NVT Ratio & Signal

NVT is the ratio of what the market values Bitcoin at to what the network actually moves. The numerator is market cap — price times circulating supply. The denominator is the estimated value of transactions settling on-chain in a day, in USD. Divide one by the other and you get a price-to-utility multiple: how many dollars of valuation the market assigns for each dollar of economic throughput the chain clears. When that multiple runs high, price has outrun on-chain usage; when it runs low, usage has outrun price.

Today’s NVT Signal is 217.1; raw NVT is 857.4; the regime read against Kalichkin’s canonical thresholds is Overvalued. The implicit market cap at $65,837.03 spot is $1.3T; the latest day’s on-chain volume sits near $1.5B. The thinner line on the chart is raw NVT against same-day volume; the thicker line is the smoothed NVT Signal, and it is the one that carries the regime call.

Willy Woo introduced raw NVT in early 2017 as “Bitcoin’s P/E ratio.” The intuition is direct: if the market pays a hundred dollars of valuation for every dollar of utility flowing through the network in a day, the ratio is high; if it pays ten, the ratio is low. The framework was structurally defensible and operationally noisy. Daily on-chain volume jumps an order of magnitude on event days — an exchange rebalancing, a single whale transfer, a stress-day liquidation cascade — and raw NVT inherits all of that volatility one-for-one. It is a real ratio that flickers for reasons that have nothing to do with valuation.

Dmitry Kalichkin published Rethinking NVT Ratio: introducing the NVT Signal on 4 February 2018 with a single change: replace the same-day denominator with a 90-day moving average of on-chain volume. The smoothing does two things. It suppresses single-day bursts that would otherwise drag the ratio around without any shift in the underlying value/utility mix. And it gives the ratio a memory — the denominator now reflects the last quarter of network activity, so the line reads cycle phases as extended regimes rather than as daily noise.

The cost is responsiveness. A 90-day SMA bakes in roughly a quarter-cycle of lag at turning points; NVT Signal at the actual cycle low typically arrives two to three months after price has already bottomed. That is the central trade you accept by reading the smoothed line: cleaner regimes, later turns. The chart shows both so the noisy-but-fast and the smoothed-but-lagged reads sit side by side, but the regime label uses NVT Signal because raw NVT is operationally unusable as a regime gauge.

For every day d with both a price close and a non-zero on-chain volume reading:

market_cap(d) = price(d) × circulating_supply(d)
NVT(d) = market_cap(d) / tx_volume_usd(d)
NVT_Signal(d) = market_cap(d) / SMA(tx_volume_usd, 90)(d)

The market-cap numerator uses the protocol’s halving schedule for circulating supply. The denominator is blockchain.info’s estimated on-chain transaction volume in USD — a quantity that strips out most change outputs but does not net out exchange-internal transfers, mixer noise, or batched custody flows. That distinction matters more for NVT than for almost any other on-chain ratio, because the denominator is the whole signal: every quirk in how transaction volume is estimated lands directly on the reading.

The 90-day SMA is a simple unweighted mean over the trailing window, and the series begins after that warm-up so neither line has a discontinuity at the start. The reading row recomputes both ratios against live spot — the market-cap numerator scales with the BTC price through the day while the 90-day denominator stays anchored to the latest baked row. The cycle-anchor table below uses daily-close pricing for cross-cycle consistency. Reconciliation choices for the volume series live on the methodology page.

Reading NVT Signal at the canonical cycle dates is the cleanest summary of how the ratio moves through regimes — and of how its reach has shrunk. The 2017 cycle top printed 322.6, the 2021 top printed 273.1, and the 2022 post-FTX low printed 131.4 — the first cycle low in our record that never fell into the Undervalued band. The 2024 halving day registered 136.0, a post-halving entry already sitting in the upper Normal band.

The chart uses Kalichkin’s canonical NVT Signal thresholds directly: 45 for the undervalued boundary and 150 for the overvalued boundary. Holding the lines fixed keeps the visual bands aligned with the original NVT Signal literature and with the reference most market-data providers publish. The lines are not the problem. The denominator under them is.

NVT depends entirely on how on-chain volume is cleaned. Adjusted-volume frameworks net out change outputs and some exchange self-transfers; raw-volume frameworks do not. Two charts can both label themselves “NVT” and disagree by a factor of two on the absolute level — not because one has a data error, but because they count different things in the denominator. If you compare our reading to one elsewhere and the numbers diverge, the volume method is almost always the explanation.

