Bitcoin's worst month has missed three years running

Bitcoin enters September 2026 carrying a nickname it has not earned since 2022: "Rektember." The calendar says the month is the worst of the year — the last three Septembers say otherwise.

Is September Really Bitcoin's Worst Month?

Yes, on the historical record — but by a margin thinner than the nickname suggests, and one that has failed three years running. Since 2013, September carries the worst average and median return of any calendar month for Bitcoin, with 8 of 13 completed Septembers closing red, a 61.5% loss rate . The exact size of that edge depends entirely on whose dataset you use, and the spread between methods is wider than the effect itself.

Quick Answer: September is statistically Bitcoin's weakest month since 2013 — 8 of 13 closed red, with average returns cited between −2.97% and −4.02% depending on dataset. But 2023, 2024 and 2025 all closed green, so the September 2026 risk is the Sept 16 Fed decision, not the calendar.

Reported figures diverge by source and method. Decrypt, working from the CoinGlass monthly series, cites averages of −2.87%, −2.97% and −3.08% in circulation with a median near −2.44%; a separate DefiLlama-derived open-to-close study compiled by CoinJuice puts the average at −4.02% and the median at −4.72%, with a win rate of just 5 of 13 (38%) . The worst single instances remain 2014 (−18.59%) and 2019 (−16.36%) . Note the underlying data has no first-party issuer — no exchange, regulator or statistical agency publishes an official Bitcoin monthly seasonality series, so every number here traces to a private aggregator such as CoinGlass.

The streak matters more than the average. Bitcoin closed green in each of the last three Septembers — roughly +3.91% in 2023, +7.29% in 2024 and +5.16% in 2025 . Three consecutive misses out of a 13-observation sample is not a rounding error; it is close to a quarter of the entire dataset pointing the other way.

The 2026 setup sharpens the question. Bitcoin rallied roughly 24.95% in August 2026 — its strongest month since November 2024 and its best August since 2017 — then opened September near $77,500–$78,000, down about 1% on the first session . What follows examines whether the seasonal record deserves any weight in that decision, and what actually sits on the September 2026 calendar.

Bitcoin September Returns, Year by Year (2017-2025)

Bitcoin's September record since 2017 splits cleanly into two eras: five consecutive red Septembers from 2017 through 2021, then three consecutive green ones from 2023 through 2025. The red run printed −7.44% (2017), −5.58% (2018), −13.38% (2019), −7.51% (2020) and −7.03% (2021); the green run printed roughly +3.91% to +4.0% (2023), +7.25% to +7.29% (2024) and +5.16% to +5.36% (2025) . Across the full series, Bitcoin closed lower in 8 of 13 completed Septembers since 2013 .

Two prints anchor the extremes. The worst September on record is 2014 at −18.59%, followed by 2019 at −16.36% on the open-to-close series compiled from DefiLlama data — a figure that differs from the −13.38% widely quoted for the same year, which is exactly the kind of divergence readers should expect . The best September on record is 2024 at +7.25% to +7.29%. Depending on which tracker you use, the average September return lands anywhere between −2.87% and −4.02%, with the median between −2.44% and −4.72% .

The ranges in that table are not sloppiness — they are the honest state of the data. There is no regulator, exchange or statistical agency that publishes an official Bitcoin monthly return series. CoinGlass's monthly performance table is the dataset most traders cite , while open-to-close math derived from DefiLlama produces a materially more negative picture . Different exchange references, different UTC cutoffs and different open/close conventions move each year's figure by one to three percentage points. Anyone sizing a position off a single decimal place is over-reading the source.

What survives across every version of the dataset is the shape, not the precision:

• September is the only month with a clearly negative average and median on every tracker — it is the sole outlier, not one of several weak months.
• June is the nearest comparison, with an average near −1.59%, roughly half to a third of September's drawdown depending on the series.
• The red era ended in 2022. Every September since has closed green, and the three-year run of gains covers the entire spot-ETF era.
• Thirteen observations is a small sample, and the early years carry thin liquidity that inflates both tails.

Read as a ledger rather than a forecast, the year-by-year record shows a real historical tilt that has been contradicted for three consecutive years by the most liquid, most institutionally accessible Septembers in Bitcoin's history.

