Bitcoin price on September 15, 2026: BTC closed at $75,584, down 3.31%, in a $3,355 range between a $78,243 high and a $74,888 low — the largest single-day decline since June 5, on the heaviest volume since September 3, on an afternoon when the Senate’s CLARITY Act cloture vote failed and the close finished $96 above the lower Bollinger Band with the Federal Reserve’s decision one session away. The snapshot and charts are first, the short version follows, and the full explanations are at the end for anyone who wants them.
Informational only. This brief reports market data and publicly available news. It is not investment, trading, or financial advice, and it does not recommend buying, selling, or holding anything. The author is not responsible for any decision made using this information. Numbers are as of the 00:00 UTC daily close on September 15, 2026 (8 p.m. New York time) unless a different time is stated.
New to the indicators? Every name below links to the plain-English explanations in Bitcoin Indicators Explained. Yesterday’s edition: September 14, 2026. About this series: Bitcoin Daily Brief.
Bitcoin price on September 15, 2026, at a glance
The charts
The 30-second read
Bitcoin closed Tuesday at $75,584.17, down $2,590.84 or 3.31%, in a $3,355.26 range — 4.439% of the close and 1.52 times Monday’s average true range — on 10,155 BTC of volume, 1.72 times the twenty-day average and the heaviest session since September 3. Every near-term layer gave way at once: RSI fell 9.63 points to 48.15 and closed below 50 for the first time since August 16, the MACD histogram widened again to −787.06 and is now its most negative since June 9, %B collapsed from 0.44 to 0.02 with the close just $96.26 above the lower Bollinger Band, and the squeeze that had held since September 9 broke open with band width up 0.72 points — while the long frame held, a twenty-eighth consecutive close above the 200-day average, still 7.61% above it. What the numbers do not contain is Tuesday’s ETF print, which had not been published when this brief was written, or the Federal Reserve decision at 2 p.m. Eastern on Wednesday that Yahoo Finance quoted at a 92.5% chance of a rate increase — the first in three years.
What happened in the last 24 hours
- Price. Opened $78,175.00, ranged between $74,887.50 and $78,242.76 and closed $75,584.17 (−3.31%, −$2,590.84) — a $3,355.26 span on volume of 10,155.20 BTC, 1.72 times the twenty-day average; the body measured $2,590.83, or 77.2% of the range, with a $67.76 upper wick (the thin line above the candle body marking where price traded but did not close) and a $696.67 lower wick, the close finishing 20.8% of the way up from the low. It is the fifteenth-largest single-day decline of 2026’s 258 daily returns and the largest since June 5.
- Flows. Monday’s US spot ETF figure arrived at +$159.9 million, ending a four-session outflow run of −$462.7 million and taking September’s nine reported sessions to +$467.2 million against August’s completed +$3,524 million. Tuesday’s figure — the session that would actually cover the decline — had not been reported when this brief was written. Separately, CryptoSlate described “Bitcoin ETF outflows exceeding $2 billion over two weeks,” a characterisation that does not reconcile with Farside Investors’ daily table and is reported here as their wording rather than adopted.
- Positioning. CryptoSlate, citing CoinGlass, put twenty-four-hour crypto liquidations at $941 million across 161,200 traders — $378 million in Bitcoin contracts and $255 million in Ether, roughly $870 million of it long against $71.4 million short, with the single largest entry a $32.4 million Bitcoin swap on Bitget. A separate account put the total at $771 million with $568.5 million long. CoinStats’ September 15 page, captured earlier in the UTC day and before most of the decline, showed funding at +0.0031% per four hours (annualized there at about 6.82%), open interest of $52.61–$52.66 billion and only $41.13 million of liquidations split 81% short — a snapshot that the afternoon overtook.
- On-chain. CoinWarz put difficulty at 127.45T at block 967,206 with 474 blocks left in the cycle and now estimates the September 19 retarget at +5.01%, to about 133.84T — revised sharply up from the +3.13% it projected a day earlier, reversing three days of downward revisions. The same source marked network hashrate near 986 EH/s, down 1.6% on the day but up 17.01% over seven days, and average block time at 9 minutes 33 seconds against the ten-minute target. On the corporate side, Decrypt reported Strive’s holdings reaching an even 25,000 BTC after a $36.6 million purchase funded entirely through preferred stock.
- Macro and policy. The Senate’s cloture vote on the CLARITY Act failed to reach the 60 votes needed, and Cointelegraph reported Coinbase and Circle shares falling about 10% alongside miners and treasury companies. US equities closed lower on TheStreet’s marks — the S&P 500 at 7,585.73 (−0.45%), the Nasdaq Composite at 25,981.57 (−0.78%) and the Dow at 52,093.11 (−0.63%) — with the ten-year Treasury yield at 5.006% intraday, a fifty-two-week high, and the thirty-year at 5.369%. WTI crude rose 2.64% to $104.10 and gold futures eased 0.73% to $4,320.10. Yahoo Finance put the odds of a 25 basis-point increase at Wednesday’s FOMC at 92.5%, against 86.5% a day earlier and 69.4% on Friday.
Section summaries
Price and trend
Tuesday gave back Monday’s rebound and a great deal more. The $3,355.26 range was 4.439% of the close, the seventh-widest of the twenty-eight sessions since price reclaimed the 200-day average, and the $2,590.84 decline is the fifteenth-largest of 2026’s 258 daily returns and the largest since June 5. The close of $75,584.17 is the lowest since August 20 and the lowest of September, sitting at the very bottom of the month’s closing range. The long frame held and the near frame did not: a twenty-eighth consecutive close above the 200-day average, with the golden-cross gap widening an eighth straight session to $1,413.10 — but a seventh consecutive close below the 20-day, and this time $2,704.94 below it against $286.20 on Monday. The cushion above the 200-day narrowed $2,639.46 to $5,345.15, the largest single-session contraction of this run. Details ↓
Momentum and volatility
Monday’s momentum repair lasted exactly one session. RSI fell 9.6283 points to 48.15 — the fifth-largest single-day fall of 2026 and the largest since August 28 — closing below 50 for the first time since August 16. The MACD histogram, which had narrowed once after ten widenings, widened again by $91.99 to −787.06, its most negative reading since June 9, and this time the MACD line did the work, falling $288.75 against the signal line’s $196.77. ADX eased an eighth session to 41.88 and the directional spread, which had widened to 18.38 on Monday, closed back to 11.90 as +DI fell 3.33 and −DI rose 3.14. Volatility rose across all three measures for a second day: ATR $81.98 higher at $2,289.55, band width 0.7208 points higher at 7.156% — the largest one-day expansion since August 27 — and realized volatility 1.99 points higher at 49.64%, its highest since March 31. Details ↓
Flows and positioning
The pending number arrived and it was positive: Monday’s US spot ETF flow printed at +$159.9 million, ending the four-session outflow run of −$462.7 million and lifting September’s nine reported sessions to +$467.2 million. That figure describes Monday, not Tuesday, and Tuesday’s print — the one that would say whether the fund channel participated in the sell-off — had not been published when this brief was written. Derivatives told the clearer story: CryptoSlate, citing CoinGlass, put twenty-four-hour liquidations at $941 million across the crypto complex with roughly $870 million of it long, against a CoinStats snapshot taken earlier in the day that showed only $41.13 million and an 81% short skew. The two are not contradictory so much as differently timed, and the gap between them is a fair measure of how much of the day happened after the New York lunch hour. Details ↓
On-chain and network
The difficulty projection reversed direction. After three consecutive downward revisions — +4.25%, then +3.7%, then +3.13% — CoinWarz now estimates the September 19 retarget at +5.01%, to about 133.84T, on 1,542 of the cycle’s 2,016 blocks with 474 remaining. A convergent sequence that jumps on the penultimate day is a reminder that these are extrapolations from partial block counts, not measurements; the retarget itself will settle it. The same source marked hashrate near 986 EH/s, down 1.6% on the day and up 17.01% over seven days, at 68.4% of the 1.44 ZH/s peak recorded on September 20, 2025. Decrypt reported Strive’s stack reaching an even 25,000 BTC after a $36.6 million purchase. No on-chain valuation metrics — MVRV, exchange balances, SOPR — were sourced on this run, a fifth consecutive day. Details ↓
Sentiment
alternative.me’s index dated September 15 reads 69 (“Greed”), up twelve points from Monday’s 57 and the largest single-day rise of the eight-day series this brief has assembled — on a day price fell 3.31%. The timing explains most of it: the API’s own countdown places its publication at roughly 05:23 UTC, which is before the Senate vote and before most of the decline, so the reading describes the first quarter of the session rather than the whole of it. The Alpha Vantage news feed carried 18 Bitcoin-tagged articles, scored one bullish, five somewhat bullish, ten neutral, one somewhat bearish and one bearish. Published views split hard: CryptoQuant’s Julio Moreno said spot demand is “contracting at the fastest pace since Jan. 10,” and Alphractal’s Joao Wedson noted Bitcoin’s annualized Sharpe ratio turning negative, while CryptoQuant’s own short-term-holder work, reported by Cointelegraph the same morning, had those holders approaching a full month in partial profit and called it a bull signal. Details ↓
