The Bull Market Support Band: Every Bitcoin Test Since 2016, Measured
The 20-week SMA and 21-week EMA, explained with real math, plus every weekly close that crossed the band from 2016 to November 2026, computed from free exchange data with the code to check it.
Published 2026-11-21 · Updated 2026-11-21 · Deep Blue Alpha
Learn the basics: Trend and regression models (free lesson)
The bull market support band is the zone between Bitcoin’s 20-week simple moving average and its 21-week exponential moving average, both taken on weekly closes. We rebuilt it from free daily data (Bitstamp, cross-checked with Coinbase) and listed every weekly close that crossed it from January 2016 to November 15, 2026. Bitcoin lost the band 19 times and reclaimed it 18 times: 37 crosses in total, and none at all in 2016 or 2017.
Six losses during uptrends (September 2019, March 2020, May 2021, August 2023, July 2024, March 2025) were reclaimed in 7 to 16 weeks, median 10. Three losses (February 2018, December 2021, October 2025) arrived 2 to 7 weeks after a record weekly close and opened bear markets that took 44 to 61 weeks to reclaim the band for good. The catch: the May 2021 loss also came 6 weeks after a record, and it turned out to be a shakeout. At the moment of the cross, the two kinds looked alike. The band is widely tied to analyst Benjamin Cowen, but no source we found shows he invented it. Updated November 2026.
What is the bull market support band?
Take Bitcoin’s closing price at the end of every week. Average the last 20 of those closes and you get the 20-week SMA. Run a 21-week exponential average over the same closes and you get the 21-week EMA. Draw both on a weekly chart and shade the space between them. That shaded zone is the band.
The common reading has three states:
- Weekly close above both lines: read as an uptrend. In past bull markets, pullbacks often stopped near the band, which is where the “support” name comes from.
- Weekly close below both lines: read as a downtrend. In past bear markets, rallies often stalled near the band from below, so the same zone acted as a ceiling.
- Weekly close between the lines: undecided. In our data this happened on only 24 of 568 weekly closes since 2016, about 4%.
Nothing about the band is special math. It is two ordinary moving averages placed close together so they form a zone instead of a single line. A zone is more forgiving than a line: a close a few dollars under one average does not flip the reading if it is still above the other.
SMA vs EMA: the math, with a real example
The two lines are built differently, and that difference is the whole reason for using both.
The simple moving average gives every week in the window equal weight. For 20 weeks, each close counts for 5%. When a new week arrives, the oldest week drops out. So the SMA moves by (new close − dropped close) / 20. What leaves the window matters exactly as much as what enters it.
The exponential moving average never drops anything. Each week it blends the new close with its own previous value:
EMA_today = α × close + (1 − α) × EMA_last_week, with α = 2 / (21 + 1) ≈ 0.0909
So the newest close gets about 9.1% of the weight, nearly twice the SMA’s 5%, and older weeks fade out gradually. A week’s influence halves roughly every 7.3 weeks. That is why the EMA reacts faster to a sharp move.
Worked example: the week ending November 15, 2026
One week, two very different reactions
| Item | Value |
|---|---|
| Weekly close, Nov 8, 2026 | $74,743 |
| Weekly close, Nov 15, 2026 | $65,312 (−12.6%) |
| 21-week EMA: $76,535 → | $75,514 (−$1,020) |
| Close dropping out of the 20-week window (Jun 28, 2026) | $59,473 |
| 20-week SMA: $73,570 → | $73,862 (+$292) |
EMA: 0.0909 × $65,312 + 0.9091 × $76,535 = $75,514. SMA: ($65,312 − $59,473) / 20 = +$292. Source: Bitstamp daily closes, our calculation.
Bitcoin fell 12.6% in a week, and the 20-week SMA went up. The reason is the week leaving the window: June 28, 2026, the lowest weekly close of the 2025–26 decline. Swapping a $59,473 week for a $65,312 week lifts the average. The EMA has no window, so it simply moved toward the new, lower close. This is a useful thing to know when reading the band: the SMA can rise during a sell-off, purely because of what happened 20 weeks earlier.
Why weekly closes, and which weekly close?