Why September Has a Bad Reputation — And Why the Sample Is Too Small to Trust

September's bad name in crypto is inherited, not invented. Equities carry the same scar: the S&P 500 has averaged roughly a 0.6% September decline since 1945, widening to between 1.1% and 1.2% on data going back to 1928, making it the only calendar month with a consistently negative long-run average . Bitcoin's 13-year September record inherited that narrative frame before it had enough observations to earn one, and the two datasets get cited together as if one confirms the other.

The standard explanations for the equity effect are structural and specific. Mutual funds with October 31 fiscal year-ends harvest tax losses into the autumn, forcing mechanical selling of the year's laggards. Institutional desks return from summer holidays at roughly the same time and re-examine risk books simultaneously, concentrating position changes into a narrow window. And the Federal Reserve's mid-September meeting concentrates macro volatility into the same fortnight — this year on September 15–16, 2026, a meeting that carries a Summary of Economic Projections and dot plot . Midterm election years sharpen the pattern further: historically those cycles have bottomed near September 2 and involved drawdowns on the order of 17% before recovering .

Note what those mechanisms have in common: none of them describe Bitcoin. Bitcoin has no fiscal year-end, no mutual-fund tax-loss regime, and a spot ETF complex that is barely two years old. The one channel that plausibly transmits — a mid-month Fed decision hitting a risk asset — is macro, and macro operates the same way in March or June. That is the case against treating the crypto version as a distinct calendar effect rather than a coincidence of overlapping narratives.

"More myth than math," — Ben Kurland, CEO of DYOR, on Bitcoin's September effect (source: Decrypt, 2026-09)

The statistical objection is the harder one to argue with. Even the source data disagrees with itself. Reported September averages for the CoinGlass series since 2013 circulate as −2.87%, −2.97% and −3.08% depending on method, with a median closer to −2.44%, while open-to-close math derived from DefiLlama and compiled by CoinJuice produces a −4.02% average and −4.72% median . A spread that wide across a single question — what does Bitcoin do in September — is a signature of a sample too small to be stable under different definitions.

Outliers do most of the work. With 13 completed observations, a single extreme month moves the mean materially:

• 2014: −18.59% — the single worst September on record, from a period of thin exchange liquidity and pre-institutional market structure .
• 2019: −16.36% — the second worst, and the other half of the pair that anchors the negative average .
• 8 of 13 down — the directional tilt is real, but a 62% down-rate across 13 trials is not far from what coin-flip variance produces by chance .

Strip out 2014 and 2019 and the remaining eleven Septembers cluster near flat. That is the practical test for any seasonal claim: if removing one or two observations reverses the conclusion, the conclusion was describing those observations, not the month. Bitcoin's September average fails that test, which is why the three-year green streak is better understood as the sample filling in than as a curse being broken.

The Three-Year Green Streak: What Broke the Curse in 2023-2025

The green streak is the single most important fact in the Rektember dataset: Bitcoin closed September higher in 2023, 2024 and 2025 — roughly +4.0%, then +7.25% to +7.29%, then +5.16% to +5.36% . That is the first three-in-a-row positive stretch in the 13-year series, and it arrived while the same series was still being quoted as evidence of a seasonal curse. Three consecutive contradictions out of thirteen observations is not noise at the margin; it is roughly a quarter of the entire sample voting the other way.

What the three years share is not a calendar effect but a flow catalyst. Each has a distinct, identifiable driver:

• September 2023 — pre-approval positioning. Traders front-ran the US spot Bitcoin ETF decision, and demand arrived ahead of the product rather than because of the month.
• September 2024 — post-election risk appetite. A macro and political repricing pulled capital into risk assets generally, with Bitcoin among the beneficiaries.
• September 2025 — sustained spot ETF demand. The vehicle that was speculative in 2023 was, by 2025, a standing bid that operated independently of seasonal sentiment.