Macro and cross-asset
Two things hit the same market on the same afternoon. The Senate failed to reach the sixty votes needed to end debate on the CLARITY Act, and Cointelegraph reported Coinbase and Circle down about 10% with miners and treasury companies following. Behind that sat a global bond move: the US ten-year yield touched 5.006% on TheStreet’s marks, a fifty-two-week high, with Cointelegraph quoting 5.041% and calling it the highest since June 2007, the G7 average ten-year at 4.285%, the UK thirty-year at 5.95% and the Japanese ten-year at 3.04%. US equities closed lower across the board, WTI rose 2.64% to $104.10 and gold eased 0.73%. Rate expectations firmed again into Wednesday’s decision, with Yahoo Finance quoting 92.5% odds of a 25 basis-point increase against 86.5% a day earlier. Details ↓
Calendar
Wednesday September 16: Tuesday’s ETF flow is published, August retail sales land at 8:30 a.m. ET, then the Fed decision at 2 p.m., a press conference at 2:30 and a fresh Summary of Economic Projections. Thursday September 17: the Bank of England at 7 a.m., then US jobless claims at 8:30, where Coinpedia’s calendar carries a 209,000 estimate. Friday September 18: the Bank of Japan at 3 a.m., which Coinpedia’s calendar expects to raise to 1.25%, Japan’s August CPI, the September options expiry and industrial production at 9:15 a.m. Around September 19 at 03:57 UTC, the difficulty retarget now estimated at +5.01%. Details ↓
Reading the combination
The two frames separated again, and further than at any point in this run. The long frame is intact — a twenty-eighth close above the 200-day, a 50-day rising for a thirty-fifth session, a golden-cross gap out to $1,413.10 — while the near frame broke on every measure at once: RSI through 50, the MACD histogram to its most negative since June, %B at 0.02 with the close $96.26 off the lower band, and the shortfall to the 20-day average back out to $2,704.94. The historical record splits on a seam that is worth naming rather than averaging: the filters describing what happened to volatility read well (band width expanding more than 0.7 points, higher ten sessions later 66% of the time on 981 instances; all three volatility measures rising together, 65% on 612), the filters describing where price landed read close to baseline (%B under 0.05, 60% on 151), and the filter describing what momentum did reads worst (RSI crossing below 50, 49% on 159 against a 58% above-200-day baseline). None of them is strong, and the one scheduled event that will most likely settle the week has not happened yet. The indicators do not say which way this resolves. Details ↓
Learn the indicators: every reading above is explained, with charts and the combinations traders watch, in Bitcoin Indicators Explained.
The details, for readers who want more
Signal board
| Indicator | Reading (Sep 15 close) | Zone | Change since Sep 14 | What this zone has historically meant |
|---|---|---|---|---|
| Bollinger Bands (20, 2) | %B 0.02; band width 7.16% | At the lower band; the squeeze broke open | %B 0.44 → 0.02; width 6.44% → 7.16% | %B, which locates the close between the bands with 0 at the lower band and 1 at the upper, fell from 0.4433 to 0.0172 — the largest single-day fall of the twenty-seven day-over-day changes since price reclaimed the 200-day average, and the lowest reading since June 25. The close finished just $96.26 above a lower band at $75,487.91, and the session low of $74,887.50 pierced that band by $600.41 before the recovery into the close. This is the first day of 2026 on which price closed with %B below 0.05 while holding above its 200-day average; the previous instance in this series was September 26, 2025. Band width, which had spent six sessions creeping off a September 9 trough of 5.574%, jumped 0.7208 points to 7.1560% — the largest single-day expansion since August 27 and a fourth consecutive rise. The mechanism is the reverse of Monday’s: the standard deviation of the last twenty closes rose $138.30, from $1,262.30 to $1,400.60, because a $75,584 close is far enough from the middle of that window to stretch it. The envelope widened $553.22 to $5,602.40, and it widened asymmetrically — the upper band rose $104.51 while the lower band fell $448.71, which is what a squeeze breaking downward looks like in the arithmetic. Of 151 earlier days in this series with %B below 0.05 while price held above its 200-day average, the close was higher ten sessions later 60% of the time (median +0.99%) and twenty sessions later 54% (median +0.81%). Of 981 earlier days on which band width expanded more than 0.7 points in a session above the 200-day, the figures are 66% (median +3.79%) and 63% (median +6.14%), with the standing caveat that such days cluster in a handful of volatile stretches, so this is not 981 independent observations. |
| Price vs 20-day SMA | −3.46% below ($78,289) | Below, seventh consecutive close — and by the widest margin of the run | −0.36% → −3.46% | The gap widened from $286.20 to $2,704.94, and almost all of it came from price: the close fell $2,590.84 while the average fell $172.10. At $2,704.94 the 20-day average is 1.18 times one day’s ATR away, against 0.13 times on Monday — which is to say it has moved from within a routine session’s reach to something more than a full session’s range above the market. The 20-day itself fell for a fourth session, and faster than on any of the previous three, because the close entering its window was far below the one leaving it: Tuesday’s $75,584.17 replaced August 26’s $79,026.18, a shortfall of $3,442.01 spread over twenty days. The window now runs from August 27 to September 15, and August 27’s $80,275.34 leaves next — $4,691.17 above Tuesday’s close — so on an unchanged close the average would fall about $235, roughly a third faster than it just did. Of 68 earlier days in this series that were exactly the seventh consecutive close below the 20-day while price held above its 200-day, which is precisely today’s condition, the close was higher ten sessions later 60% of the time (median +2.27%) and twenty sessions later 54% (median +1.13%). Widening to every instance of seven or more consecutive closes below the 20-day above the 200-day — 603 earlier days — gives 55% (median +0.68%) and 53% (median +0.69%). The two point mildly different ways, the 68-observation version is small enough that a handful of clustered instances could move it several points, and neither is far enough from the 58% above-200-day baseline to carry weight. |
| 50-day vs 200-day SMA | $71,652 vs $70,239 (gap +$1,413.10) | Golden cross — 50-day above, eighth day | Gap +$1,224 → +$1,413 | The gap widened for an eighth consecutive session since the cross printed, by $189.18, as the 50-day rose $237.79 to $71,652.12 — a thirty-fifth straight daily rise — against the 200-day’s $48.62 to $70,239.02. The post-cross sequence now runs $88.65, $272.36, $436.77, $617.00, $814.62, $1,020.67, $1,223.92 and $1,413.10. A golden cross is the name given to a shorter moving average crossing above a longer one; it is a statement about the two averages, not about the day. That it widened on a session when price fell 3.31% is the whole point of the mechanism: the 50-day window is still shedding the low closes of late July, which sit far below anything the 200-day window is shedding, so the shorter average gains ground whichever way price goes — it gained $203.26 on Monday while price rose $1,375, and $189.18 on Tuesday while price fell $2,591. The base rates are struck on the cross date rather than on subsequent days and are therefore unchanged: of the twelve earlier golden crosses this series records, beginning with February 9, 2012, price was higher thirty sessions later in six of twelve cases (median −2.62%), sixty sessions later in eight of twelve (median +12.60%) and ninety sessions later in eight of twelve (median +10.67%). Twelve observations is a sample in which one instance moves the median by several points, and that is how it should be read. |
| Price vs 200-day SMA | $75,584 vs $70,239 (+7.61%) | Above (28th straight close, since Aug 19) | +11.38% → +7.61% | Price above the 200-day average is the most common definition of a long-term uptrend, and this is a twenty-eighth consecutive close above it. The distance narrowed 3.766 percentage points and the cushion shrank $2,639.46 to $5,345.15 — the largest single-session contraction of the twenty-seven day-over-day changes inside this twenty-eight-session run, ahead of August 28’s $2,474.84. The composition is almost entirely price: the close took away $2,590.84 and the rising average absorbed a further $48.62, so 98% of the narrowing came from the market and the remainder from the line continuing to climb underneath it. It would now take a 7.07% decline from Tuesday’s close to reach the line, against 10.21% on Monday. The 200-day has risen on each of the last twenty-five sessions, though its rate of increase slowed to $48.62 from $53.42. A long average still rising while the distance to it compresses by a third in one session describes a trend that is intact in definition but has given up a large part of its margin, and both halves of that sentence are load-bearing. |
