Kumo Research · Ethereum
Does ETH rally on Ethereum upgrades? Price and volatility around all 16 upgrades, from Byzantium to Fusaka
"Buy the upgrade, sell the news" is one of the oldest ideas in Ethereum trading. We measured what ETH actually did in the 30 days before and the 7, 30 and 90 days after all 16 mainnet upgrades since Byzantium in 2017 – in dollars, against Bitcoin and in volatility.
By Alex Wu · Published · 7 min read
Key findings
- ETH rose in the 30 days before 9 of 13 feature upgrades, with a median gain of +21.7%. A random 30-day window since 2017 has a median of just +1.1%, so the run-up is real – it would happen by chance less than 2% of the time.
- But Bitcoin rallied almost as much over the same weeks (median about +19%). Measured against BTC, ETH gained a median of only +0.9% before upgrades. The run-up mostly reflects when upgrades shipped – often in strong markets – not extra demand for ETH.
- After upgrades there was no reliable pattern either way. In the 30 days after a feature upgrade, ETH/BTC changed by a median of −2.3%, almost exactly the −2.4% of an average 30-day window. "Sell the news" did not show up as a consistent effect.
- Volatility usually went up after an upgrade: in 9 of 13 feature upgrades, ETH swung more in the two weeks after activation than in the two weeks before (median +9%).
- The outliers are memorable: after Pectra (May 2025) ETH gained 44% in a week and beat BTC by 35%; after the Merge (September 2022) it fell 13% in a month and lagged BTC by 11%.
Why we looked at this
Every Ethereum upgrade comes with weeks of headlines: testnet dates, client releases, and confident takes that the price will "pump into the upgrade" or "dump on the news". Ethereum has now been through enough upgrades to test that with data instead of memory. Since Binance listed ETH in August 2017, the network has activated 16 mainnet upgrades – from Byzantium to Fusaka – and we have daily ETH and BTC prices for every one of them.
We asked three questions:
- Does ETH tend to rise in the month before an upgrade?
- Does it fall or rise in the weeks after – is there a "sell the news" effect?
- Does the market get calmer or more volatile around activation?
The upgrade dates come from the official Ethereum history page. We split the upgrades into two groups. Feature upgrades (13) change how the network works – EIP-1559 fee burning in London, proof-of-stake in the Merge, staking withdrawals in Shapella, blobs in Dencun. Difficulty-bomb delays (Muir Glacier, Arrow Glacier and Gray Glacier) only postponed the "difficulty bomb" and changed nothing for users, so we report them but keep them out of the averages.
The run-up before upgrades is real
In the 30 days before activation, ETH rose ahead of 9 of the 13 feature upgrades. The median move was +21.7%. To judge whether that is unusual, we compared it with every 30-day window in the data: the median 30-day change in ETH since 2017 is only +1.1%, and ETH was up in just 52% of all windows. If you pick 13 random dates, a median run-up of +21.7% or more appears in fewer than 2 out of 100 tries. On its face, ETH does tend to climb into upgrades.
ETH around 13 feature upgrades (median path)
Price indexed to 100 on upgrade day. Days before and after activation.
- ETH in USD
- ETH relative to BTC
Source: Binance ETHUSDT and BTCUSDT daily closes (UTC), 2017–2026 · Chart: Kumo Research, kumocast.io
The chart shows the median path of ETH across the 13 feature upgrades, set to 100 on upgrade day. The solid line (ETH in dollars) climbs steadily from about 82 a month before activation to 100 on the day itself. Then it goes flat.
…but it is mostly the whole market
The dashed line in the same chart tells the second half of the story. It shows ETH measured in bitcoin – the ETH/BTC ratio – and it barely moves before upgrades: from 99.1 to 100. Bitcoin, which does not care about Ethereum upgrades, gained a median of roughly 19% over the same 30-day windows. Ahead of Dencun in March 2024, for example, ETH gained 50.6% in a month – but BTC gained 46.4% in the same month, during the run-up to the US spot Bitcoin ETF flows.
So the honest reading is that upgrades have tended to ship during strong markets. Core developers do not schedule forks around prices, but it is easier to ship large upgrades when the ecosystem is busy, funded and optimistic, and those are also the periods when crypto rises. A trader who bought ETH a month before each upgrade did well – but would have done about as well buying BTC.
