On 10 October 2025 Benjamin Cowen wrote on X: "Ethereum has now dropped 30% and is at its bull market support band. I think ETH can now rally to new all time highs." He had taken the band he made popular for Bitcoin and laid it over Ether.
Ether rose to $4,241 (13 October) and never made a new high. At the weekly close of 9 November it sat at $3,583, below both lines. In June 2026 it closed at $1,568.
As support, the band did not hold. As an exit rule, it worked: selling on the break on 9 November and buying back only at the next close above both lines meant re-entering on 23 August 2026 at $2,463, 31 % cheaper.
The band can be read as support or used as a rule, and those are two different tools. This piece is about the second one. The question: does the rule carry over to other coins, to stocks and to gold, or is it a Bitcoin tool?
Two lines on the weekly chart
The Bull Market Support Band is the 20-week simple moving average together with the 21-week exponential moving average. Who invented it is not documented. Cowen made it known through his channel Into the Cryptoverse; the most-used TradingView script was published by zkdev in March 2021.
Unlike a single 200-day line, the band has a width. That gives a rule with three states:
- weekly close above both lines: invested
- weekly close below both lines: cash
- weekly close between the lines: do nothing
This rule runs unchanged on every asset here, with a 0.1 % fee per side. No parameters were searched. What would be tested was fixed before the first run.
In Bitcoin, the band delivers what the rule promises
On Binance data from January 2018:
| BTC, 2018 to October 2026 | Band rule | Holding |
|---|---|---|
| return per year | 17.8 % | 20.4 % |
| worst loss | โ66 % | โ80 % |
| time in market | 54 % | 100 % |
From 2012 (Bitstamp data) the pattern is the same: a worst loss of โ72 % instead of โ83 %, and 83 % instead of 89 % a year. In every window tested, the worst loss was 11 to 19 points shallower than holding. Return came out slightly ahead in some windows and slightly behind in others.
In Bitcoin the band is a brake, and it adds no return. Does the brake still work when the asset is not Bitcoin?
In stock indices, the brake works too
The same rule on price indices, without dividends, with cash earning nothing:
| Index, window | return p.a. band / hold | worst loss band / hold |
|---|---|---|
| S&P 500, 1950โ2026 | 6.2 / 8.2 % | โ44 / โ56 % |
| S&P 500, 1950โ1999 | 7.5 / 9.2 % | โ21 / โ48 % |
| S&P 500, 2000โ2026 | 3.8 / 6.4 % | โ39 / โ56 % |
| Nasdaq-100, 1986โ2026 | 12.0 / 14.0 % | โ53 / โ83 % |
The Bitcoin pattern repeats: the worst loss is 13 to 30 points shallower, and return runs about 2 points a year lower. For the Nasdaq-100 that means losing about half of the portfolio instead of four fifths.
In stocks that price is not small. In the S&P 500, 2 points less a year over 76 years leave just under a quarter of the final wealth. Here the band sells you calm, and it pays nothing extra for it.
In gold, you pay and the brake never grips
| Gold | return p.a. band / hold | worst loss band / hold | round trips (winners) |
|---|---|---|---|
| gold future, 2001โ2026 | 5.4 / 11.3 % | โ45 / โ44 % | 74 (25) |
| PAXG (gold token), 2021โ2026 | 13.1 / 15.4 % | โ26.2 / โ26.6 % | 11 (3) |
On the gold future the rule cost half the return and did not make the worst loss any shallower; on the token it saved less than one point. There were 74 entries and exits in 25 years, 25 of them profitable. Between 2011 and 2015 gold fell for years in small steps, and the band sold after every dip and bought back after every bounce.
On gold, the band mostly racks up fees. That leaves the altcoins, where most people use it.
In altcoins, the brake grips in every second coin
The test covered the 50 largest coins with a Binance pair, from 2018 or from listing. 35 of them, Bitcoin included, had at least three years of trading. Fixed in advance: the rule "protects" a coin when its worst loss is at least 10 points shallower than holding.
| 35 coins, at least three years | result |
|---|---|
| protection of at least 10 points | 17 of 35 |
| worst loss, median | band โ79 %, hold โ94 % |
| deeper than โ70 % even with the band | 27 of 35 |
| band fell deeper than holding | 3 (BNB, QNT, XRP) |
| return above holding | 19 of 35 |
The median looks like protection: โ79 instead of โ94 %. But at โ79 % four fifths are gone even with the band, and that happened to 27 of 35 coins. In BNB the rule fell to โ85 %, holding only to โ72 %.