| RSI (14) | 48.15 | Lower half of the neutral zone (30–70), first close under 50 since Aug 16 | 57.78 → 48.15 | Down 9.6283 points, the fifth-largest single-day fall of 2026 and the largest since August 28, and enough to erase Monday’s 4.6153-point gain twice over. RSI compares the average size of up-days with down-days over fourteen sessions, so a −3.31% close after a run of small moves registers heavily; the reading is now below 50 for the first time since August 16, below the 52.77 low of September 10, and 38 points below the 86 it printed on August 21. Of 378 earlier days in this series with RSI between 46 and 50 and price above its 200-day average, the close was higher ten sessions later 50% of the time (median −0.05%) and twenty sessions later 54% (median +0.81%) — eight points under the 58% above-200-day baseline at ten sessions. The more specific filters disagree with each other, and the disagreement is instructive. Of 159 earlier days on which RSI crossed from at-or-above 50 to below it while price held above its 200-day, the close was higher ten sessions later 49% of the time (median −0.29%) and twenty sessions later 50% (median −0.04%) — the weakest pair on this board. But of 155 earlier days on which RSI simply fell more than nine points in that same setting, the close was higher 63% of the time at ten sessions (median +2.01%) and 62% at twenty (median +3.91%) — the strongest. Today is both. One filter selects for a level being lost, the other for a violent move that has already happened, and the record says the second has been the better predictor without saying why. |
| MACD (12, 26, 9) | 1,217 vs signal 2,004; histogram −787.06 | Below signal for a twelfth session; most negative since June 9 | Histogram −695.07 → −787.06 (widening) | Monday’s single narrowing was reversed in one session. The histogram sequence since the September 4 crossover now reads −49.50, −108.96, −139.48, −261.93, −396.55, −500.09, −678.69, −739.02, −759.80, −785.98, −695.07 and −787.06, so the ten consecutive widenings and one narrowing are followed by a widening of $91.99 that takes the reading one dollar past its previous extreme and to its most negative since June 9, 2026, when it printed −925.67. What matters is which line moved. On Monday the histogram narrowed only because the signal line fell more than twice as fast as the MACD line — convergence from behind. On Tuesday the MACD line fell $288.75 to 1,216.77 while the signal line fell $196.77 to 2,003.83, so the faster line resumed leading, which is the ordinary shape of momentum still draining rather than a smoothing artefact. The MACD line remains 1,217 above zero, so this continues to describe upward momentum unwinding rather than a change of trend direction, but that buffer has now shrunk 19.2% in a single day and 65% since September 3. Of 634 earlier days in this series with a negative and widening histogram, the MACD line still above zero and price above its 200-day average, the close was higher ten sessions later 55% of the time (median +0.80%) and twenty sessions later 53% (median +0.80%) — three points under the baseline at ten sessions, which is close enough to it to say very little. The broader condition — histogram negative and MACD line above zero, regardless of direction — gives 1,124 instances at 50% and a median of exactly 0.00%. |
| ADX (14) | 41.88 (+DI 28.5, −DI 16.6) | Strongly trending (above 25), falling for an eighth day | 43.08 → 41.88 | An eighth consecutive fall, and at 1.195 points the second-largest of the eight, against Monday’s 0.196. ADX measures the strength of a trend without saying which way it points, and it is computed from a smoothed average of the directional indices, so it lags a turn in those indices by several sessions. The internals gave back Monday’s move almost exactly: +DI fell from 31.8575 to 28.5227 and −DI rose from 13.4819 to 16.6248, closing the directional spread from 18.3756 to 11.8980. That 6.478-point contraction is a fraction larger than Monday’s 5.674-point expansion, so two sessions have left the spread narrower than it was on Sunday. A $3,355.26 range whose low extended $1,459.77 below Monday’s is exactly the low-to-low movement −DI is built from. A falling ADX with a narrowing positive spread is the signature of a trend losing definition rather than reversing — the index is still well above the 25 threshold that marks a trending market, and +DI is still comfortably above −DI. Days in this series with ADX between 41 and 45, falling, with +DI above −DI and price above the 200-day — 68 earlier instances with a full ten-session forward record — were followed by a higher close ten sessions later 59% of the time (median +1.09%) and twenty sessions later 54% (median +2.24%). Sixty-eight observations is a thin sample, and it sits close enough to the 58% baseline to be indistinguishable from it. |
| ATR (14) | $2,290 (3.0% of price); range $3,355.26, lower wick $696.67 | Rising for a second day, highest since Sep 11 | $2,208 → $2,290 | Up $81.98, a second consecutive rise and the highest reading since September 11. The cause is a high-to-low range of $3,355.26, which is 1.52 times Monday’s ATR and 4.439% of the closing price, ranking fifty-fourth widest of 2026’s 258 sessions on that measure and seventh widest of the twenty-eight-session run above the 200-day. ATR is a fourteen-day Wilder average, so a session whose true range exceeds the running average lifts it by roughly a fourteenth of the difference, which is what happened here. The candle’s shape inverts Monday’s: the body was $2,590.83, or 77.22% of the range — a decisive, one-directional session — with a $67.76 upper wick and a $696.67 lower wick, and the close sat only 20.76% of the way up from the low. Where Monday’s long upper wick said a high was reached and rejected, Tuesday’s short upper wick and long lower wick say the session opened at its high and spent the day going down, with a partial bounce off $74,887.50 into the close. At 3.03% of price, ATR is back above the 3% threshold for the first time since September 11. Of 401 earlier days in this series on which price fell 3% or more while holding above its 200-day average, the close was higher ten sessions later 57% of the time (median +1.51%) and twenty sessions later 54% (median +1.74%) — within a point of the 58% baseline at ten sessions, which is one more reading on this board that says almost nothing on its own. |
| 30-day realized volatility | 49.6% annualized | Elevated; the highest since March 31 | 47.65% → 49.64% | Up 1.9906 points, by far the largest move in this measure in more than a week — the previous four daily changes were −0.083, −0.008, −0.039 and +0.105, none of them a fifth of a point. Realized volatility measures how much daily returns have varied over thirty days, so a −3.31% session sits far enough into the tail of that thirty-day window — only the three August 19–21 sessions, at +7.14%, +5.36% and +7.28%, are larger in absolute terms — to move the measure materially where a 1.79% session did not. At 49.64% the reading is the highest since March 31, 2026 and comfortably above anything printed between April 15 and September 8. This is the second consecutive session on which all three volatility measures moved the same way, and the first on which all three moved appreciably: ATR, which sees the intraday range, rose 3.7%; band width, computed from the standard deviation of twenty closes, rose 11.2%; realized volatility, computed from thirty closes, rose 4.2%. When the three agree in both direction and magnitude, they are usually describing the same underlying event rather than three different ones, and here that event is a single wide down-day. Of 352 earlier days in this series with realized volatility between 47% and 53% and price above its 200-day average, the close was higher ten sessions later 53% of the time (median +0.66%) and twenty sessions later 50% (median +0.25%). |
| Volume vs 20-day average | 10,155 BTC vs 5,912 average (1.72×) | Well above the average — heaviest session since September 3 | 6,547 → 10,155 BTC | Up 55.1% from Monday and 1.7178 times the twenty-day average — the twenty-sixth highest ratio of 2026’s 258 sessions, the highest since August 21, and in absolute terms the heaviest single session since September 3. The twenty-day average itself rose from 5,728.92 to 5,911.81, so the ratio improved despite a rising denominator, which is the cleaner version of this reading. For scale, Tuesday traded 10,155.20 BTC against 10,435.01 BTC for Saturday, Sunday and Monday combined — one session within 3% of the previous three put together. A falling price on heavily above-average volume is the configuration that conventionally counts as conviction on the sell side, and it is a different thing from the low-volume drift that characterised most of the preceding week. Of 333 earlier days in this series with volume above 1.5 times its twenty-day average while price held above the 200-day, the close was higher ten sessions later 61% of the time (median +2.34%) and twenty sessions later 60% (median +2.79%) — three points above the above-200-day baseline at ten sessions, and notably, that filter does not condition on direction. Narrowing it to days that combined a fall of 3% or more with volume above 1.5 times average leaves 64 earlier instances, higher ten sessions later 56% of the time (median +1.54%) but only 48% at twenty sessions (median −0.26%). Sixty-four observations is thin, and the two horizons pointing opposite ways on so small a sample is a reason to treat it lightly rather than a finding. |