The slower structural pressure is base-layer migration. As Lightning, sidechains, and federated systems absorb payment activity, the on-chain volume denominator shrinks for reasons that have nothing to do with utility loss, mechanically pushing NVT readings higher in the late-2020s than the same usage would have produced in 2017. The fixed 45/150 lines do not move to compensate, which means a reading in the Overvalued band today is not strictly comparable to one from a cycle ago.

The most common misuse of NVT is to treat readings near the threshold rails as entry triggers. The record argues against that on both ends. The extreme prints below are the five highest and five lowest NVT Signal readings in the post-warmup history — and the story they tell is one of decay.

Top extremes. The highest readings cluster around the 2013 and 2017 cycle tops, with NVT Signal printing near 388 at the peak. Each cluster led the actual price top by weeks to months and stayed elevated through the early descent — a slow warning, not a sell trigger. The 2021 cycle top printed only 273 on its price-top date, materially below the 2013 and 2017 highs, consistent with the broader pattern of successive cycles producing lower extremes on ratio-based gauges.

Bottom extremes. The lowest readings all sit in late 2011, with prints below 20 — a regime that has not repeated since. The 2011–12 bear spent extended periods in the canonical Undervalued band and early 2015 only briefly dipped into it; the 2018–19 and 2022 bears did not reach it at all. The decay at the bottom mirrors the decay at the top: extremes are narrowing, partly from cycle maturity, partly from the denominator decoupling from real user activity.

2022: the cycle low that never printed Undervalued. The single most important failure to internalise. The early bears — 2011–12 and, briefly, early 2015 — drove NVT Signal below 45, so a generation of analysts learned to wait for the sub-45 print as the deep-accumulation tell. But the reach was already fading: the 2018–19 bear bottomed near 50, never touching the band. The 2022 post-FTX bottom on 21 Nov 2022 — the most violent drawdown of the cycle — bottomed at NVT Signal 131.4, deep in the Normal band. Anyone treating sub-45 as a precondition for buying simply never received the signal at the actual low. The thresholds drifted out from under their own history.

2023–24: inscriptions distorted the denominator. Ordinals (early 2023) and then Runes (April 2024) loaded blocks with data-carrying transactions that move little or no economic value but still register against on-chain throughput. Where that activity inflated the USD-volume denominator it pushed NVT artificially low; where it congested blockspace without adding measured volume it did the opposite. Either way the readings through that span are not cleanly comparable to the payment-only 2017 cycle, and the 90-day SMA only partially absorbs sustained inscription regimes.

Layer-2 migration biases the level upward over time. NVT counts base-layer volume only. As Lightning, Liquid, Rootstock, and exchange-internal ledgers absorb more payment activity, the denominator structurally shrinks for reasons unrelated to any decline in utility. A slowly rising NVT Signal in a calm market is consistent with both “the market is overvalued” and “the denominator is under-counting” — and the ratio cannot tell you which.

Raw NVT spikes on quiet days. The same-day line jumps whenever on-chain volume falls below its quarterly average, even with utility unchanged in any meaningful sense — which is exactly why Kalichkin smoothed it. Use NVT Signal for regime calls; read raw NVT only as transparency, or for events too short for the SMA to register. A single-day raw print is headline context, not a regime read.

If you are sizing a position, read NVT Signal as cycle phase, not as an entry. The Overvalued band has flagged late-cycle euphoria at every top with weeks to months of lead time — useful for trimming exuberance, useless for calling the exact high. The Undervalued band identified deep buy windows twice — the 2011–12 bear and briefly early 2015 — and missed each cycle low by months. And after 2022, do not treat a sub-45 print as a precondition for accumulating; the indicator may simply never deliver one. Cross-check with the MVRV ratio for profit-position and realized price for cohort cost basis before acting on either rail.

If you are researching the cycle structure, the cleanest finding this page records is that NVT extremes are narrowing — each cycle’s top and bottom prints sit closer to the Normal band than the last, and the 2022 bottom never reached Undervalued at all. Whether that compression reflects market maturity, denominator drift from Layer-2 migration, or a structural change in how Bitcoin trades is a question the data poses but does not settle. Triangulate the denominator itself with the transaction-count series before drawing a conclusion about utility.