Return beat holding in 19 of 35 coins. Often that only means the rule lost less: in 16 of the 35, the band's return was negative too.
The protection was large in Ether (โ74 instead of โ94 %), Solana (โ67 instead of โ96 %) and in DOT, HBAR, APT, ARB, AVAX and ALGO, each by at least 25 points. These are coins with long, unbroken downtrends. Where a coin swings back and forth, the band exits late and re-enters late.
A brake that grips in every second coin does not work as a rule for an altcoin basket. It works for single coins whose trends run long enough, and that is only known afterwards. What remains is whether the two common readings of the band hold up.
"Above the band is a bull market" holds clearly only for Bitcoin
Sam Daodu wrote at 24/7 Wall St in May 2026 that the band "historically separates bull and bear phases". The measurement: what followed in the 26 weeks after a weekly close.
| next 26 weeks, median | above the band | below the band |
|---|---|---|
| Bitcoin (from 2012) | +43.5 % (73 % positive) | +1.8 % (51 % positive) |
| Ether (from 2017) | +11.3 % (59 %) | +6.7 % (56 %) |
| S&P 500 (from 1950) | +5.3 % (75 %) | +4.3 % (62 %) |
In Bitcoin the band separates the next six months clearly. In Ether and the S&P 500 the values above and below the band sat close together. Below the band does not mean the next months fall.
Then the second reading, the band as support. Nic Puckrin of Coin Bureau called it "a key support for bull markets" in April 2026. The count covers every week in a long phase that came within 2 % of the band:
- Bitcoin: 15 cases, 8 held, 7 broke within four weeks
- Ether: 10 cases, 2 held, 8 broke
- S&P 500: 576 cases, 59 % held
Cowen himself wrote in January 2024 that the band held in 2012 and 2016 and failed in 2020. At eight of 15 in Bitcoin, the band as support is roughly a coin flip. Buying on it is a bet, not a rule.
Whether the band pays depends on the asset
| Asset | softens the worst loss? | cost in return |
|---|---|---|
| Bitcoin | yes, 11 to 19 points | about even |
| Ether, Solana | yes, about 20 to 30 points | none, return higher |
| other large altcoins | in 14 of 32 | mixed |
| S&P 500, Nasdaq-100 | yes, 13 to 30 points | about 2 points a year |
| gold | no | 2 to 6 points a year |
Holding Bitcoin or a stock index and not wanting to sit through the deep crashes, the band gives you a rule that has demonstrably done that, with a return about level with holding or a little below.
Before laying the band on an altcoin, check whether the protection ever carried in that coin. In gold, you can skip it.
Test your own coin and window in the tool: Bull Market Support Band.
Still open is whether you can tell in advance in which altcoin the band will grip. Band width or trend length would be candidates. Either would be its own measurement with its own rules fixed beforehand.
How this was measured
The rules were fixed before the first run (prereg hash b5a4ec61, 23 October 2026). Crypto ran in the Backtesting Arena (strategy bmsb_band, weekly candles, Monday UTC). Holding's worst loss was recomputed per coin over the same window as the rule, with an own engine that matches the Arena within 1.2 points of return and 0.5 points of drawdown. Stocks and gold ran on the same engine with Yahoo daily data.
Limits: the 50 coins are today's top 50. Coins that disappeared since 2018 are missing, and that is exactly where the band would have protected most. Cash earns nothing and dividends are left out, which both slightly disadvantage the rule in stocks. The gold future is a rolling contract.
FAQ
Why weekly candles and not daily? The band is defined on the weekly chart. A script that calculates on the daily chart shows a 20-day and a 21-day line, a different tool with the same name.
Why does a close inside the band not count as an exit? That is how the rule is defined: the state changes only once the close clears both lines. In Bitcoin, one candle in May 2021, which closed $27 above or $404 below the EMA depending on the exchange, decided whether the band sits slightly ahead of or behind holding since 2011.
Do the numbers differ from the tool? Not for Bitcoin from 2018: 17.8 % return, โ66 % worst loss, 17 round trips, the same values as in the tool. A different start date gives different numbers; the pattern of a brake without extra return held in every window.
No investment advice, no recommendation, no forecast โ historical patterns are no promise.
Study the Past โ Improve your Future ๐ฅ