| US spot ETF net flows | Sep 14 +$159.9M (reported); Sep 15 not yet reported when this brief was written; September ≈ +$467.2M; August +$3,524M | One inflow day ended a four-day outflow run; the session that matters is pending | Latest print −$13.2M → +$159.9M | The number this brief flagged as pending yesterday arrived, and it was an inflow: Farside Investors’ table now carries September 14 at +$159.9 million, ending a four-session redemption run that had totalled −$462.7 million and taking September’s nine reported sessions to +$467.2 million against August’s completed +$3,524 million. Read carefully, that print covers Monday — the day price rose 1.79% — and says nothing about Tuesday. The September 15 row had not appeared when this brief was finished, which is the ordinary settlement lag rather than an anomaly, and it is the single most consequential pending figure in this edition: it is the only direct read on whether the fund channel participated in a 3.31% decline or sat it out. One discrepancy is worth flagging rather than resolving. CryptoSlate’s Tuesday piece described “Bitcoin ETF outflows exceeding $2 billion over two weeks,” calling it one of the fastest two-week exits on record. That characterisation cannot be reconciled with the daily table this brief tracks: the ten reported sessions from August 31 through September 14 sum to roughly +$684 million, and the largest cumulative drawdown inside September is the −$462.7 million four-session run that Monday’s print ended. Either a different universe of funds, a different window, or a gross-redemption rather than net-flow convention is in play. Both figures are reported here; only the Farside series is used for the chart and the running totals. |
| Perpetual funding rate | +0.0031% per 4 hours (CoinStats, pre-decline snapshot) | Mildly positive; longs paying shorts | Not comparable — the interval changed between the two pages | Funding is the periodic payment between holders of perpetual futures that keeps the contract tethered to spot; positive means longs pay shorts, which indicates net long positioning. CoinStats’ September 15 page put it at +0.0031% and, this time, described the interval as four hours, annualizing at about 6.82%. Yesterday’s page described the same measure as +0.0059% per eight hours, annualizing at about 6.42%. The two annualized figures are close — 6.82% against 6.42% — but the raw numbers are on different conventions, so no day-over-day comparison of the raw rate is drawn here, and the four-to-eight-hour ambiguity flagged twice in the previous week has reopened rather than settled. What can be said is that on either convention the level remains far below the crowded-long readings that have preceded sharp unwinds in this series, which have generally run several times higher. The larger caveat on this row is timing: the page was captured before the bulk of Tuesday’s decline, as its own price field confirms — it showed $78,186.86 and a +1.78% daily change, both of which belong to Monday’s close. Funding after a 3.31% fall and $870 million of long liquidations is very likely a different number, and this brief does not have it. |
| Open interest / liquidations | Liquidations $941M, ~92% long (CryptoSlate/CoinGlass); OI $52.6B (CoinStats, pre-decline) | A large, one-sided long flush | Liquidations $16.9M → $941M across the crypto complex | Liquidations are forced closures of leveraged positions by an exchange when margin runs out. The day produced the largest such event in this brief’s short series: CryptoSlate, citing CoinGlass, put twenty-four-hour crypto liquidations at $941 million across more than 161,200 traders, of which roughly $870 million was long against $71.4 million short — about 92% one-sided — with $378 million in Bitcoin contracts, $255 million in Ether, and a single $32.4 million Bitcoin swap on Bitget as the largest entry. A separate account of the same day put the total at $771 million with $568.5 million long, so the two sources differ by about a fifth on the total and rather more on the split; both are reported and neither is adopted as the figure. Against those, CoinStats’ page showed only $41.13 million with an 81% short skew and open interest of $52.61–$52.66 billion — a snapshot taken before the Senate vote, as its stale price field shows, and therefore describing the early hours rather than the session. TheStreet separately noted Bitcoin open interest down about 3.5% over seven days. The useful reading is not any single number but the shape they agree on: leverage that had been net long was removed involuntarily and in size, which is the deleveraging pattern rather than the distribution one, and it is the mechanism most likely to explain why a $2,591 decline happened in a few hours rather than over several days. |
| Fear & Greed | 69 (“Greed”) | Greed, upper part of the band | 57 → 69, but published before most of the decline | alternative.me’s dated API returned data for a fifth consecutive run, extending the series to eight days: 69 on September 8, 66 on September 9, 69 on September 10, 56 on September 11, 63 on September 12, 61 on September 13, 57 on September 14 and 69 on September 15. Tuesday’s twelve-point rise is the largest single-day move of the series in either direction, and it came on a day price fell 3.31% — which looks like a contradiction and is mostly a clock problem. The API’s time_until_update field, read at 00:25 UTC on September 16, placed the next refresh roughly five hours later, which puts the publication of each dated reading at about 05:23 UTC. The entry dated September 15 therefore describes conditions early in that UTC day — before the Senate’s afternoon vote in Washington and before most of the decline — and not the session as a whole. That is worth stating plainly because a reader comparing 69 against a −3.31% close would otherwise conclude the index is broken; it is simply stamped with a date it only partly covers. The same caveat applied in the other direction on September 14, when the index fell four points on a day price rose. The reading remains inside the “Greed” band, which begins at 55, and the September 16 entry — the first to cover the sell-off in full — was not available when this brief was written. |
| Hashrate / difficulty | Difficulty 127.45T; next retarget estimated Sep 19, +5.01% to ~133.84T; hashrate ~986 EH/s | Between retargets, 474 blocks from the adjustment | Estimate +3.13% → +5.01% | CoinWarz shows difficulty unchanged at 127.45T — it changes only every 2,016 blocks — with the chain at block height 967,206 and 474 blocks remaining in the current cycle, and now estimates the next retarget for September 19 at 03:57 UTC at +5.01%, taking difficulty to about 133.84T. That reverses three consecutive downward revisions: +4.25% on 1,019 observed blocks, +3.7% on 1,169, +3.13% on 1,379, and now +5.01% on roughly 1,542. Yesterday’s edition described that sequence as converging in the way an extrapolation from a partial block count is supposed to; Tuesday’s figure is a reminder that convergence is not monotone, because the estimate is driven by the block intervals of the most recent stretch and a fast run of blocks late in a cycle moves it sharply. The supporting datum is consistent: the same source marks average block time at 9 minutes 33 seconds against the ten-minute protocol target, which is what an upward adjustment is. None of the projections is a measurement; the September 19 retarget, computed on 2,016 actual blocks, will be. Spot hashrate is quotable this run for once — CoinWarz marked it near 986 EH/s at block 967,196, down 1.6% on the day but up 17.01% over seven days, and 68.4% of the 1.44 ZH/s peak recorded on September 20, 2025 — with the standing caveat that hashrate is inferred from block intervals rather than measured, so short-window estimates scatter by roughly a tenth across the same moment. |
Price and trend (chart 1)
Tuesday opened at its high and closed near its low. Bitcoin opened at $78,175.00 — Monday’s close, as the daily candles here are continuous — reached $78,242.76 within the first hours, then fell to $74,887.50 and closed at $75,584.17, a loss of $2,590.84 or 3.3142%. The range was $3,355.26, which is 4.439% of the closing price and 1.52 times Monday’s ATR of $2,207.57. That ranks fifty-fourth widest of 2026’s 258 sessions by percentage of the close and seventh widest of the twenty-eight sessions since price reclaimed the 200-day average. The decline itself is the fifteenth-largest of the year’s 258 daily returns and the largest since June 5’s −4.35%. The body — the distance between open and close — was $2,590.83, or 77.22% of the range, with a $67.76 upper wick and a $696.67 lower wick, and the close finished 20.76% of the way up from the low. That shape is the mirror image of Monday’s: a session with almost no upper wick, a body filling three-quarters of the range and a modest bounce off the low is a one-directional day with a late stabilisation, not a fight. Volume was 10,155.20 BTC against a twenty-day average of 5,911.81, a ratio of 1.718 — the heaviest session since September 3, and within 3% of the combined volume of Saturday, Sunday and Monday, which together traded 10,435.01 BTC.