Bitcoin trades all day, every day, so daily charts carry a lot of noise: weekend thin liquidity, liquidation wicks, headline spikes that reverse within hours. A weekly close compresses seven days into one number and ignores intraweek wicks. People who use the band usually stress that only the weekly close counts, not a dip during the week.
But “weekly close” needs a definition, because crypto has no closing bell. This study uses the close of the Sunday UTC daily candle, which is the same as a weekly bar running Monday 00:00 to Sunday 23:59 UTC. That matches how most charting platforms draw weekly bars for crypto. A different convention, such as a Friday close or a local-time Sunday, would shift a few borderline crosses by a week.
Who came up with it? The Ben Cowen question
The band is most closely linked to Benjamin Cowen, who runs Into The Cryptoverse (intothecryptoverse.com, YouTube, X). He has published a recurring series of “Bitcoin Bull Market Support Band” update videos; transcript summaries on Rosetta list them from at least July 2021 through January 2026. Those summaries are auto-generated, so treat their details as secondhand.
Cowen’s framework, as reported in news coverage, treats the band as support during bull phases and resistance during bear phases, with heavy weight on weekly closes. Coverage also reports that he points to the EMA reacting faster than the SMA because it weights recent prices more.
What we could not find is any proof that he invented it. A widely used TradingView script credits the original concept only as “popularized in crypto analysis,” without naming anyone. Pairing a 20-period SMA with a 21-period EMA is a generic technique. The fair summary: Cowen popularized the band and has tracked it publicly for years; its origin is not documented.
Our method, and the code
CoinGecko’s free API now limits history to the last 365 days, so we used Bitstamp’s public OHLC endpoint, which serves BTC/USD daily candles back to 2011. We pulled January 1, 2012 to November 20, 2026 (5,438 finished days, no gaps). Starting in 2012 gives the EMA four years to settle before 2016 and makes the 200-week SMA available from October 2015. We then repeated the run on Coinbase daily candles from July 2015. The two sources matched on 36 of 37 crosses. CoinGecko’s own Monday 00:00 UTC prices for the four most recent crosses were within 0.05% of the Bitstamp closes.
The whole thing fits in standard-library Python:
import json, urllib.request, datetime as dt
URL = "https://www.bitstamp.net/api/v2/ohlc/btcusd/?step=86400&limit=1000&start="
today = dt.datetime.now(dt.timezone.utc).date()
daily = dict()
for start in range(1325376000, int(dt.datetime.now().timestamp()), 1000 * 86400):
for c in json.load(urllib.request.urlopen(URL + str(start)))["data"]["ohlc"]:
day = dt.datetime.fromtimestamp(int(c["timestamp"]), dt.timezone.utc).date()
if day < today: # skip the unfinished day
daily[day] = float(c["close"])
# Weekly close = close of the Sunday UTC daily candle (Monday-to-Sunday weeks)
weeks = [(d, daily[d]) for d in sorted(daily) if d.weekday() == 6]
closes = [c for _, c in weeks]
def sma(x, n):
return [None if i < n - 1 else sum(x[i - n + 1:i + 1]) / n for i in range(len(x))]
def ema(x, n):
a, out, e = 2 / (n + 1), [], None
for v in x:
e = v if e is None else a * v + (1 - a) * e
out.append(e)
return out
s20, e21 = sma(closes, 20), ema(closes, 21)
prev = None
for i, (d, c) in enumerate(weeks):
if s20[i] is None:
continue
hi, lo = max(s20[i], e21[i]), min(s20[i], e21[i])
state = "above" if c > hi else "below" if c < lo else "inside"
if state == "inside":
continue
if prev and state != prev and d.year >= 2016:
print(d, "LOST" if state == "below" else "RECLAIMED", round(c))
prev = state
Run on November 21, 2026, it prints the 37 crosses listed below. Two choices are worth stating plainly. First, a cross means a close moving from above both lines to below both lines (or back); a close inside the band does not end a run. Second, the EMA here starts from the first 2012 close, while some charting tools start it from an average. By 2016 the starting value has decayed to effectively zero, so it does not change any cross.