After the upgrade: no consistent "sell the news"
If traders really bought the rumour and sold the news, ETH should lag after activation. The data does not show a reliable effect.
ETH vs BTC in the 30 days after each upgrade
Change in the ETH/BTC ratio, %
- Byzantium-21.4%
- Constantinople-2.3%
- Istanbul-12.5%
- Muir Glacier7.1%
- Beacon Chain genesis-18.4%
- Berlin95.9%
- London (EIP-1559)12.5%
- Altair12.0%
- Arrow Glacier-14.4%
- Gray Glacier33.6%
- Bellatrix-18.4%
- The Merge (Paris)-10.6%
- Shapella5.1%
- Dencun-12.0%
- Pectra27.2%
- Fusaka1.8%
Source: Binance daily closes; ethereum.org upgrade dates · Chart: Kumo Research, kumocast.io
In the 30 days after a feature upgrade, ETH/BTC changed by a median of −2.3%. That sounds like mild selling – until you compare it with an ordinary month: the median 30-day change in ETH/BTC across all of 2017–2026 is −2.4%, because ETH has slowly lost ground against BTC over the period. Only 6 of the 13 upgrades saw ETH beat BTC in the following month, and a random-date test puts the post-upgrade average squarely in the middle of normal outcomes.
The spread is wide, which is why anecdotes are so misleading:
- Berlin (April 2021) was followed by ETH almost doubling against BTC (+95.9%) – in the middle of the 2021 bull run, and while the London upgrade with EIP-1559 fee burning was being finalised.
- Pectra (May 2025) saw ETH jump 44.1% in seven days and 36.7% in 30 days, beating BTC by 27.2% over the month.
- The Merge (September 2022), the largest change in Ethereum's history, was followed by a 13.4% fall in ETH and a 10.6% loss against BTC in the next month – a textbook "sell the news" that is often quoted as proof of the rule.
- Dencun (March 2024) saw ETH fall 19.2% in 30 days, and 12.0% against BTC.
Two of the most famous cases point in opposite directions. Taken together, the after-upgrade record looks like ordinary crypto noise rather than a pattern.
Volatility tends to rise after activation
The clearest pattern is in volatility, not direction. For each upgrade we compared how much ETH swung in the 14 days before activation with the 14 days after (annualised standard deviation of daily returns).
Volatility in the 14 days before and after
Annualised volatility of daily ETH returns, %
- 14 days before
- 14 days after
Source: Binance ETHUSDT daily closes · Chart: Kumo Research, kumocast.io
In 9 of the 13 feature upgrades, the two weeks after activation were more volatile than the two weeks before. The median change was +9%. Some of the jumps were large: volatility doubled after Shapella (from 30% to 65%) and Dencun (53% to 100%), and almost quintupled after Pectra (25% to 121%), when ETH rallied hard. The quiet weeks before an upgrade are often the calmest of the quarter, and the market tends to reprice once the event is behind it.
For anyone watching an upgrade, this is the practical lesson: the date is better treated as a volatility event than as a directional one. Ranges widen after activation more often than not, while the direction has been close to a coin flip.
Every upgrade, one table
| Upgrade | Date | Type | ETH 30d before | ETH 7d after | ETH 30d after | ETH 90d after | ETH/BTC 30d after |
|---|---|---|---|---|---|---|---|
| Byzantium | 2017-10-16 | Feature | +34.4% | -16.1% | -1.2% | +301.9% | -21.4% |
| Constantinople | 2019-02-28 | Feature | +29.6% | +1.1% | +5.3% | +98.6% | -2.3% |
| Istanbul | 2019-12-08 | Feature | -18.1% | -5.3% | -5.1% | +57.7% | -12.5% |
| Muir Glacier | 2020-01-02 | Bomb delay | -13.6% | +8.3% | +44.4% | +6.7% | +7.1% |
| Beacon Chain genesis | 2020-12-01 | Feature | +47.5% | -5.2% | +25.8% | +168.2% | -18.4% |
| Berlin | 2021-04-15 | Feature | +39.3% | -4.6% | +45.1% | -20.7% | +95.9% |
| London (EIP-1559) | 2021-08-05 | Feature | +21.7% | +7.7% | +37.4% | +62.8% | +12.5% |
| Altair | 2021-10-27 | Feature | +34.1% | +17.4% | +3.0% | -37.3% | +12.0% |
| Arrow Glacier | 2021-12-09 | Bomb delay | -13.2% | -3.6% | -25.0% | -33.6% | -14.4% |
| Gray Glacier | 2022-06-30 | Bomb delay | -44.8% | +15.5% | +58.4% | +24.9% | +33.6% |
| Bellatrix | 2022-09-06 | Feature | -8.3% | +1.0% | -13.3% | -19.2% | -18.4% |
| The Merge (Paris) | 2022-09-15 | Feature | -21.5% | -9.9% | -13.4% | -11.2% | -10.6% |
| Shapella | 2023-04-12 | Feature | +14.6% | +0.9% | -5.7% | -2.0% | +5.1% |
| Dencun | 2024-03-13 | Feature | +50.6% | -12.2% | -19.2% | -12.7% | -12.0% |
| Pectra | 2025-05-07 | Feature | +16.6% | +44.1% | +36.7% | +99.4% | +27.2% |
| Fusaka | 2025-12-03 | Feature | -11.5% | +4.3% | -2.0% | -37.8% | +1.8% |
Dates from ethereum.org/history (UTC day of activation). "Bomb delay" upgrades only postponed the difficulty bomb. Windows use daily closes; day 0 is the close on activation day.