The level structure below the market is where the day stopped, and it stopped at the band. The $74,887.50 low pierced the lower Bollinger Band at $75,487.91 by $600.41 before the close recovered to $75,584.17, finishing $96.26 above it — the closest a close has finished above the lower band since February 3, 2026, when it ended $19.97 clear of it, and the first close of this year with %B below 0.05 while price held above its 200-day average. Round numbers were taken out on the way: $76,000 and $75,000 both gave way intraday, and several outlets led with the second. Feeds disagree about how far it went, which is normal and worth stating: this brief’s 00:00 UTC Alpha Vantage candle marks the low at $74,887.50, Cointelegraph reported a September low of $75,560 on its own feed at 16:02 UTC, and CryptoSlate later cited $74,255. The close of $75,584.17 is the lowest since August 20’s $73,011.87 and the lowest close of September, which places it at the very bottom of the month’s closing range between itself and the $81,263.99 high close of September 3. Against longer reference points the close sits 40.15% below the all-time high of $126,296, 13.62% below the 2025 year-end close of $87,497.94, and 29.15% above the June 30 low close of $58,523.93 — or, stated the other way round, that June low is 22.57% below Tuesday’s close. September is now 3.79% lower across its first fifteen sessions, a $2,978.57 decline from August 31’s $78,562.74. This record contains fifteen completed Septembers from 2011 to 2025; nine finished negative, and the month’s average return of −3.8% makes it the weakest of the twelve on that sample, though the last three Septembers closed +4.0%, +7.3% and +5.4%.
The averages diverged from the price and from each other. The 50-day rose $237.79 to $71,652.12 — a thirty-fifth consecutive daily rise — and the 200-day rose $48.62 to $70,239.02, taking the golden-cross gap from $1,223.92 to $1,413.10, a widening of $189.18 and the eighth consecutive session of widening since the cross printed on September 8. The 20-day, meanwhile, fell $172.10 to $78,289.11, its fourth consecutive decline and the largest of the four. Bitcoin closed 7.61% above the 200-day average, 5.49% above the 50-day, 12.04% above the 100-day at $67,459.82 and 3.46% below the 20-day — a twenty-eighth consecutive close above the 200-day in the run that began on August 19, and a seventh below the 20-day. The cushion above the 200-day is the reading that moved most and the one most worth watching: it shrank $2,639.46 to $5,345.15, the largest single-session contraction of the twenty-seven day-over-day changes inside this run, ahead of August 28’s $2,474.84. Its composition is almost purely price — the close took away $2,590.84 and the average’s own $48.62 climb removed the rest — and it means the decline required to reach the line has compressed from 10.21% to 7.07% in one session. The near frame moved the other way with equal force: the shortfall to the 20-day widened from $286.20 to $2,704.94, of which the close supplies $2,590.84 and the average’s own $172.10 decline hands $172.10 back, for a net $2,418.74. At $2,704.94 the 20-day sits 1.18 times one day’s ATR above the market, against 0.13 times on Monday.
Levels being discussed (attributed)
| Level | Why it is being watched | Who cites it |
|---|---|---|
| $88,700 | Cited as the upper boundary of the trader realized price band | CryptoQuant via Parameter.io (Sep 14) |
| $82,000 | Upper bound of the resistance zone described as running from $80,000 | Social-media analyst summary via CoinStats (Sep 15) |
| $81,700 | Cited as the 365-day moving average and the threshold that would confirm a fresh bull cycle | Julio Moreno, CryptoQuant, via Parameter.io (Sep 14) |
| $81,264 | The September 3 close — September’s highest close, 7.51% above Tuesday’s | Alpha Vantage feed |
| $81,090 | Upper Bollinger Band, 7.28% above the close and $105 higher than yesterday | This brief’s calculation |
| $81,087 | 365-day simple moving average, computed here; $613 below the figure CryptoQuant cites for the same measure | This brief’s calculation |
| $80,000 | Lower bound of the named resistance zone, and the round number chart 1 marks with a dashed line | Social-media analyst summary via CoinStats (Sep 15) |
| $79,635 | 50-week simple moving average, computed here on completed weekly closes through September 13 | This brief’s calculation |
| $78,289 | 20-day simple moving average, $2,705 above the close and falling for a fourth session | This brief’s calculation |
| $78,243 | Tuesday’s high on the 00:00 UTC candle — reached in the session’s first hours | Alpha Vantage feed |
| $78,175 | Monday’s close and Tuesday’s open | Alpha Vantage feed |
| $77,421 | The 10-period exponential moving average, computed here | This brief’s calculation |
| $77,364 | 50-week exponential moving average, computed here on completed weekly closes | This brief’s calculation |
| $77,000 | Upper bound of the support band described as running to $76,000 | Social-media analyst summary via CoinStats (Sep 15) |
| $76,800 | The September 13 close — the last completed weekly close | Alpha Vantage feed |
| $76,000 | Lower bound of the named support band; broken intraday on Tuesday | Social-media analyst summary via CoinStats (Sep 15) |
| $75,584 | Tuesday’s close — September’s lowest, and the lowest since August 20 | Alpha Vantage feed |
| $75,560 | Cited as the September low on Cointelegraph’s feed at 16:02 UTC, before the later decline | William Suberg, Cointelegraph (Sep 15) |
| $75,488 | Lower Bollinger Band, $96 below the close and $449 lower than yesterday | This brief’s calculation |
| $75,000 | The round number broken during the session; several outlets led with it | KuCoin news flash (Sep 15) |
| $74,888 | Tuesday’s low on the 00:00 UTC candle — $600 below the lower Bollinger Band | Alpha Vantage feed |
| $74,255 | Cited as the session low on CryptoSlate’s feed — $633 below this brief’s candle low, a feed difference rather than a dispute | CryptoSlate (Sep 15) |
| $73,056 | 200-day exponential moving average, computed here | This brief’s calculation |
| $71,722 | 21-week exponential moving average, computed here on completed weekly closes | This brief’s calculation |
| $71,652 / $70,239 | 50-day and 200-day simple moving averages — the two lines that crossed on September 8, now $1,413 apart | This brief’s calculation |
| $70,000 | Cited as support, described as the 200-day moving average | Parameter.io (Sep 14) |
| $69,300 | The August 19 close — the lowest close of the twenty-eight-session run above the 200-day | Alpha Vantage feed |
| $58,524 | The June 30 low close of 2026 — 22.6% below Tuesday’s close, which sits 29.2% above it | Alpha Vantage feed |
One note on the table. Of these twenty-eight entries, seven carry an attribution to published coverage dated September 15, three carry one dated September 14 and are marked as such, ten are this brief’s own calculations, and eight come straight off the price feed. Those four counts sum to twenty-eight. The structural feature of Tuesday’s version is that the cluster has moved from above the market to beneath it. On Monday four separate reference points sat inside $554 above the close; on Tuesday the three nearest levels in either direction are all below — the lower Bollinger Band at $75,487.91, $96.26 under the close; the $75,000 round number the session broke and reclaimed; and the $74,887.50 low itself. Above the market the nearest reference is now the 10-period exponential average at $77,420.77, $1,836.60 away, with nothing else until $78,175.00. That is an unusually empty $1,800, and it is the arithmetic consequence of a session that moved through its own recent range in a few hours rather than trading within it. Stripping out the long-horizon projections, the working band named in this week’s published analysis runs from $70,000 to $88,700, which is $18,700 wide, or about 8.2 times Tuesday’s ATR of $2,289.55. The Bollinger envelope widened to $75,487.91–$81,090.31, a $5,602.40 span against $5,049.18 a day ago — an expansion of $553.22, and an asymmetric one, with the upper band rising $104.51 while the lower fell $448.71.