Every weekly cross, 2016–2026
Coming into 2016, Bitcoin had closed back above the band on October 18, 2015 at $261. It then went 120 straight weeks without a weekly close below the band, through the whole 2016–17 bull market. Only three weeks in September and October 2016 closed inside the band; none closed below it. The first loss came after the December 2017 top.
All 37 weekly-close crosses of the 20W SMA / 21W EMA band, Jan 2016 – Nov 15, 2026
| Week ending (Sun UTC) | Event | Weekly close | Band (low–high) | Weeks to next cross |
|---|---|---|---|---|
| Feb 4, 2018 | Lost | $8,191 | $9,755–$9,953 | 2 |
| Feb 18, 2018 | Reclaimed (failed) | $10,421 | $9,840–$10,277 | 1 |
| Feb 25, 2018 | Lost | $9,590 | $9,817–$10,527 | 1 |
| Mar 4, 2018 | Reclaimed (failed) | $11,463 | $9,967–$10,816 | 1 |
| Mar 11, 2018 | Lost | $9,535 | $9,927–$10,994 | 20 |
| Jul 29, 2018 | Reclaimed (failed) | $8,217 | $7,666–$7,784 | 1 |
| Aug 5, 2018 | Lost | $7,033 | $7,608–$7,715 | 35 |
| Apr 7, 2019 | Reclaimed | $5,191 | $3,825–$4,352 | 25 |
| Sep 29, 2019 | Lost | $8,048 | $9,188–$9,859 | 16 |
| Jan 19, 2020 | Reclaimed | $8,697 | $8,217–$8,376 | 7 |
| Mar 8, 2020 | Lost | $8,034 | $8,446–$8,677 | 8 |
| May 3, 2020 | Reclaimed | $8,916 | $7,802–$7,906 | 55 |
| May 23, 2021 | Lost | $34,707 | $45,353–$48,626 | 11 |
| Aug 8, 2021 | Reclaimed | $43,858 | $39,854–$43,836 | 17 |
| Dec 5, 2021 | Lost | $49,463 | $51,680–$52,463 | 16 |
| Mar 27, 2022 | Reclaimed (failed) | $46,864 | $43,969–$45,505 | 2 |
| Apr 10, 2022 | Lost | $42,134 | $43,722–$44,003 | 40 |
| Jan 15, 2023 | Reclaimed | $20,885 | $18,517–$19,429 | 31 |
| Aug 20, 2023 | Lost | $26,192 | $27,806–$28,526 | 6 |
| Oct 1, 2023 | Reclaimed (failed) | $27,995 | $27,254–$27,915 | 2 |
| Oct 15, 2023 | Lost | $27,174 | $27,303–$27,929 | 1 |
| Oct 22, 2023 | Reclaimed | $30,002 | $27,548–$28,073 | 37 |
| Jul 7, 2024 | Lost | $55,829 | $60,908–$64,990 | 2 |
| Jul 21, 2024 | Reclaimed (failed) | $68,158 | $61,560–$65,695 | 2 |
| Aug 4, 2024 | Lost | $58,142 | $61,804–$65,147 | 3 |
| Aug 25, 2024 | Reclaimed (failed) | $64,262 | $61,517–$63,826 | 1 |
| Sep 1, 2024 | Lost | $57,310 | $61,135–$63,404 | 3 |
| Sep 22, 2024 | Reclaimed | $63,563 | $60,719–$62,679 | 24 |
| Mar 9, 2025 | Lost | $80,708 | $89,271–$92,735 | 7 |
| Apr 27, 2025 | Reclaimed | $93,769 | $87,226–$92,163 | 25 |
| Oct 19, 2025 | Lost | $108,684 | $110,900–$113,002 | 1 |
| Oct 26, 2025 | Reclaimed (failed) | $114,551 | $111,232–$113,440 | 1 |
| Nov 2, 2025 | Lost | $110,530 | $111,168–$113,685 | 25 |
| Apr 26, 2026 | Reclaimed | $78,660 | $77,072–$78,428 | 5 |
| May 31, 2026 | Lost | $73,568 | $74,156–$78,056 | 12 |
| Aug 23, 2026 | Reclaimed | $77,716 | $69,497–$69,565 | 12 |
| Nov 15, 2026 | Lost | $65,312 | $73,862–$75,514 | open |
“Failed” = the reclaim was lost again within two weekly closes. Source: Bitstamp BTC/USD daily closes, Sunday UTC weekly close, our calculation.
