Every cycle someone pulls up the MVRV Z-Score chart, points at the red band sitting up at 7, and tells me the top is nowhere close because the score is only at 3. I used to make the same argument. Then I watched the metric peak lower in each successive cycle while the bands stayed frozen on every charting site, and it slowly sank in that most people quoting this indicator are trading against thresholds calibrated to a market that no longer exists. The metric itself has held up better than almost anything else in the on-chain toolkit, but the numbers painted on it have not.
What the score actually measures
Market cap is the spot price times circulating supply. Realized cap values each coin at the price it last moved on-chain, which makes it a rough aggregate cost basis for every holder on the network. The MVRV ratio compares the two, and the Z-Score version takes the dollar gap between them and divides it by the standard deviation of market cap across the asset's entire history. That normalization is what lets you compare a reading from one cycle against a reading from another, at least in theory.
The logic behind it is why the metric works at all. When market cap sits far above realized cap, the average holder is carrying large unrealized profit, and unrealized profit is the raw material for sell pressure. When market cap drops below realized cap, the average holder is underwater, and historically that is where forced sellers finish selling and the market carves a bottom. The ratio came out of work by Murad Mahmudov and David Puell, building on the realized cap concept from Coin Metrics, and the Z-Score wrapper was added by another analyst shortly after. All of that happened around 2018, which matters, because the bands everyone still quotes were drawn from the cycles visible at that time.
The track record, stated honestly
At the extremes, the historical hit rate is genuinely impressive. The 2011 top, the double top in 2013, and the 2017 top all printed scores in the high single digits or above. The major bottoms of 2011, 2015, late 2018 and 2022 all pulled the score to around zero or below. Very few indicators in any market flagged every one of those turns, and I want to give it that credit before I pick it apart.
The decay shows up after 2017. The first top in 2021 pushed into the old danger zone around 7 and did mark a real cycle peak, but the second top later that year, which came at a higher price, registered less than half that reading. The cycle after that topped out in the same low single digits. Anyone who kept their sell alert set at 7 has now sat through multiple market tops without the alert ever firing. The bottom signal has aged much better, and readings near zero have kept marking deep value even as the top thresholds drifted.
The three failure modes
The first is the diminishing peaks, and it is mostly mechanical. The denominator is the standard deviation of market cap over the full history of the asset, and that number only grows. Every cycle adds bigger absolute dollar swings to the record, which inflates the divisor permanently. Meanwhile the numerator grows more slowly each cycle because price multiples shrink as the market gets larger. Divide a slower-growing gap by a faster-growing spread and the score compresses even if holder behavior is identical. On top of the math, coins also turn over faster near modern tops, so realized cap climbs during the rally itself and closes the gap in real time.
The second is lost coins. Realized cap carries Satoshi-era coins at the prices they last moved, which is either cents or effectively nothing. Estimates vary a lot, but a meaningful slice of supply, plausibly a few million coins, has not moved in well over a decade and much of it is gone for good. Those coins drag realized cap down and push the score up, so the metric has always overstated how extended the market really is. Worse, that distortion shrinks over time as fresh capital comes to dominate realized cap, which compresses the score cycle after cycle in the same direction as the denominator problem, so the old bands end up wrong twice over.
The third is structural change in how coins move. ETF and custody flows re-mark large tranches of old coins at high prices the moment they change hands, which lifts realized cap and pushes the score down. Derivatives pull the other way, since price discovery increasingly happens in futures without any coins moving on-chain, which leaves realized cap stale and props the score up. I do not know which effect dominates in any given quarter, and I doubt anyone does with real confidence. What I do know is that a fixed threshold drawn before either of those things existed cannot be trusted to mean what it once meant.
The bands I use instead
For accumulation, the classic levels have mostly survived. A score below roughly 0.5 is where I start paying attention, and a dip below zero has marked every major bottom so far. The catch is that it can sit down there for months, so this is tranche-buying territory rather than a single all-in entry.
For distribution, I have moved everything down. A score above roughly 3 reads to me the way 7 read in 2017, meaning late cycle, stop adding, start planning exits. Somewhere around 4 to 5 is as stretched as the modern market has managed, and waiting for more than that is betting on a return to a market structure that has been shrinking for years. If you want something more rigorous than eyeballed levels, recompute the score with a rolling standard deviation over roughly one halving cycle instead of the full history, or fit a simple declining trendline across the past few cycle peaks and treat that line as your ceiling.
- Treat readings as zones. The score can hold extreme levels for weeks at tops and months at bottoms, so scale in and out rather than acting on the first touch.
- Confirm with a second cost-basis metric before acting, something like long-term holder spending or a spike in realized profits.
- Re-derive your thresholds at least once per cycle instead of inheriting them from old charts. The drift is slow but it compounds.
- Respect the muted-score scenario. Long consolidations let realized cap catch up, so price can push to new highs while the score stays unremarkable, and a quiet score does not guarantee a safe market.
The bottom signal I still trust close to as published. The top signal I treat as a reference that decays and needs recalibrating every cycle. If the chart you are looking at still paints the danger zone at 7, assume the band is a historical artifact, set your own alerts a few levels lower, and revisit them each cycle. It costs nothing, and it is the difference between distributing into strength and riding the whole round trip.