A few more details stand out in the full table:
- Byzantium (2017) came at the start of the 2017–18 bubble: ETH fell 16.1% in the week after, but was up 301.9% three months later.
- The difficulty-bomb delays had no meaningful effect. Gray Glacier (June 2022) happened near the bottom of the 2022 bear market and ETH gained 58.4% over the next month – a market bounce, not an upgrade effect.
- Fusaka (December 2025), the most recent upgrade, saw a quiet month (ETH −2.0%, +1.8% against BTC) and then a 37.8% fall over 90 days as the wider market weakened.
What this means
Ethereum upgrades matter enormously for the network – fee burning, proof-of-stake, withdrawals and blobs changed how ETH works. But the price record says something more modest:
- ETH has usually risen into upgrades, largely because the whole crypto market was rising at the same time.
- After activation, the direction has been unpredictable. There is no dependable "sell the news" effect.
- Volatility tends to increase after an upgrade, so price ranges widen.
With 13 feature upgrades, the sample is small and every result has wide error bars; a different market cycle could change the averages. We will update this study after Glamsterdam, the next planned upgrade. For current ETH price levels, ranges and volatility, see our live ETH market data.
Method, data and limits
Data: daily closing prices (00:00 UTC) of ETH/USDT and BTC/USDT on Binance from 17 August 2017, and the ETH/BTC ratio computed from them. Upgrade dates and times are taken from ethereum.org/history.
Windows: day 0 is the close on the UTC day the upgrade activated. "30 days before" is the change from the close 30 days earlier to day 0; "7/30/90 days after" are changes from day 0. Volatility is the standard deviation of daily log returns in the 14 days before and after activation, annualised with √365.
Baseline and test: we compare upgrade windows with every overlapping window of the same length in the full sample, and run a bootstrap (20,000 random draws of 13 dates) to estimate how often a result as large could appear by chance.
Limits: 13 feature upgrades is a small sample, and upgrades are not random events – they cluster in certain market phases. Results describe the past and are not a forecast. Binance data starts in August 2017, so the five upgrades from Frontier to Spurious Dragon (2015–2016) are not included.
Download the data (CSV, CC BY 4.0)
Sources
How to cite
Kumo Research (2026). "Does ETH rally on Ethereum upgrades? Price and volatility around all 16 upgrades, from Byzantium to Fusaka". Kumocast. https://www.kumocast.io/research/ethereum-upgrades-price-volatility
This study describes past market data. It is not investment advice and past behaviour does not guarantee future results.
Live data for these markets
More from Kumo Research
- BitcoinBitcoin halving cycles compared: price, peaks and crashes after all four halvings (2012–2026)6 min read →
- DogecoinDogecoin and Elon Musk: what 12 events did to the DOGE price, from "One word: Doge" to D.O.G.E.6 min read →
- EthereumETH vs BTC, 2017–2026: correlation, beta and the long cycles of the ETH/BTC ratio5 min read →
About the researcher
Alex Wu
Crypto, DeFi, perps and FX · driven by data
Alex designed the Kumocast model behind these signals, sets its rules, reviews its accuracy record and updates the methodology whenever the model changes. Signals are refreshed every 30 minutes with live market data.