The cycles in plain words
2018 bear market. The first loss came on February 4, 2018, seven weeks after the record weekly close of $18,953. Three quick reclaims in February, March and July each failed after one week. The lowest weekly close was $3,194 on December 16, 2018. The band was not reclaimed for good until April 7, 2019, at $5,191: 61 weeks after the first loss.
2019–2020. The 2019 rally lost the band on September 29, 2019 ($8,048), reclaimed it in January 2020, then lost it again in the March 2020 crash, when the weekly close fell to $5,344. It was back above by May 3, 2020 ($8,916), eight weeks later, and stayed above for 55 weeks.
2021: two losses, two outcomes. On May 23, 2021, Bitcoin closed at $34,707, far under the band, six weeks after a record $59,979 close. It reclaimed the band on August 8, 2021 and set a new record close of $65,510 in November. Then on December 5, 2021, three weeks after that record, it lost the band again at $49,463. That one became the 2022 bear market. A brief reclaim in late March 2022 lasted two weeks.
2022–2023. The low weekly close was $16,256 on November 20, 2022. The band came back on January 15, 2023, at $20,885, 58 weeks after the December 2021 loss. A dip in August to October 2023 was reclaimed for good on October 22, 2023 at $30,002.
2024–2025 chop. Summer 2024 was the messiest stretch on record: three losses and two failed reclaims between July and September, all within about 10% of the band. March 2025 brought another loss ($80,708), reclaimed on April 27, 2025 at $93,769. The cycle’s record weekly close was $123,519 on October 5, 2025.
2025–2026 bear market. Two weeks after that record, on October 19, 2025, the band was lost ($108,684), reclaimed for one week, then lost on November 2, 2025. Bitcoin stayed below it for 25 straight weekly closes. It closed above on April 26, 2026 ($78,660), slipped back under on May 31, and made its lowest weekly close of the decline, $59,473, on June 28, 2026. It reclaimed the band on August 23, 2026 ($77,716) and held for 12 weeks, peaking at an $86,510 close on October 4, before losing it again on November 15, 2026 at $65,312.
Summary: shakeouts vs bear markets
Raw crosses overstate how often anything happened, because whipsaws come in clusters. So we grouped them into episodes: from a loss to the first reclaim that held more than two weekly closes.
Every band-loss episode since 2016
| First loss | Weeks since record close | Lowest close before reclaim | Weeks to lasting reclaim | What followed |
|---|---|---|---|---|
| Feb 4, 2018 ($8,191) | 7 | $3,194 (−61%) | 61 | Bear market |
| Sep 29, 2019 ($8,048) | — | $6,908 (−14%) | 16 | Reclaimed |
| Mar 8, 2020 ($8,034) | — | $5,344 (−33%) | 8 | Reclaimed |
| May 23, 2021 ($34,707) | 6 | $31,777 (−8%) | 11 | Reclaimed |
| Dec 5, 2021 ($49,463) | 3 | $16,256 (−67%) | 58 | Bear market |
| Aug 20, 2023 ($26,192) | — | $25,836 (−1%) | 9 | Reclaimed |
| Jul 7, 2024 ($55,829) | 14 | $54,866 (−2%) | 11 | Reclaimed |
| Mar 9, 2025 ($80,708) | 12 | $78,342 (−3%) | 7 | Reclaimed |
| Oct 19, 2025 ($108,684) | 2 | $59,473 (−45%) | 44* | Bear market |
| Nov 15, 2026 ($65,312) | — | — | open | Open |
Percentages are from the first-loss close. “Weeks since record close” is shown only when the loss came within 14 weeks of an all-time-high weekly close. *The April 26, 2026 reclaim held five weeks before failing; counting it, that episode lasted 27 weeks and a second one (May 31 to Aug 23, 2026) lasted 12. Source: Bitstamp, our calculation.