Momentum and volatility (charts 2 and 3)
Both momentum measures reversed Monday’s improvement, and one of them did so in a way that matters more than the other. RSI fell 9.6283 points to 48.1537, the fifth-largest single-day decline of 2026 and the largest since August 28, taking the reading below 50 for the first time since August 16 — twenty sessions ago — and below the 52.77 low of September 10 that had marked the floor of the recent range. A fourteen-day relative strength index at 48 is not an extreme; it is the middle of the neutral band, three points below the midpoint, and 18 points above the 30 that conventionally marks oversold. What changed is the side of the midpoint, and the record treats that as a different question from the size of the move, which is why this brief reports both filters below rather than choosing. The MACD histogram widened $91.99 to −787.06, a twelfth consecutive session below the signal line and the most negative reading since June 9, 2026. The mechanism is the reverse of Monday’s and is the more ordinary of the two: the MACD line fell $288.75 to 1,216.77 while the signal line fell $196.77 to 2,003.83, so the faster line is again leading the slower one down. Monday’s narrowing came entirely from the signal line catching down; Tuesday’s widening comes from the MACD line resuming its fall, which is what momentum draining looks like without the smoothing artefact. The MACD line is still 1,216.77 above zero, so this remains a description of an upward impulse unwinding rather than a trend reversing — but that buffer has shrunk 19.2% in one session and roughly 65% since September 3, and the zero line is the level at which the description would change.
ADX eased an eighth consecutive session, to 41.8843 from 43.0790, and the 1.195-point fall is the second-largest of the eight — the sequence now runs 0.431, 0.315, 1.244, 0.692, 0.642, 1.048, 0.196 and 1.195. The internals returned almost exactly to where they stood on Sunday: +DI fell from 31.8575 to 28.5227, a 3.33-point loss, while −DI rose from 13.4819 to 16.6248, a 3.14-point gain, closing the directional spread from 18.3756 to 11.8980. That 6.478-point contraction slightly exceeds Monday’s 5.674-point expansion, so the two sessions together have left the spread narrower than it was before either. The arithmetic is not mysterious: −DI is built from how much each session’s low undercuts the last, and a low $1,459.77 beneath Monday’s supplies a great deal of it. ADX is a smoothed average of the difference between the two directional indices and therefore lags them; at 41.88 it remains far above the 25 threshold that marks a trending market, and +DI remains 11.90 above −DI, so what the pair describes is a trend that is losing definition rather than one that has changed sides. Volatility, for a second consecutive session, moved the same way on all three measures — and this time appreciably on all three. ATR rose $81.98 to $2,289.55, its highest since September 11, because the $3,355.26 range exceeded the running average and a fourteen-day Wilder average lifts by roughly a fourteenth of the excess. Band width rose 0.7208 points to 7.1560%, its largest single-day expansion since August 27 and a fourth consecutive rise, because the standard deviation of the last twenty closes jumped $138.30 to $1,400.60 — Monday’s move had shifted only $1.66. Thirty-day realized volatility rose 1.9906 points to 49.6366%, the highest since March 31. The squeeze that had held since September 9’s 5.574% trough has broken, and it broke downward.
The base rates for this session are unusually scattered, and the scatter is the finding. Three filters sit clearly above the 58% baseline for days above the 200-day average. Of 981 earlier days on which band width expanded more than 0.7 points in a session above the 200-day, the close was higher ten sessions later 66% of the time (median +3.79%) and twenty sessions later 63% (median +6.14%). Of 612 earlier days on which ATR, band width and realized volatility all rose together in that setting, 65% and 65% (medians +3.58% and +5.86%). Of 333 with volume above 1.5 times its twenty-day average, 61% and 60%. Two sit close to baseline: 401 days with a fall of 3% or more give 57% and 54%, and 151 days with %B below 0.05 give 60% and 54%. And two sit below it: 378 days with RSI between 46 and 50 give 50% and 54%, while 159 days on which RSI crossed from at-or-above 50 to below give 49% and 50% — the weakest pair on the board. Set against that last figure, 155 earlier days on which RSI fell more than nine points in a single session above the 200-day were higher ten sessions later 63% of the time, median +2.01%, and 62% at twenty. Tuesday satisfies both. The honest resolution is the same one this brief drew on Monday from the opposite side: these are different questions. A filter keyed to a level being lost selects for the market having changed state; a filter keyed to the size of a move selects for something violent having already happened, and violent days above a rising 200-day average have historically been followed more often by continuation of the longer trend than by its end. Neither is strong enough to overrule the other, both are computed from the same fourteen-year record, and the sample sizes are similar. Narrowing to the intersection — a fall of 3% or more on volume above 1.5 times average — leaves 64 instances at 56% ten sessions out and 48% at twenty, which is too few and too internally inconsistent to carry anything.
Flows and positioning (chart 4)
Chart 4 has a new bar for the first time in four days, and it is green. Farside Investors’ table now carries September 14 at +$159.9 million, which ends the four-session redemption run that had totalled −$462.7 million and takes September’s nine reported sessions to +$467.2 million against August’s completed +$3,524 million. The reading that matters, though, is the one the chart cannot show: that print covers Monday, the session on which price rose 1.79%, and the September 15 row had not appeared when this brief was finished. Until it does, there is no direct evidence on whether the US spot ETF channel sold into a 3.31% decline, sat it out, or bought it — and that single figure will say more about the durability of this move than any indicator in this edition. One characterisation in Tuesday’s coverage needs flagging rather than resolving. CryptoSlate, in the piece that supplied this brief’s liquidation figures, described “Bitcoin ETF outflows exceeding $2 billion over two weeks” and called it one of the fastest two-week exits on record. That does not reconcile with the daily table tracked here: the ten reported sessions from August 31 through September 14 sum to roughly +$684 million net, and the largest cumulative drawdown anywhere inside September is the −$462.7 million four-session run that Monday’s print ended. A different fund universe, a different two-week window, or a gross-redemption rather than net-flow convention would each explain the gap. This brief reports both and uses only the Farside series for the chart and the running totals.
The derivatives picture is the clearest part of Tuesday and the part where sources agree on shape while disagreeing on size. CryptoSlate, citing CoinGlass, put twenty-four-hour liquidations across the crypto complex at $941 million affecting more than 161,200 traders — $378 million in Bitcoin contracts, $255 million in Ether, roughly $870 million long against $71.4 million short, with the largest single entry a $32.4 million Bitcoin swap on Bitget. A separate account of the same session put the total at $771 million with $568.5 million long. The two differ by about a fifth on the total and by more on the long share, which is ordinary for liquidation data — venues covered and the exact twenty-four-hour window both vary — and neither is adopted here as the figure. What they agree on is that the flush was large and overwhelmingly one-sided against longs, which is the deleveraging pattern: positions closed involuntarily by margin engines rather than sold by their owners. That is also the most plausible mechanism for the day’s shape, a $2,591 decline concentrated into a few hours with a $696.67 lower wick and a partial recovery into the close. Against this, CoinStats’ September 15 page showed $41.13 million of liquidations with an 81% short skew, open interest of $52.61–$52.66 billion and funding at +0.0031% per four hours; its price field read $78,186.86 and +1.78%, both of which belong to Monday’s close, which dates the snapshot to before the Senate vote. TheStreet separately noted Bitcoin open interest down about 3.5% over seven days, and this brief has no post-decline open-interest reading at all — the figure that would show whether the $941 million of forced closures reduced total positioning or was replaced. On the wider market, tv-hub.org put Bitcoin’s share of total crypto value at 58.4% on September 15 against 58.9% from the same source a day earlier, with Ether at 11.4% and total market capitalization at $2.67 trillion.
On-chain and network
The difficulty projection reversed, and the reversal is a better lesson than the convergence this brief described yesterday. CoinWarz shows difficulty at 127.45T — unchanged, as it must be between retargets — at block height 967,206 with 474 blocks remaining in the current 2,016-block cycle, and now projects the September 19 retarget at 03:57 UTC at +5.01%, taking difficulty to about 133.84T. The four-day sequence of projections reads +4.25% on 1,019 observed blocks, +3.7% on 1,169, +3.13% on 1,379 and now +5.01% on roughly 1,542. Monday’s edition noted that each successive figure was smaller and rested on more evidence, which is how an extrapolation from a partial count is supposed to behave; Tuesday’s figure shows that the behaviour is not guaranteed, because what the estimate actually tracks is the pace of the most recent blocks, and a fast stretch late in a cycle moves it a long way. The supporting datum points the same direction: average block time is running at 9 minutes 33 seconds against the ten-minute protocol target, which is precisely the condition an upward adjustment corrects. None of the four projections is a measurement. The retarget itself will be, computed on 2,016 actual blocks, and at +5.01% it would be the largest single adjustment since this brief began tracking the series, against +1.31% on September 5. Spot hashrate is quotable this run, which it has not been for most of the week: CoinWarz marked the network near 986 EH/s at block 967,196, down 1.6% on the day but up 17.01% over seven days, and at 68.4% of the all-time peak of 1.44 ZH/s set on September 20, 2025. The standing caveat still applies — hashrate is inferred from block intervals rather than measured, and short-window estimates from the same sources scatter by roughly a tenth across the same moment — so the seven-day direction is worth more than the level.