The through-line matters more than any individual year. In all three cases the marginal buyer was structural — a fund launch, an allocation shift, a persistent creation flow — and none of those forces consult a calendar. That is the mechanical reason the seasonal edge decayed: as ETF and institutional flow became the dominant marginal input, the month-of-year variable lost whatever explanatory weight thin, retail-dominated liquidity had given it in 2014 and 2019. The same logic runs in reverse and is worth holding onto for 2026 — when flows turn negative, as they did with roughly $236 million of net US spot ETF outflows on September 1, 2026, led by BlackRock's IBIT , the calendar gets the credit for something the order book actually did.

There is also a shape inside the month worth separating from the direction of the month. Intra-month analysis compiled by CoinJuice from DefiLlama open-to-close data describes a repeatable profile even in red years: relative strength in the first week, the sharpest drawdown clustering roughly September 15–21, and a partial recovery in the final week . For 2026 that mid-month window overlaps almost exactly with the September 15–16 FOMC meeting — which is a reason to watch the date, not the season.

Finally, note how unstable the framing itself is. One count says Bitcoin closed lower in 8 of 13 completed Septembers since 2013; another framing of the same period yields a 5-of-13 win rate, and the average return quoted ranges from −2.87% and −2.97% to −3.08% on the CoinGlass series, against −4.02% average and −4.72% median from the DefiLlama-derived math . Those are not different markets. They are different start dates, different close conventions and different averaging methods applied to one small dataset. A pattern that changes size depending on which tracker you open is not a pattern you should size a position around.

What's Different in September 2026: A Near Coin-Flip Fed Hike

The variable that separates September 2026 from the three green Septembers before it is the direction of Fed risk: for the first time this cycle, the live question at the next meeting is a hike, not a cut. The FOMC held the federal funds target range at 3-1/2 to 3-3/4 percent on July 29, 2026, but did so on a 9–3 vote, with Beth M. Hammack, Neel Kashkari and Lorie K. Logan all dissenting in favor of a 25 basis point increase . A three-member dissent bloc pushing for tighter policy is the macro backdrop no September since 2022 has had to price.

Chair Kevin Warsh's Jackson Hole keynote on August 28, 2026 hardened that read. Warsh described the 2 percent PCE objective as a "firm, fixed target," argued that disinflation is "not self-executing, nor is inflation necessarily mean-reverting," and disclosed that the 12-month PCE change stands at 3.7 percent while the six-month change runs at 4.1 percent — the shorter window hotter than the longer one . The Bureau of Economic Analysis independently confirms the 3.7 percent headline reading for July 2026 . On breadth, Warsh noted 54 percent of the PCE basket showed 12-month price increases above 3 percent, and 49 percent on a six-month annualized basis, against a 32 percent pre-pandemic norm .

"We must be confident that underlying inflation is moving to our objective, clearly and at sufficient speed. Otherwise, we have work to do," — Kevin Warsh, Chair, Federal Reserve, Jackson Hole, August 28, 2026 (source: Federal Reserve, 2026-08).

He also declined to pre-commit, calling himself "committed to a discipline, not to a decision" . That is why the September 15–16 meeting, which carries a Summary of Economic Projections and dot plot, functions as a binary event rather than a formality .

Two accuracy notes belong with that table. First, the hike odds are second-hand: the CME FedWatch tool is the first-party surface, and probability estimates circulating in the mid-to-high 60s come from news reports of it, with prediction-market pricing running visibly lower. Second, widely repeated figures of a 4.784% 10-year and a 5.28% 30-year yield do not reconcile with the Treasury's own published daily par yield curve, which last printed the 10-year at 4.51% on August 26 . Use the Treasury series. The genuine rate stress is directional, not a crisis-level print.

Bitcoin entered the month priced for that uncertainty. After rising roughly 24–25% in August — its best month since November 2024 — it opened September near $77,500–$77,800 and slipped about 1% on the first session . US spot Bitcoin ETFs absorbed about $3.5 billion in August, their strongest month of 2026 with inflows on 16 of 21 sessions, then flipped to roughly $236 million of net outflows on September 1, led by BlackRock's IBIT . Wintermute OTC trader Jasper De Maere frames the decision window with two levels: $75,000 and $82,000 . The calendar is not what moves that range — the September 16 statement and dot plot are.