What the numbers show:
- Uptrend losses were all reclaimed, and fairly fast. The six losses inside the 2019–2021 and 2023–2025 uptrends were reclaimed in 7, 8, 9, 11, 11 and 16 weeks (median 10). Five of the six never closed more than 14% below the loss-week close; March 2020 was the exception at −33%.
- Three losses opened bear markets. February 2018, December 2021 and October 2025. Each came 2 to 7 weeks after a record weekly close, and each took 44 to 61 weeks to reclaim the band in a way that lasted.
- Timing alone did not separate them. The May 2021 loss came 6 weeks after a record close, almost the same setup as December 2021. One was a shakeout, the other a bear market. Labels like “uptrend loss” are only clear in hindsight.
- Reclaims fail often. 8 of 18 reclaims were lost again within two weekly closes. Five of the eight came during bear markets (three in 2018, one in March 2022, one in October 2025); the other three came in the 2023–2024 chop.
- Price ran less far above the band each cycle. At the record weekly close of each cycle, Bitcoin sat 2.47 times the top of the band in December 2017, 1.27 times in November 2021, and 1.10 times in October 2025.
The one-line version: since 2016, losing the band during an uptrend was always temporary, and losing it right after a new record was the start of all three bear markets. But nobody knows a record is the cycle’s last one until long afterward. This is a description of a small historical sample, not a rule about what happens next.
How it compares to the 200-week moving average
The 200-week SMA averages almost four years of weekly closes, roughly one full Bitcoin cycle. It moves slowly and, in bull markets, sits far below the band: $1,245 at the December 2017 top (the band was near $7,660), $17,359 at the November 2021 top, and $53,475 at the October 2025 top.
Bitcoin weekly closes vs the 200-week SMA, 2016–2026
| Period | Weeks closed below 200W SMA | Lowest close vs 200W SMA |
|---|---|---|
| 2018 bear | 0 | $3,194 vs $3,178 (+0.5%) |
| March 2020 crash | 1 | $5,344 vs $5,503 (−2.9%) |
| 2022–2023 bear | 46 | $16,256 vs $24,007 (−32%) |
| 2026 (to Nov 15) | 4 | $59,473 vs $62,442 (−4.8%) |
Weekly closes from Jan 2016 to Nov 15, 2026: 51 of 568 were below the 200-week SMA. 2022–2023 count includes later 2023 dips near the line (Jun, Aug–Oct). Source: Bitstamp, our calculation.
The two tools answer different questions. The band describes the trend of the last five months; it flipped 37 times in ten years. The 200-week SMA describes the trend of the last four years; Bitcoin closed under it in only about 9% of weeks, almost all in 2022–2023. In 2018 it never closed below it at all. In 2022 it spent months far underneath. In 2026 it dipped below in June, August and again on November 15, each time by single-digit percentages so far. Each bear market interacted with the 200-week line differently, which is itself a reason to be careful with any fixed rule built on it.
For a sense of scale, the band itself dropped below the 200-week SMA in 41 weeks between October 2022 and October 2023, and at no other time in this period. That is how deep the 2022 bear market went compared with the others.
Cowen’s past stated views, checked against the data
Two of Cowen’s band calls in this cycle are documented in news coverage, both secondhand.
March 2025. Benzinga reported that Cowen flagged Bitcoin’s first weekly close below the band outside a third quarter that cycle as a possible warning sign for the cycle top. What happened: the band was reclaimed seven weeks later, on April 27, 2025, and the cycle’s record weekly close came in October 2025, about seven months later. So the warning was early; the loss that preceded the bear market was the October 2025 one.
2026. Coverage reported that in this midterm year he treated the band as resistance, a “bear market resistance band.” What happened: every weekly close from November 2, 2025 to April 19, 2026 (25 weeks) stayed below the band. Then Bitcoin closed above it twice, in late April and again from August 23 to November 1, 2026, before losing it on November 15. The resistance reading held for about six months, then did not hold for parts of 2026.
We could not open the original videos or every article, so these describe reported views, not verified quotes. Neither is a view of ours, and neither implies any endorsement of this analysis by Cowen.
