWG | 0.15 | 0.26 | 14th | 0.27 | 0.31 | −0.24 |
| FTSE MIBEWI | 0.22 | 0.28 | 38th | 0.30 | 0.30 | −0.40 |
| FTSE 100EWU | 0.19 | 0.24 | 36th | 0.24 | 0.25 | −0.44 |
| ASX 200EWA | 0.06 | 0.17 | 1st | 0.26 | 0.21 | +0.16 |
The first three columns come from the R&D note "Stock correlation" (data to 24–25 September, percentile since 2021). The next two are computed month by month on the stocks, 2016–2025, with monthly rather than 30-session windows, so they compare with each other but not with the first columns. The last column relates, over ten years, the October correlation to the equal-weighted return of the same October.
What it says. For a given stock-level volatility, low correlation mechanically lowers index volatility, which fits the 21-day volatility being below its long-run average in all seven markets (scenario table). The data show no stable link with the direction of October: the sign is negative in 5 markets of 7 (down to −0.44 for the FTSE 100 and −0.40 for the FTSE MIB), meaning Octobers with more correlation tended to be worse Octobers, but with ten observations per market this is a hint, not a rule. Australia and the S&P 500 go the other way.
The economic calendar for October
Events that can move the seven markets, in date order. The tag shows how each date was checked.
- 29 SepRBA decisionbefore the monthA hike to 4.60% is expected by the major Australian banks. It shapes the start of the ASX 200's month.
- 2 OctUS jobs report for SeptemberBLS, checked8:30 ET. About 85,000 jobs expected, unemployment 4.1%.
- 13 OctJPMorgan opens third-quarter earnings seasonseveral sourcesResults before the open, call at 8:30 ET.
- 14 OctUS CPI for SeptemberBLS, checked8:30 ET. One search result said the 13th: the official BLS date is the 14th, and that is the one used. PPI on the 15th, import/export prices on the 16th.
- 21 OctUK inflation for September (ONS)several sources7:00. Relevant for the FTSE 100 ahead of the Bank of England on 5 November.
- 27–28 OctFed meeting, decision on the 28th at 14:00 ETseveral sourcesNo new projections or dot plot. Four participants had pointed to two more hikes by year-end, so a hike in October is possible.
- 29 OctECB decisionseveral sourcesStatement at 14:15 CET, press conference at 14:45. First decision after the September hike, the second since the war began.
- 29–30 OctBank of Japan, with the Outlook Report on the 30thseveral sourcesQuarterly meeting with new forecasts.
- 30 OctUS Employment Cost Index, third quarterBLS, checked8:30 ET.
- 2–5 NovJust after the monthafter OctoberRBA on 2–3 November, US midterm elections on the 3rd, Bank of England on the 5th. Neither the BoE nor the RBA meets in October.
Not found or not checked: specific German and Italian data dates, US third-quarter GDP and PCE. They are not in the calendar.
Scenarios for October on the seven markets
Three outcomes, each with the historical October return range of the index attached: down = average of the worst fifth of Octobers, middle = average of the middle 60%, up = average of the best fifth. The historical probability of each is therefore 20%, 60% and 20%. These are the only probabilities used.
Down
Rates and oil prevail.
- The Fed hikes again on the 28th or signals more
- US CPI on the 14th above expectations, strong jobs on 2 October
- 10-year Treasury holding above 5.25%, oil rising again
- Bank and tech earnings that do not absorb the cost of money
Middle
Markets looking for direction.
- Mixed data, yields on a plateau near their highs
- Fed and ECB as expected, no surprise on the path
- Earnings good enough to keep indices in a narrow range
- Low correlation: stock selection matters more than the market
Up
Relief on rates and oil.
- Weak jobs on 2 October and softer CPI reduce hike bets
- Hopes of Middle East progress push oil down
- 10-year yields retrace
- October closes in recovery, like the favourable half of history
Historical return ranges by market
| Index | Down | Middle | Up | Octobers ≤ −3% / ≥ +3% | 21-day volatility vs long-run |
|---|---|---|---|---|---|
| S&P 500SPY | −6.7% | +1.4% | +7.5% | 10% / 32% | 11% vs 18% |
| Nasdaq 100QQQ | −9.8% | +2.3% | +11.3% | 20% / 50% | 16% vs 26% |
| Euro Stoxx 50EZU | −8.1% | +1.0% | +8.7% | 32% / 32% | 13% vs 22% |
| DAX 40EWG | −7.2% | +2.3% | +10.4% | 18% / 45% | 13% vs 22% |
| FTSE MIBEWI | −7.8% | +2.5% | +8.6% | 21% / 46% | 14% vs 23% |
| FTSE 100EWU | −7.8% | +1.6% | +6.0% | 17% / 34% | 10% vs 17% |
| ASX 200EWA | −6.0% | +1.4% | +5.8% | 15% / 36% | 8% vs 15% |
Ranges computed on the whole available history of each index (same series as the seasonality table). 21-day volatility is annualised on daily index returns to 25 September. With current volatility below the long-run average, the historical ranges, which include 1987 and 2008, are wider than what the market prices today; they have not been rescaled.