Decision Framework: Bull Case vs. Bear Case for Rektember 2026

The September 2026 decision reduces to a single question: does the liquidity that drove August's rally survive a Federal Reserve that may hike? The bull case rests on demonstrated flow — U.S. spot Bitcoin ETFs absorbed roughly $3.5 billion in August, their strongest month of 2026 , a step-change from July's negligible intake that pulled year-to-date net outflows back toward roughly $1.8 billion. The bear case rests on positioning: a coin-flip rate decision landing on a market that just ran about 24–25% in a month . Both are real. Neither is a calendar argument.

Bull case components:

• ETF demand is structural, not tactical. Inflows landed on 16 of 21 August sessions — breadth across sessions, not one or two block trades, which is the profile of allocation rather than momentum-chasing.
• An underdiscussed Treasury liquidity offset. The Treasury's buyback program raises its operational minimum starting September 9, adding a bid to the front of the plumbing in the same window as the FOMC meeting. It is a small mechanical support, but it runs counter to the "Fed tightens, liquidity drains" narrative traders will default to.
• The bid may not be a risk-on bid at all. Decrypt argues Bitcoin's recent strength looks like a debasement hedge — yields, gold and oil moving together — rather than a pure risk asset trade . If that framing is right, a hawkish Fed confirming a 3.7% PCE problem is not automatically bearish for Bitcoin. It changes which macro signals matter: watch real yields and the inflation-breadth data Chair Warsh emphasized, not the equity-beta correlation.

Bear case components:

• Crowded positioning. Short liquidations reportedly reached roughly $3 billion in 24 hours during the August squeeze — a signal that leverage was heavily one-sided going into September, and forced-flow moves cut both ways.
• The flow already turned. September 1 produced net ETF outflows, led by BlackRock's IBIT . One session is noise; a week of it is a trend break.
• A hike is the base case, not the tail. Three FOMC members already dissented in favor of a 25bp hike on July 29 , and Warsh explicitly declined to pre-commit to easing .

One contested signal deserves flagging rather than resolving. Early-September ETF flow reports do not agree: one tracker showed roughly +$142 million of net inflows on September 1, another about −$236 million of net outflows the same day. The most likely explanation is a trade-date versus settlement-date convention difference, not a data error — but the practical consequence is that a trader reading one dashboard and a counterparty reading another will disagree on whether demand broke. Before acting on a single-day flow print, reconcile it against both Farside Investors and CoinGlass, and prefer the multi-session trend over any one session.

The framework, then, is not bullish-or-bearish. It is: which of these two variables do you have an edge on? If your read is on flow durability, the ETF dashboards are your instrument and the September 16 decision is secondary. If your read is on the Fed reaction function, the FOMC calendar and the dot plot are your instrument and daily flows are noise. Holding both views at once with size behind them is how traders get caught on a 25% month's giveback.

Which Trader Profile Fits Which Approach?

The right response to Rektember depends entirely on your holding period and what you actually get paid for. A swing trader with a two-week horizon and a dollar-cost-averaging holder with a five-year horizon are looking at the same calendar and the same FOMC date, but only one of them has a decision to make in September 2026. Below, four common profiles and what the evidence supports for each — not a prediction, a mapping from time horizon to the signal that is actually informative for it.

Short-term and swing traders. Your exposure is to realized volatility, not to the monthly average. The September 16 FOMC decision falls inside the mid-month window where sharp drawdowns have historically clustered in red Septembers, and this meeting carries a Summary of Economic Projections and dot plot . That combination — a live rate decision plus a fresh projection path — is a volatility event on its own terms, independent of any seasonal belief. Wintermute OTC trader Jasper De Maere framed the levels around it as $75,000 and $82,000 . If you trade this window, size for gap risk across the announcement rather than positioning on the month's reputation.

Long-term holders and DCA buyers. The seasonal case is too weak to justify a tactical exit. You are being asked to act on 13 completed observations, an average that varies from −2.87% to −4.02% depending on whose method you use, and a pattern that has produced green closes three years running — roughly +4.0% in 2023, +7.25% in 2024 and +5.16% in 2025 . Exiting and re-entering costs spread, tax events and the risk of missing a month like August 2026, when Bitcoin gained roughly 24–25% . Continuing the schedule is the defensible default here.