The limits of the band
- It lags by design. A 20-week average reflects about five months of prices. By the time a weekly close breaks below it, much of a move has often already happened. In December 2021 the loss came at $49,463, already about 25% under the record close.
- Whipsaws are common. 8 of 18 reclaims failed within two weeks. In summer 2024 the band was crossed six times between July 7 and September 22.
- The sample is tiny. Three bear markets and six uptrend losses since 2016. With samples that small, almost any pattern can look reliable. Results are also specific to one asset that survived; the same method on failed coins would look worse.
- Convention matters at the edges. A different weekly close or exchange moves borderline crosses. Our two sources disagreed on the August 2021 reclaim by one week, because that close sat only $23 above the 20-week SMA.
- Weekly closes hide wicks. This study counts closes only. Price traded under the band intraweek many more times than the table shows.
- Phase labels need hindsight. “Uptrend loss” and “bear-market loss” are assigned using peaks and lows that were only known later.
- Markets change. Spot ETFs, treasury companies and derivatives changed who trades Bitcoin after 2024. Behavior around any average from 2016–2022 may not carry over.
Where whale data fits
The band is a price tool; it says nothing about who is moving coins. Deep Blue Alpha tracks 20,000+ whale wallets on Ethereum and shows their real DEX buys and sells first, with exchange flow as a supporting line: coins leaving exchanges read as the accumulation side, coins arriving as the distribution side. Pairing a trend reading like the band with what large wallets actually did in the same weeks gives a fuller picture than either alone. The live data is on the whale feed and the Whale Sentiment Index.
Frequently asked questions
What is the bull market support band?
The zone between Bitcoin’s 20-week SMA and 21-week EMA on weekly closes. Above both lines is read as an uptrend, below both as a downtrend, between them as undecided.
What is the difference between the 20-week SMA and 21-week EMA?
The SMA weights each of the last 20 weeks at 5% and drops the oldest week each time. The EMA gives the newest week about 9.1% and never drops anything, so it reacts faster. On November 15, 2026 the EMA fell $1,020 while the SMA rose $292 in the same week.
Did Ben Cowen invent the bull market support band?
He popularized it and has tracked it publicly since at least 2021, but no source we found proves he invented it. Its origin is undocumented.
How many times has Bitcoin lost the band since 2016?
19 losses and 18 reclaims through November 15, 2026, grouped into 10 completed episodes plus one open. No losses happened in 2016 or 2017.
Does losing the band mean a bear market started?
Not on its own. Six uptrend losses were reclaimed in 7–16 weeks. Three losses that came soon after a record close opened bear markets. The May 2021 case shows the two looked similar at the time.
How does it compare with the 200-week moving average?
The 200-week SMA is much slower. Bitcoin closed below it in 51 of 568 weeks since 2016, never in 2018, and mostly in 2022–2023.
Where was the band in November 2026?
At the November 15, 2026 weekly close: 20-week SMA $73,862, 21-week EMA $75,514, close $65,312, 200-week SMA $67,977. CoinGecko showed Bitcoin near $58,300 early on November 21, 2026, before that week closed. These readings change weekly.
Bottom line
The bull market support band is two ordinary moving averages, and its history is easy to check. From 2016 to November 2026, Bitcoin crossed it 37 times on a weekly-close basis. It held through all of 2016 and 2017. Losses during uptrends were reclaimed within 7 to 16 weeks every time. The three losses that came soon after a new record close were the doorways into the 2018, 2022 and 2025–26 bear markets, and each took most of a year or more to undo.
The same data also shows why the band is a description, not a decision rule. It lags, it whipsaws, nearly half of reclaims failed within two weeks, and the cases that mattered most could not be told apart from shakeouts when they happened. It is attributed to Benjamin Cowen, who has done the most to popularize it, though its true origin is not documented. Use the table above as a record of what happened, and the code to check it yourself.
See what whales are doing this week
Price averages show the trend. Deep Blue Alpha shows the wallets behind it: real DEX buys and sells and exchange flow from 20,000+ tracked whale wallets on Ethereum, updated block by block.
Open the live whale feed →