Corporate treasury news was thin and, unusually for a day like this, positive. Decrypt reported that Strive’s Bitcoin holdings reached an even 25,000 BTC after a $36.6 million purchase, funded entirely through preferred stock, with the notional value of its SATA issuance passing $1 billion for the first time. That is the same programme this brief described on Monday and the same holdings total; the purchase disclosed Tuesday completes the round number rather than adding a new leg. No corresponding disclosure arrived from Strategy, which has now gone two reported weeks without buying. A separate item worth a line for reasons that have nothing to do with price: the Bank for International Settlements published a study finding that widely used crypto on-chain metrics can substantially misstate underlying economic activity, with measurement gaps spanning Bitcoin, Ethereum and stablecoins. Readers of this series will recognise the problem from the opposite end — this brief has spent five consecutive days unable to source MVRV, exchange balances or SOPR from any free-tier provider with a current date, and the BIS paper is a reminder that the difficulty is not only availability. MVRV compares market value with the aggregate cost basis of coins last moved; exchange balances proxy the supply immediately available to sell; SOPR measures whether coins moving on-chain are being spent at a profit or a loss. None of the three is reported here, and on a day with $941 million of forced liquidations, the absence of a cost-basis reading is a real gap rather than a formality.
Sentiment
alternative.me’s dated API returned data for a fifth consecutive run, giving this brief a continuous eight-day series: 69 on September 8, 66 on September 9, 69 on September 10, 56 on September 11, 63 on September 12, 61 on September 13, 57 on September 14 and 69 on September 15. Tuesday’s twelve-point rise is the largest single-day move of the series in either direction, and it landed on a day price fell 3.31% — a combination that looks absurd until the publication clock is checked. The API’s time_until_update field, read at 00:25 UTC on September 16, showed 17,867 seconds to the next refresh, which places each dated reading’s publication at roughly 05:23 UTC. The entry stamped September 15 therefore describes the first quarter of that UTC day, before the Senate’s afternoon vote in Washington and before the bulk of the decline. It is not wrong; it is stamped with a date it only partly covers, and the same caveat cut the other way on September 14, when the index fell four points on a day price rose. Readers tracking this series day to day should treat each reading as describing the small hours of its date rather than the session that bears its name. The September 16 entry, the first that will cover Tuesday’s decline in full, was not available when this brief was written. CoinStats’ separately constructed index was not quotable this run, its page having been captured before the move.
Of the 18 Bitcoin-tagged articles in the Alpha Vantage news feed published between 00:00 UTC on September 15 and 04:00 UTC on September 16, the feed’s sentiment model scored one as bullish, five as somewhat bullish, ten as neutral, one as somewhat bearish and one as bearish. That is a narrower tally than Monday’s 20 and a flatter one, with three-quarters of the coverage in the neutral or somewhat-bullish bands on a day the asset fell more than 3% — which is what happens when the news cycle is dominated by a legislative event whose market consequence is contested rather than by the price itself. The single bearish score went to Cointelegraph’s report on Coinbase and Circle shares falling about 10% after the vote; the single somewhat-bearish score to its news write-up of the vote itself. The bullish score went to a Cointelegraph piece on short-term-holder profitability, and four of the five somewhat-bullish scores went to items with no connection to Tuesday’s session — a Motley Fool assessment of Michael Saylor’s twenty-year growth forecast, Decrypt’s report on Strive’s purchase, and two general market pieces. Eighteen articles is a workable sample for tone and a poor one for direction, and a tally that leans positive on a sharply negative day is a reason to distrust the tally as a market read rather than a reason to distrust the day.
Published views, in both directions, and they were unusually far apart. On the cautious side, CryptoQuant’s head of research Julio Moreno — the same analyst this brief quoted constructively on Monday — told CryptoSlate that “Bitcoin spot demand is contracting at the fastest pace since Jan. 10,” framing the decline as a demand problem rather than a leverage accident. Alphractal chief executive Joao Wedson noted that Bitcoin’s annualized Sharpe ratio, which measures return per unit of volatility, had turned negative, indicating what he described as elevated pressure and poor return efficiency. The Kobeissi Letter, quoted by Cointelegraph, framed the macro backdrop bluntly: “Monetary policy is shifting, rate hikes are returning, and the next battle against inflation has started.” On the constructive side, CryptoQuant’s own short-term-holder work, reported by Cointelegraph’s William Suberg on Tuesday morning, had short-term holders approaching a full month in partial profit — a condition that analysis associates with improving odds of a durable upward reversal, and one that is not obviously compatible with the same firm’s demand-contraction read published hours later. QCP Capital, writing before the vote, noted that passage would have clarified the respective regulatory roles of the SEC and the CFTC but cautioned that a procedural vote does not determine a final outcome. Senator Cynthia Lummis, the bill’s most prominent Republican advocate, called it “now or never” on the morning of the vote, saying the final text conceded more than 120 Democratic demands including new ethics rules; Senator Elizabeth Warren described the revised provisions as a “weak fig leaf.” Polymarket’s odds of the CLARITY Act passing in 2026 stood at 14% before the vote, per Cointelegraph, and at least one account described the bill as likely dead for the year afterwards.
Macro and cross-asset
Two distinct pressures landed on the same afternoon, and separating them is the useful exercise. The first is legislative. The Senate’s cloture vote on the CLARITY Act — the procedural vote to end debate, which requires sixty senators — failed. CNBC reported the Senate voting to block the bill from advancing; Decrypt named Gillibrand, Warner, Booker, Warnock, Gallego, Alsobrooks and Cortez Masto among Democrats who voted no after months of negotiation, and reported that Bitcoin fell as the no tally mounted while describing the move as stopping well short of panic. Cointelegraph reported Coinbase and Circle shares each down about 10%, with Bitcoin miners and treasury companies also declining — an equity reaction several times larger than Bitcoin’s own, which is the ordinary pattern when a regulatory event bears on US-listed intermediaries more directly than on the asset. The second pressure is the bond market, and it was running before the vote. The US ten-year Treasury yield touched 5.006% on TheStreet’s intraday marks, a fifty-two-week high; Cointelegraph quoted 5.041% and called it the highest since June 2007, noting the G7 average ten-year yield at 4.285%, the highest since mid-2008, the UK thirty-year at 5.95%, its highest since March 1998, and the Japanese ten-year at 3.04%, a thirty-year high. The US thirty-year was marked at 5.369%. US equities closed lower across the board — the S&P 500 at 7,585.73 (−0.45%), the Nasdaq Composite at 25,981.57 (−0.78%) and the Dow at 52,093.11 (−0.63%) — so this was a broad risk-off session in which Bitcoin fell roughly four times as much as the index, not a crypto-specific dislocation. WTI crude rose 2.64% to $104.10 on continuing Middle East supply concerns, and gold futures eased 0.73% to $4,320.10.
The rate decision is now one session away and the expectation has hardened. Yahoo Finance put the odds of a 25 basis-point increase at Wednesday’s FOMC at 92.5%, against 86.5% a day earlier and 69.4% on Friday; Coinpedia’s calendar carried roughly 87% and CoinStats about 85%, so the quoted spread this run runs from 85% to 92.5%, narrower than Monday’s 85%-to-93% scatter but still wide enough that this brief quotes all three rather than choosing. Decrypt reported that nearly every major bank now expects the Fed to raise rates for the first time in three years, and noted that markets have largely priced it while the political fallout may not be. The current target range is 3.50%–3.75%, so a hike takes it to 3.75%–4.00%. The Summary of Economic Projections published alongside the decision carries the committee’s own rate path into 2027 and matters as much as the decision itself; the press conference follows at 2:30 p.m., and Coinpedia flagged that traders will parse Chair Kevin Warsh’s remarks for whether further increases are contemplated. What remains genuinely unusual, and worth restating for readers who joined this series recently, is the direction: the base case priced by futures is a rate increase, the opposite of the easing cycle most of this market’s 2026 commentary was written around, and a ten-year Treasury yield at 5% is the direct competitor to an asset that yields nothing. Cross-asset, tv-hub.org put Ether’s dominance at 11.4% against Bitcoin’s 58.4% on September 15, with total crypto market capitalization at $2.67 trillion — a half-point fall in Bitcoin’s share on a day Bitcoin fell, which implies the rest of the complex fell rather less.
