The claim
Every cycle, the same advice goes around: a big token unlock is coming, so sell before the cliff and buy back the dump. People treat it as obvious. In aggregate it's also wrong. Unlocks aren't harmless, but the drop people blame on the unlock was already happening.
I pulled every discrete cliff unlock DefiLlama tracks: 372 protocols and 22,168 historical unlock events. I priced each one against its token's own history and stripped out two things that fool the naked eye:
- The market. If BTC fell 10% that week, that isn't the unlock. Every return here is measured relative to BTC over the same window.
- The token's own baseline decay. Alts bleed against BTC most weeks, unlock or not. The only honest question is whether an unlock adds anything beyond that normal bleed.
Here's the answer across 669 meaningful unlocks, each releasing at least 1% of circulating supply:
The excess return an unlock adds, over the token's own normal drift, is +0.2%. That's statistically indistinguishable from zero.
The −2.3% you see in the week around an unlock is real. But it's the same −2.3% the token was posting in weeks with no unlock at all.
The one chart that says it
This is an event study: 669 unlocks stacked on top of each other, aligned at the unlock date (day 0). It shows the median token's cumulative return versus BTC, from two weeks before to a month after.
The token is already sliding into the unlock, by about −0.32% per day. The dashed line projects that pre-unlock trend forward: what the price would do if the unlock changed nothing.
The solid line, which is what actually happens, lands right on the dashed line. At T+7 the real median is −2.3% against a −2.5% projection. At T+30 it's −8.4% against −9.9%. After the unlock, the token does slightly better than its own pre-unlock trend.
There is no kink at the cliff. The slope through day 0 is the same as the week before. Whatever is dragging these tokens down, the unlock itself isn't a turning point.
"But surely in a bear market…"
That was my prior too, and the raw numbers seem to back it hard:
| Median return T-1 → T+7 (vs BTC) | Bull | Bear |
|---|---|---|
| Raw | +0.9% | −4.3% |
A five-point gap looks like proof that unlocks are dangerous in bear markets. It's a trap. In a bear market, alts bleed against BTC harder regardless of unlocks: quiet-week drift is −2.25% a week vs BTC. So compare each bear-market unlock with that token's normal bear-market week:
| Median excess over same-regime baseline | Bull | Bear |
|---|---|---|
| Controlled | −1.4% | −2.6% |
The five-point gap collapses to 1.1 points. Roughly 80% of "unlocks are worse in bear markets" is really "everything is worse in bear markets." What's left is a small residual, and it doesn't behave. Sorting by unlock size gives non-monotonic noise (mid-size bear unlocks are actually positive, +0.8%), and the residual effect has about a 55% hit rate. That's a coin flip, not a trade.
What to actually do with this
- Stop paying the "unlock tax." Selling ahead of a scheduled unlock and buying the "dump" is, on average, selling a normal week and buying a normal week, minus two sets of fees and slippage.
- The unlock isn't the risk; the trend is. A token grinding −0.3% a day against BTC will keep grinding whether or not tokens vest this Tuesday. Trade the trend and ignore the calendar entry.
- If you fear anything, fear thin floats, not big headlines. Even there, the effect in this dataset is too weak and noisy to size a position around. Honestly, it mostly doesn't matter.
Method, limits and honesty
So you can shoot holes in it.
- Events: discrete cliff unlocks from DefiLlama's
unlockEvents. Continuous or linear vesting is excluded because it isn't an event. Size = cliff tokens ÷ circulating supply at that moment, with circulating supply derived as cumulative unlocked to date. "Meaningful" means at least 1% of circulating; n = 669 with price coverage. - Returns: daily prices from DefiLlama's coins API, measured T-1 → T+k and market-adjusted by subtracting BTC's return over the same window.
- Baseline: for each token, the median 7-day forward market-adjusted return on quiet days (no unlock within ±10 days), computed separately for bull and bear regimes. The regime is BTC above or below its own 200-day moving average on the unlock date.
- I used the median, not the mean. The mean is useless here: a handful of small-caps that 10×'d after an unlock drag the average to +130%, which describes survivorship, not unlocks.
- Known limits: (1) Tokens that unlock monthly produce overlapping windows, so the effective sample is smaller than it looks. (2) Survivorship: tokens that died after an unlock may be missing from current data. If anything that understates the downside, yet the effect is still about zero. (3) One asset class (DefiLlama-tracked tokens) and one era (2020–2026, about one and a half cycles). (4) I deliberately stopped slicing after regime and size. Keep cutting the data by liquidity, sector, chain and so on, and eventually some cell looks tradeable by pure chance. That isn't a finding, it's p-hacking.
The point isn't that unlocks can never move a specific coin. It's that the general, tradeable rule everyone repeats, "unlocks dump, so sell the cliff", isn't in the data once you stop blaming the unlock for a decline that was already under way.