Deep research on market seasonality — why the calendar effects exist, what the data actually shows, and how to use it.
The only month to average a net loss in every measured decade. 70 years of data, four structural reasons, and what to actually do about it.
The most famous phrase in seasonality tested against 75 years of data. The answer is more nuanced — and more useful — than the slogan suggests.
+21.4% average. Positive in 9 of 12 years. The mechanics behind Bitcoin's most reliable seasonal pattern — and why it keeps working.
If you have to choose one seasonal pattern to remember, it's this: September loses money. Not sometimes. Not in bad years. On average, consistently, across every decade of modern market history, in the US, UK, Germany and Japan simultaneously.
The S&P 500 averages -0.7% in September over 15 years — the only month to average a net loss. The NASDAQ is worse at -1.4%. The win rate — the percentage of Septembers where the market ends the month higher than it started — sits at just 42% for the S&P and 39% for the NASDAQ. You're more likely to lose money in September than not.
This isn't a recent phenomenon driven by a few bad years. Academic studies going back to the 1950s consistently identify September as the weakest month in US equities. The pattern pre-dates the internet, the Fed's current framework, algorithmic trading and every other structural change in markets over the last 70 years. It's not noise.
"September is the only month to average a net loss in the S&P 500 across every measured decade."
Many institutional investors — particularly US mutual funds — have fiscal years ending on October 31. That means September is their last full month to realise losses for tax purposes (tax-loss harvesting), lock in gains, and rebalance portfolios toward their year-end target allocations. The net effect is systematic selling pressure that shows up in the data every year.
August is already thin and low-volume as institutional traders take summer holidays. September is when they come back — and the first thing many portfolio managers do in September is reassess risk positions that were left running on autopilot through July and August. If those positions have run up, the temptation to lock in gains is highest in the first weeks of September.
Academic research by Ritter and Chopra has documented elevated mutual fund redemptions in September, linked to school-year financial planning (tuition payments, end-of-summer expenses) and the psychological tendency to review financial decisions at the start of autumn. Net outflows from equity funds create selling pressure that market makers must absorb.
Behavioural finance research has found correlations between seasonal affective patterns and risk appetite. Investors are measurably more risk-averse in autumn than spring, even controlling for fundamentals. The end of summer also tends to bring renewed focus on geopolitical risks, Federal Reserve meetings (September hosts one of the eight annual FOMC decisions), and the beginning of earnings pre-announcement season — all of which can add uncertainty.
What makes September seasonality particularly credible is that it isn't a US anomaly. The FTSE 100 averages roughly -0.6% in September. The DAX averages -0.7%. The Nikkei shows -0.7%. Four major markets in four different countries, with different regulatory environments, different investor bases and different fiscal years, all producing the same result. That level of consistency points to structural behavioural factors, not US-specific mechanics.
The average masks some genuinely severe individual Septembers. September 2022 delivered -9.3% for the S&P 500 — the bear market's worst single month, amplified by rate-hike panic. September 2001 saw the post-9/11 reopening plunge. September 2008 had Lehman. September 1998 had the LTCM crisis. September is consistently where bad things get worse.
The data doesn't say "sell everything on September 1st and buy back on October 1st." Seasonal patterns are tendencies, not certainties — and trading costs matter. What the data supports is a set of more modest adjustments:
The key is using this as one input among several, not a standalone trading rule. Combined with a read on macro conditions, valuations and trend, the September tendency gives you a meaningful calibration point that most retail investors don't have.
"Sell in May and go away, come back on St Leger's Day." It's the most quoted piece of market folklore in history. It's also — in its rough outline — supported by 75 years of data. The more interesting question is what exactly the data shows, and what it doesn't.
The academic version of the same observation is called the Halloween Indicator, named for the switching date of October 31. First documented in the American Economic Review in 2002 by Bouman and Jacobsen, it found that equity markets in 36 of 37 countries showed meaningfully higher returns in the November–April period than the May–October period. Not just the US. Not just recently. 36 of 37 countries.
Over the last 15 years (2009–2024), the S&P 500's average monthly return in the November–April window was approximately +1.2%, against +0.3% for May–October. That 0.9 percentage point difference per month compounds significantly over time. A portfolio that was fully invested November–April and in cash May–October would have outperformed buy-and-hold in 11 of 15 years, with materially lower maximum drawdowns during the weak period.