Calendar: what’s ahead
| Date (ET) | Event | Why Bitcoin traders watch it |
|---|---|---|
| Wed, Sep 16 | Tuesday’s US spot ETF flow is published | The only direct read on whether the fund channel participated in a 3.31% decline. Monday’s print was +$159.9 million, which ended a four-session outflow run of −$462.7 million |
| Wed, Sep 16, 8:30 a.m. | August retail sales and import/export prices | Lands the morning of the decision; the consumption side of the same inflation question, and the last US data the committee sees before it votes |
| Wed, Sep 16, 2:00 p.m. | FOMC decision; press conference 2:30 p.m., with a Summary of Economic Projections | Quoted odds of a 25 basis-point increase ran from about 85% (CoinStats) to 87% (Coinpedia) to 92.5% (Yahoo Finance), against 69.4% on Friday; the target range is 3.50%–3.75%. The projections matter as much as the decision, because they carry the committee’s own path for rates into 2027, and Coinpedia flags Chair Kevin Warsh’s remarks on whether more increases follow |
| Thu, Sep 17, 7:00 a.m. | Bank of England rate decision | The first read on whether other central banks follow the Fed’s direction or diverge from it |
| Thu, Sep 17, 8:30 a.m. | Weekly US jobless claims; August housing starts and building permits; Philadelphia Fed index | The day after the Fed, and the first check on whether the labour and housing data support the path the committee has just published. Coinpedia’s calendar carries an initial-claims estimate of 209,000 |
| Fri, Sep 18, 3:00 a.m. | Bank of Japan rate decision and Japan’s August CPI | Coinpedia anticipates a 25 basis-point increase to 1.25%, with core inflation around 1.8%–2.0%. It matters for the yen carry trade — borrowing in a low-rate currency to buy higher-yielding assets elsewhere — which unwinds when Japanese rates rise, and Japanese ten-year yields are already at a thirty-year high |
| Fri, Sep 18 | September options expiry; US industrial production at 9:15 a.m. | The expiry’s notional size was not yet reported when this brief was written |
| ~Sat, Sep 19, 03:57 UTC | Next Bitcoin difficulty adjustment | CoinWarz now estimates +5.01%, to about 133.84T, with 474 blocks remaining — revised sharply up from +3.13% a day earlier after three days of downward revisions. The retarget itself is a measurement on 2,016 actual blocks, so it settles the week’s shifting projections |
| Tue, Sep 29, 10:00 a.m. | August JOLTS job openings | The labour-market side of the case for or against further tightening, two weeks after the decision |
| Wed, Sep 30, 8:30 a.m. | August PCE inflation and the third estimate of Q2 GDP | The Fed’s preferred inflation measure, arriving two weeks after the decision it will not have informed |
| Through September | Calendar-month seasonality | September has finished negative in nine of the fifteen completed years from 2011 to 2025 in this series and is the weakest month by average return on that sample, at −3.8% — though the last three Septembers closed positive, at +4.0%, +7.3% and +5.4%. The month is 3.79% lower so far |
| Ongoing | The long end of the global bond curve and the oil price | A US ten-year yield at 5%, a G7 average at an eighteen-year high and a Japanese ten-year at a thirty-year high are all direct competitors to a non-yielding asset, and WTI above $104 keeps the inflation question live. Both feed the same Wednesday decision |
| Unscheduled | What follows the failed CLARITY Act vote | Decrypt reported that the SEC is already working on a standalone “Regulation Crypto Assets” framework and the CFTC on rules for prediction markets and commodity spot markets, so the regulatory path does not end with the bill. Polymarket’s odds of 2026 passage stood at 14% before the vote |
Reading the combination
The two ends of the time axis have not been this far apart at any point in this run. The long frame extended, and it did so mechanically: a twenty-eighth consecutive close above the 200-day average, a 50-day rising for a thirty-fifth straight session, a golden-cross gap out from $1,223.92 to $1,413.10 in its eighth consecutive widening. Every one of those held on a day price fell 3.31%, because each is a statement about averages whose windows are still shedding the low closes of July. The near frame broke on every measure at once, and one of them for the first time in months: RSI through 50 for the first time since August 16; the MACD histogram to −787.06, its most negative since June 9, with the MACD line resuming the lead down; %B at 0.0172 with the close $96.26 above the lower band and the session low $600.41 through it; band width up 0.7208 points in its largest expansion since August 27, ending the squeeze; and the shortfall to the 20-day average back out from $286.20 to $2,704.94. The cushion above the 200-day gave up $2,639.46 to $5,345.15, the largest single-session contraction of this run, taking the decline required to reach the line from 10.21% to 7.07%. Three qualifications belong beside that. First, this was a broad risk-off session, not a crypto-specific one: the S&P fell 0.45%, the Nasdaq 0.78%, and global ten-year yields printed multi-decade highs — Bitcoin fell further, as a higher-beta asset does, but it did not fall alone. Second, the mechanism looks more like forced deleveraging than distribution: roughly $870 million of long liquidations against $71.4 million of short, on CryptoSlate’s CoinGlass figures, is positions closed by margin engines rather than sold by owners, and it fits a $2,591 decline concentrated into hours with a $696.67 lower wick and a bounce into the close. Third, the two numbers that would settle the most important questions do not exist yet — Tuesday’s ETF flow, which would say whether the fund channel joined in, and Wednesday’s Fed decision, which roughly nine-tenths of quoted odds expect to be a rate increase. The base rates split along a seam this brief has now described twice in two days from opposite directions: the filters keyed to volatility expanding and to heavy volume sit well above the 58% above-200-day baseline (66% on 981 instances, 65% on 612, 61% on 333), the filter keyed to where price landed sits near it (60% on 151), and the filter keyed to RSI losing the 50 line sits below it (49% on 159) — while the filter keyed to the size of the RSI fall sits well above it (63% on 155). Both of those last two describe today. The indicators do not say which way this resolves, and with a Federal Reserve decision inside twenty-four hours, the honest reading is that Tuesday removed the near-term repair Monday had built and left the long-term structure standing on a margin a third smaller than it had the day before.
Sources
Price, volume and all technical indicators: Alpha Vantage BTC/USD daily series (00:00 UTC candles), computed by the author. Data note. The Alpha Vantage daily endpoint, which served stale partial candles on each of the two previous runs, returned the completed September 15 candle on the first request of this run at 00:44 UTC; the presence of a partial September 16 row confirms the September 15 candle is closed. No figure in this brief was patched or estimated. The CLARITY Act vote, the Coinbase and Circle share moves, bond yield levels, the Polymarket odds and The Kobeissi Letter quote: Cointelegraph, its report on crypto equities and its news write-up of the vote. Short-term-holder profitability: Cointelegraph. Liquidation totals, the CryptoQuant demand quote, the Alphractal Sharpe-ratio observation and the two-week ETF outflow characterisation: CryptoSlate. The named Senate votes, the Lummis and Warren quotes, Strive’s purchase and the SEC and CFTC rulemaking: Decrypt, its Strive report and its Fed preview. The Senate vote result: CNBC. Equity closes, yields, gold, oil and the open-interest change: TheStreet’s September 15 market blog. The 92.5% FedWatch figure and Friday’s 69.4%: Yahoo Finance. Funding, open interest and the named support and resistance bands: CoinStats. ETF flow history: Farside Investors and SoSoValue. Dominance, Ether’s share and total market capitalization: tv-hub.org. Difficulty, block height and the next retarget estimate: CoinWarz; hashrate: CoinWarz. The week’s central-bank and data calendar and the jobless-claims estimate: Coinpedia. The $75,000 break: KuCoin. Fear & Greed: alternative.me’s dated API. News sentiment tally: Alpha Vantage NEWS_SENTIMENT feed for September 15.
Disclaimer. This post is published for general informational and educational purposes only. It is not investment, financial, trading, tax, or legal advice, and nothing in it should be read as a recommendation or solicitation to buy, sell, or hold Bitcoin or any other asset. Indicators describe past and present data; they do not predict future prices. Figures are drawn from third-party sources believed to be reliable but not guaranteed, and may be revised. The author holds no responsibility for any loss or decision arising from the use of this content. Cryptocurrency prices are highly volatile; anyone considering a trade should do their own research and consult a licensed financial professional.