Daily research · Coin events · written
What happens to a coin in the 30 days after a TVL crash?
Short answer
The middle case of 442 since 2020 saw a coin move −8.6% in the 30 days after a TVL crash. It ended lower 67% of the time. The best 1 in 10 reached +29.1% or more, the worst 1 in 10 −36.0% or less. Not tested out of sample yet.
Typical move
−8.6%
over the next 30 days, the middle case
How often
67%
it ended lower
Best 1 in 10
+29.1%
or better
Worst 1 in 10
−36.0%
or worse
Cases
442
2020-2026, on 328 days
Higher 33% of the time against 39% on ordinary days (263 days compared): higher less often than on ordinary days.
Share on Xcopy link§ 01 · The pattern442 cases
The pattern, exactly
as of · the rule, the data, the dates
A case is a day the protocol's locked value fell 20% or more and stayed down (DefiLlama), and the move that counts is the coin's own.
Each of the 442 moves is measured from the close before the event's day to the close 30 days after that day.
In our data that gives 442 measured cases on 328 separate days, from Sep 5, 2020 to Sep 23, 2026; the numbers were written on Oct 24, 2026.
§ 02 · Why it might work
Why it might work
as of · the idea traders have in mind, not a promise
A sudden fall of a fifth or more in a protocol's locked value means money left in a hurry: from fear, for a better yield elsewhere or from trouble at the protocol. Traders watch whether the token follows the money out.
That is the reasoning behind watching TVL crashes; whether it held over our 442 cases since 2020 is what the next sections measure.
§ 03 · How we tested it
How we tested it
as of · our own closes, against ordinary days
- We measured each of the 442 cases from the close before the event's day to the close 30 days after that day, on our own daily closes (UTC).
- Each case is set against each coin's own ordinary days (30-day moves starting in the 365 days before it, at least 90 of them), and 263 of the 328 days had that many to compare.
- Cases on the same day count as one day: the 442 cases fall on 328 days, and each day's cases are averaged before anything is counted.
§ 04 · Results93 up · 223 down
Results
as of · every past case, Sep 5, 2020 to Sep 23, 2026
30 days later, the middle case stood at −8.6% and the average day at +0.55%; 93 cases ended up, 223 down and 126 sideways (within ±9.3%).
ended up (93)ended down (223)sideways, within ±9.3% (126)each line one real case · past cases, not a projection
1, 7 and 30 days against ordinary days
| After | Cases | Typical move | Typical, ordinary days | Higher | Higher, ordinary days | Against ordinary days |
|---|---|---|---|---|---|---|
| 1 day | 451 334 days | −0.39% | −0.42% | 44% | 47% | no clear edge over ordinary days |
| 7 days | 449 333 days | −2.0% | −2.1% | 41% | 45% | no clear edge over ordinary days |
| 30 days this post | 442 328 days | −8.6% | −7.8% | 33% | 39% | higher less often than on ordinary days |
At the other horizons: 1 day, typical −0.39% over 451 cases; 7 days, typical −2.0% over 449 cases.
By year, the typical move ranged from −15.7% in 2022 to +2.6% in 2023; 6 of 7 years leaned the same way as the whole.
By market cap: large caps −11.0% over 5 cases; mid caps −0.34% over 37 cases; small caps −7.7% over 84 cases; micro caps −11.0% over 316 cases.
Year by year
| Year | Cases | Typical move | Higher | Lower |
|---|---|---|---|---|
| 2020 | 9 | −2.4% | 44% | 56% |
| 2021 | 61 | −4.1% | 35% | 65% |
| 2022 | 87 | −15.7% | 23% | 77% |
| 2023 | 39 | +2.6% | 52% | 48% |
| 2024 | 36 | −13.6% | 27% | 73% |
| 2025 | 110 | −8.6% | 32% | 68% |
| 2026 | 100 | −6.2% | 34% | 66% |
By market cap on the day
| Coins worth | Cases | Typical move | Higher | Lower |
|---|---|---|---|---|
| over $5B large caps | 5 | −11.0% | 20% | 80% |
| $500M to $5B mid caps | 37 | −0.34% | 37% | 63% |
| $50M to $500M small caps | 84 | −7.7% | 39% | 61% |
| under $50M micro caps | 316 | −11.0% | 32% | 68% |
The latest 10 cases
| Day | Coin | What happened | 1 day | 7 days | 30 days |
|---|---|---|---|---|---|
| B2 | TVL fell 87% in a day | open | open | open | |
| UNI | TVL fell 35% in a day | +2.4% | open | open | |
| DOLO | TVL fell 21% in a day | −3.1% | open | open | |
| HEMI | TVL fell 100% in a day | −5.2% | open | open | |
| G | TVL fell 26% in a day | −5.7% | open | open | |
| CHIP | TVL fell 22% in a day | −11.8% | open | open | |
| BONK | TVL fell 22% in a day | −4.2% | open | open | |
| PLSX | TVL fell 25% in a day | open | open | open | |
| STRK | TVL fell 26% in a day | +1.7% | −32.7% | open | |
| CAP | TVL fell 20% in a day | −24.0% | −29.3% | open |
10 cases still open: the 30-day window of the newest closes on Nov 22, 2026.
Source: our daily closes and our dated events, measured on .
§ 05 · Out of samplenot tested out of sample yet
Out of sample
as of · calls like it, scored on data they never saw
Not tested out of sample yet: the lab has not scored calls made from TVL crashes at 30 days, so the 442 cases above describe the past and are not a prediction.
§ 06 · Verdict
Verdict
as of · plain, with the caveats
In our data a coin ended higher less often than on ordinary days (33% against 39% over 263 days), and both halves of the history lean the same way.
The typical move of −8.6% is small next to the spread between the best and worst 1 in 10 (−36.0% to +29.1%): single cases went either way by far more.
Calls made from TVL crashes have not been tested out of sample yet, so read these 442 cases as a description of the past, not a forecast.
Our price history holds today's coins: coins that died before we began collecting are missing from these 442 cases, which flatters the older years.
Markets change, so what TVL crashes did from 2020 to 2026 can stop working.
§ 07 · Questions
Questions
as of · answered from the numbers above
- How often did a coin end lower in the 30 days after a TVL crash?
- In 67% of 442 cases from Sep 5, 2020 to Sep 23, 2026.
- Measured the scoreboard's way, day by day, the share higher was 33% against 39% on each coin's own ordinary days.
- How big was the move over the next 30 days?
- The middle case moved −8.6% over the next 30 days.
- The best 1 in 10 reached +29.1% or more and the worst 1 in 10 −36.0% or less.
- Can you trade on this?
- Not on its own: over 442 cases it shows higher less often than on ordinary days, and it has not been tested out of sample yet.
- The −8.6% typical move is the past's, and a past pattern can stop working.
- When did it last happen?
- The latest was on Oct 23, 2026: B2, TVL fell 87% in a day.
- Its 30-day window closes on Nov 22, 2026.
- Does it depend on the year?
- No. The typical move was +2.6% in 2023 (39 cases) and −15.7% in 2022 (87 cases).