The "Best Six Months" strategy isn't a quirk of one good period. Looking at decade-by-decade data going back to the 1950s, the seasonal pattern holds in the 1950s, 1960s, 1970s, 1980s, 1990s, 2000s and 2010s. No decade has broken it decisively — though the 2020s introduced the complication of COVID and the 2021 bull market disrupting the May–October weakness.
"36 of 37 countries showed higher returns November–April. This is not a US anomaly."
The honest answer is that nobody has definitively proven the mechanism. The main competing theories are:
Here's what the slogan obscures: the May–October period isn't reliably negative. It's just less reliably positive. The average monthly return is +0.3%, not negative. In the 2020 COVID recovery, May–October 2020 was extraordinarily strong. In 2021, the "bad" summer months were positive. The seasonal tendency is a meaningful statistical edge, not a deterministic rule.
The practical version of the strategy isn't binary switching between stocks and cash. Most seasonal investors use it to tilt: overweight equities in the winter months, reduce exposure or move toward defensive sectors in summer, without fully exiting. That approach captures most of the seasonal edge without the whipsaw risk of missing a strong summer.
The Halloween Indicator is the original seasonal pattern — everything else on the TimingAX platform is a refinement of this basic observation applied to specific assets and time windows. November is the strongest individual month (+1.7% average, 70% win rate). September is the weakest (-0.7%, 42% win rate). That 2.4 percentage point monthly spread is the mechanical expression of the "Sell in May" effect at the individual month level.
Bitcoin averages +21.4% in October across 12 years of data. It has been positive in 9 of 12 Octobers. The win rate of 75% is higher than any other month for any major asset in the TimingAX database. In an asset class defined by volatility and chaos, that level of calendar consistency is remarkable enough to warrant serious analysis.
"Uptober" has been a fixture of crypto trading culture for years — but the cultural narrative and the actual data point in the same direction, which is unusual. This isn't manufactured hype. Something structural is happening in October that consistently benefits Bitcoin's price.
Bitcoin's halvings (which cut the issuance rate by 50%) have occurred in November 2012, July 2016, May 2020, and April 2024. The post-halving cycle typically sees the strongest price appreciation roughly 6–18 months after the halving. For multiple cycles, October has landed inside this accumulation window — when supply is tightening but the full demand response hasn't yet arrived. The 2024 halving (April) placed October 2024 at exactly the 6-month mark, which the data confirmed: Bitcoin peaked near $126,000 in October 2025, roughly 535 days post-halving.
Since the introduction of Bitcoin futures in December 2017 and the maturation of institutional crypto markets, October has become the month when institutional allocators build positions ahead of Q4. The logic mirrors traditional equity seasonality: fiscal year-end positioning, year-end performance chasing, and the beginning of the strongest seasonal window for risk assets all point to accumulation in October.
Bitcoin's September is almost as bad as its October is good: averaging -5.8% with a 33% win rate, September is Bitcoin's worst month. The pattern of September weakness followed by October strength creates a mechanical setup: sellers exhausted by September, buyers stepping in as the seasonal turn becomes apparent to experienced participants. This pattern has been self-reinforcing — as more people know about Uptober, the early positioning anticipating it creates the very effect it predicts.
"September is Bitcoin's worst month at -5.8%. October is its best at +21.4%. That 27-point swing in a single calendar turn is the largest seasonal gap of any major asset."
Of Bitcoin's 12 measured Octobers, three were negative: 2014, 2018, and 2019. All three occurred during sustained bear markets where macro and structural forces overwhelmed seasonal tendencies. The 2018 case is instructive — October 2018 was negative despite strong seasonals, as the broader crypto bear market (which would see Bitcoin fall 84% peak-to-trough) dominated every other factor.
This is the key qualifier: seasonal patterns are tendencies that hold when other forces are neutral or aligned. When a bear market is in full force, no seasonal pattern reliably reverses it. The 75% win rate already prices in three failures over 12 years — the three years where the macro backdrop was too negative for seasonality to overcome.
The October effect is strongest in Bitcoin, but adjacent assets show similar patterns. Ethereum averages +15.2% in October. Coinbase, as a proxy for crypto sentiment, shows a similar Q4 buildup. The November–December window extends the strength: November is Bitcoin's second-strongest month at +18.6% average, as the Uptober momentum tends to continue into the Singles Day period and year-end crypto positioning.
The weakest stretch for crypto is June–September — particularly June (-8.7%) and September (-5.8%). Investors who understood this calendar and used it to time entries around October have systematically positioned better than those ignoring it entirely.