On 3 November 2026 CoinDesk reported that Bitcoin's chart was flashing a signal similar to the ones seen before the February and June crashes. We tested that signal, the Guppy, all the way back to 2020. On Bitcoin it lost to holding, and seven fixes did not change that. On Ethereum, with one extra rule, it made almost as much as holding while cutting the worst drop from -79% to -50%. Verdict: fade it on BTC, trade it carefully on ETH.
Crash signals get a lot of attention because they promise you can step aside before the pain. The question is whether the signal does that over years, not just on the chart in the article.
So we turned it into a fixed rule and ran it on daily data with fees and slippage in.
What the Guppy is
The Guppy, or Guppy Multiple Moving Average (GMMA), is 12 moving averages on one chart. Six are fast (3, 5, 8, 10, 12 and 15 days) and six are slow (30, 35, 40, 45, 50 and 60 days).
The rule we tested:
- When every fast line sits above every slow line, buyers are in control. Buy.
- When every fast line drops below every slow line, that is the red signal. Get out.
- Long only, no leverage, daily candles, fees and slippage included.
Bitcoin: fine at first, then it fell behind
On Bitcoin the Guppy started well. Over most of the test it made almost the same profit as just holding, with a smaller crash. But from May 2024 it started falling behind: the rule made +30% while holding made +45%.
So we asked Claude to fix it. It tried seven versions on Bitcoin, from January 2020 to November 2025:
| Version | What changed |
|---|---|
| 1 | Only buy above the 200-day average |
| 2 | Faster lines (all lengths x0.8) |
| 3 | Slower lines (all lengths x1.25) |
| 4 | 200-day filter + faster lines |
| 5 | 200-day filter + slower lines |
| 6 | 200-day filter with higher trading costs |
| 7 | Quicker exit |
All seven still lost to just holding Bitcoin.
Ethereum with one extra rule
Then Claude tried something else: the same signal on Ethereum, plus one rule. Only buy when the price is above its 200-day average.
| ETH, Jan 2020 to Nov 2025 | Guppy + 200-day | Just holding ETH |
|---|---|---|
| Return | +2,291% | +2,518% |
| Worst drop from the top | -50% | -79% |
Holding still made a little more. The difference is the worst moment: holding took you down 79%, the rule 50%.
Where it earned its keep: 2022
In April 2022 the rule sold at about $2,930. ETH kept falling, all the way under $1,000 in June. The rule sat out for almost nine months and only bought back in January 2023, at about $1,416. Doing nothing that long is the hard part for humans.
The hidden year
We kept the last full year, November 2025 to November 2026, hidden while we built the rule. It is a year the rule had never seen.
- It had already sold in October 2025, at $3,892.
- It did not buy back until August 2026, at $2,252, skipping most of the drop.
- One trade all year, still up about 4%. Holding ETH over the same year was down about 36%.
The honest caveats
- Small sample. Only 14 trades in almost 7 years. That is too few to tell skill from luck, and one trade in the hidden year is too few to grade it.
- It is slow. In the May 2021 crash ETH topped near $4,170 on 11 May. The rule only got out on 9 June, at $2,506, after riding about 40% of the drop.
- It still trails holding in a straight bull run. What you buy with it is a smaller worst drop, not a bigger return.
Why the same signal fails on one coin and works on another
Bitcoin and Ethereum move together, but not in the same way. Ethereum's bear markets have been deeper and its rallies sharper. A trend signal like the Guppy adds the most value when crashes are deep and long, because getting out early saves a lot. On Bitcoin from 2020 to 2025, the crashes were less extreme and the recoveries came faster, so the signal's late exits and late re-entries cost more than they saved.
The 200-day filter matters for the same reason. It keeps the rule from jumping back in on short rallies inside a falling market, which is where trend rules lose most often. On Ethereum, that kept the rule out for most of the 2022 crash and most of the 2026 drop.
Verdict
- CoinDesk's crash signal on Bitcoin: fade it.
- On Ethereum, with the 200-day filter: trade it, but carefully.
- Right now the ETH rule says stay out. Its fast lines have been below the slow ones since 25 October 2026.
What this means for you
- A signal that worked on one coin can fail on its neighbor. Test it on each market you plan to trade.
- Judge a filter by the worst drop it saves you, not only by the return.
- Hold back recent data while you build, then check the rule on it once. That is the closest a backtest gets to a fresh test.
Try it yourself
You can rebuild the Guppy on any coin in Trigr, add the 200-day filter and backtest it with fees. Start from the docs, and before you trust a 14-trade result, read our overfitting warning signs and our post on paper trading crypto futures.
Sources: CoinDesk: Bitcoin's chart signal looks a lot like the ones before its big crashes (3 Nov 2026) and SuperMoney on the Guppy Multiple Moving Average.
Backtest, not live. BTC and ETH daily, January 2020 to November 2025, with November 2025 to November 2026 held back as unseen data. Fees and slippage included. Past results are not a promise of future results, and this is not financial advice.