Macro-hedge and risk-parity traders. Your signal set this month is the rate path, not the calendar. The FOMC held at 3-1/2 to 3-3/4 percent on July 29, 2026 with three dissents favoring a hike , and Chair Kevin Warsh disclosed 12-month PCE at 3.7 percent with the six-month change running hotter at 4.1 percent , a headline figure BEA independently confirms for July 2026 . Track Bitcoin against the Treasury curve — the 10-year printed 4.51% on August 26, 2026 — and against energy, given the supply shocks the Committee named. Note that widely circulated yield figures above 4.7% do not match Treasury's published table; use the official series.

Flow-driven and institutional-tracking traders. Daily ETF net flow is a current, verifiable input where a September average is neither. US spot Bitcoin ETFs absorbed about $3.5 billion in August 2026 across inflows on 16 of 21 sessions, then flipped to roughly $236 million of net outflows on September 1, led by BlackRock's IBIT . That is a same-week swing of real size. Watch the daily table rather than the monthly aggregate, and treat two or three consecutive outflow sessions as the signal, not one.

The concrete takeaway: match your instrument to your horizon. If you hold for years, do nothing differently in September. If you trade weeks, hedge the September 16 window explicitly rather than the month. If you trade macro, watch PCE breadth and the 10-year. If you trade flows, watch Farside's daily table . The one approach with no support in the data is trading the calendar itself.

Frequently asked questions

What is "Rektember" for Bitcoin?

"Rektember" is crypto-market slang that fuses "September" with "rekt," the trader shorthand for a heavy loss. It labels September's reputation as Bitcoin's statistically weakest calendar month. The basis is the CoinGlass monthly performance series, which shows Bitcoin closing lower in 8 of 13 completed Septembers since 2013 and posting the worst average and median return of any month . The nickname is descriptive, not predictive — it summarizes a backward-looking dataset rather than a mechanism.

What is Bitcoin's average September return?

There is no single agreed figure, because different providers use different price series and different open-to-close conventions. Values of −2.87%, −2.97% and −3.08% all circulate for the CoinGlass series since 2013, with a median near −2.44%, while DefiLlama-derived math compiled by CoinJuice gives −4.02% average and −4.72% median . Both datasets agree on direction — average and median negative — and both are dragged by two outliers, September 2014 (−18.59%) and September 2019 (−16.36%). Treat the range, not any one number, as the honest answer. The canonical table lives at CoinGlass.

Has Bitcoin's September curse held up recently?

No. Bitcoin closed green in each of the last three Septembers: roughly +4.0% in 2023, +7.25% to +7.29% in 2024, and +5.16% to +5.36% in 2025 . That is three consecutive years in which the most-cited seasonal pattern in crypto failed. A three-year streak does not disprove the long-run average, but it does mean the effect has produced no tradable edge for anyone who acted on it since 2022 — and it shrinks an already thin 13-observation sample to a signal that is not currently working.

Why is September 2026 different from prior Septembers?

September 2026 is the first Rektember of this cycle carrying genuine rate-hike risk rather than cut risk. At the July 28–29, 2026 meeting the FOMC held the federal funds target at 3-1/2 to 3-3/4 percent on a 9–3 vote, with Hammack, Kashkari and Logan dissenting in favor of a 25 basis point hike . On August 28, Chair Kevin Warsh told Jackson Hole the 2 percent PCE objective is a "firm, fixed target" and disclosed 12-month PCE at 3.7 percent with the six-month change running hotter at 4.1 percent . Secondary reporting of CME FedWatch put hike odds near 66–68% ahead of the September 15–16 meeting . That lands on a market that just gained roughly 24–25% in August.

Is the Bitcoin September effect statistically reliable?

Not on its own. Thirteen completed observations is a small sample, the early years carry thin liquidity, and two outlier months distort the simple average. DYOR CEO Ben Kurland calls Red September "more myth than math" . The pattern is also not crypto-specific: September is the only calendar month with a negative long-run S&P 500 average, around −0.6% since 1945 and −1.1% to −1.2% since 1928 . Shared drivers such as tax-loss selling and fiscal year-end flows explain more than any Bitcoin-specific curse. Use it as context; size positions on macro and flow data instead.

Enjoyed this article? Subscribe to get new stories by email whenever they're published.

Subscribe