Reading the markets against each other. The widest ranges are those of the Nasdaq 100 (−9.8% to +11.3%) and the DAX 40 (−7.2% to +10.4%), the narrowest those of the ASX 200 and the FTSE 100. On 18 September the European markets (FTSE 100 included) and the ASX 200 were already below their 50-day averages, while the S&P 500 and Nasdaq 100 were near their highs: a different starting point, not a forecast of which will hold up better.
How this analysis was done with AI
The analysis was carried out by Claude, an AI model from Anthropic (Sonnet 5), used inside Claude Code on the author's computer. The AI wrote and ran the scripts, searched the web and wrote the text. The questions, the direction and the decision to publish were the author's.
Step by step
- 1. Existing dataClaude looked in the TradingAlgo project for the seven markets and the daily prices of their stocks (a
prices_daily.csvfile per market). They started in January 2023: too short for seasonality. - 2. Ten years of historyClaude reused the project's download script (yfinance, adjusted prices) with a separate output folder, without touching production data. A first attempt with a different call had failed entirely; it was fixed by using the original script and its HTTP session.
- 3. Seasonality and days of the monthMonthly and daily returns computed in Python on the stocks. For the long history the seven indices were also downloaded from 1985.
- 4. Pairwise correlationReuse of the "Stock correlation" note (30 sessions) and a new month-by-month computation on the stocks to compare October with the other months.
- 5. Calendar and contextWeb search for dates and news. US release dates were checked on the official BLS page; the others against several concordant search results.
- 6. Scenarios and textThe three-scenario structure, the ranges (historical fifths) and the commentary are a synthesis written by the AI from the numbers above.
What the code computed
Every seasonality, correlation, quintile, frequency and volatility figure on this page comes from Python scripts run on downloaded data, and the page was generated from those results, without retyping by hand. The same calculation can be repeated with the same inputs.
What is the AI's judgment
The definitions (trading days, equal-weight market, fifths as scenario ranges), the three scenario narratives and the way calendar events are tied to markets. No probability was assigned beyond the historical 20–60–20 frequency.
What was not verified, and known limits
- Macro context from searches. Rates, the 5.23% yield and oil come from web-search summaries. One CNBC article could not be opened (error 403), so it was not read in full; these figures should be re-read at the source before use.
- A conflict between sources. The date of US CPI was the 13th in one result and the 14th on the BLS page. The BLS date stands.
- Survivorship bias. The stocks are today's members of each index taken back in time, and some tickers could not be downloaded. The index statistics do not have this problem, but their samples have different lengths.
- Small samples. Ten Octobers for the stocks, 19 for the Euro Stoxx 50. A year like 2022 or 2018 moves the averages.
- AI mistakes already found. In one daily chart the last point used too few years and made a line collapse artificially: it was spotted by looking at the result and corrected. Others may remain: the page was not reviewed by a second analyst.
- Not advice. Research material, not a recommendation or a forecast.
Sources
- BLS, release schedule for October 2026 (read)
- Federal Reserve, FOMC calendar
- CNBC, Fed decision of September 2026
- CNBC, outlook for the week of 28 September–2 October (search summary only)
- ECB, Governing Council calendar
- Bloomberg, ECB hike of 10 September
- Bank of England, September 2026 decision
- Bank of England, MPC dates 2026
- RBA, board meeting schedule
- Bank of Japan, meeting schedule
- ONS, UK inflation for September 2026
- JPMorgan, 2026 earnings call dates
- Prices: Yahoo Finance through yfinance. Correlations: TradingAlgo R&D notes "Stock correlation" and "October, day by day".
Research material only — not personalized financial advice. Past performance does not predict future results. Scenario ranges are historical averages, not forecasts. Contact: domderrico@gmail.com.