Methodology

How the Gresham Index works

The Gresham Index is a single number between 0 and 1 that answers one question: where are we in Bitcoin's market cycle right now?

It is not a prediction. It doesn't forecast price, call tops, or name bottoms. It reads seven measurements of the market's condition and expresses them as one position on a scale — the way a barometer reports pressure without claiming to know tomorrow's weather.

Everything below is the complete method, including what we tested and threw away.

What the number means

The index is a percentile rank against Bitcoin's own history.

READ IT LIKE THIS

A reading of 0.15 doesn't mean "Bitcoin is 15% expensive." It means: of all the days in the comparison window, only about 15% looked cheaper on these measurements than today does. A reading of 0.90 means today sits in the top tenth of that history.

That framing is what keeps the scale meaningful from $300 to $120,000. Bitcoin's absolute values change by orders of magnitude across a decade; its relative extremes recur.

So the historical readings on our chart are what the method would have printed on that day. Readings published since 13 July 2026 aren't reconstructions at all — they're the actual daily output, timestamped as published. That boundary is marked on every chart.

The seven indicators

Three families, because different kinds of evidence fail at different times. Price measures saturate in long bull runs. On-chain measures distort when holder behaviour shifts structurally. Momentum can stay hot for months. Requiring agreement across families is what stops any single distortion from running the reading.

Trend deviation — 35% of the index

How far price has stretched from its own long-run behaviour.

IndicatorIn familyWhat it measures
Mayer Multiple30%Price relative to its 200-day average
200-week extension30%Price relative to its 200-week average
Log-regression deviation40%Price relative to a regression fitted only on past data, refit as history accumulates
On-chain valuation — 40% of the index

What holders and miners are actually doing, rather than what price is doing.

IndicatorIn familyWhat it measures
MVRV-Z40%Market value against the aggregate price at which coins last moved, scaled by its own historical variability
NUPL30%Net unrealised profit or loss held across the network
Puell Multiple30%Miner revenue relative to its own annual average
Momentum — 25% of the index
IndicatorIn familyWhat it measures
Monthly RSI100%Whether the current move has cycle-scale conviction, measured on monthly closes

Momentum carries the single largest individual weight, and it is deliberately slow — measured on monthly rather than daily closes. That is what stops the index getting excited about a strong quarter.

WHAT EACH INDICATOR CONTRIBUTES TO THE FINAL NUMBER

Monthly RSI
25%
MVRV-Z
16%
Log-regression deviation
14%
NUPL
12%
Puell Multiple
12%
Mayer Multiple
10.5%
200-week extension
10.5%

How the seven become one

7 indicators
raw measurements
Ranked
against past data only, over 4-year and 8-year windows
Weighted
by family and by indicator
Smoothed
7-day average
One number
0.00 – 1.00

That is the entire calculation. The two ranking windows are averaged equally: the four-year window catches relative extremes, the eight-year window carries long memory and stops the scale saturating mid-cycle.

The weights were set before any backtest was run, from reasoning about what each family measures, and haven't been tuned since.

Why seven, and not just the best one?

We ran each indicator alone, through the same pipeline, judged the same way.

6 of 7
failed at least one cycle when used on their own
4 of 7
called a top in June 2019 — before the price rose another fivefold
1
survived every test — knowable only afterwards

The Mayer Multiple — the most cited Bitcoin valuation measure there is — was among the false alarms, and used on its own over the full period it did worse than a plain fixed schedule.

One indicator did survive every test. But you only know which one afterwards, and picking the survivor in hindsight isn't a method. The composite needed no such guess: it worked in every cycle without anyone having to choose correctly in advance.

What we tested and left out

Macro liquidity (US M2)TESTED · REJECTED

Tested directly, both directions, with a realistic publication lag. It made the index worse — and the intuitive version, treating tight liquidity as elevated risk, would have argued against accumulating at the November 2022 bottom, the best entry in the dataset.

Sentiment and social dataNOT TESTED · EXCLUDED

We'd rather say so than imply an experiment we didn't run. It's vendor-proprietary and unauditable, its history is short and its definition moves, and it mostly measures the crowd reacting to price — which is price again.

The four-year halving cycleTESTED · REJECTED

Tested as an overlay. It made the backtest slightly better, and we left it out anyway: a metric that assumes the four-year rhythm persists will fail exactly when that rhythm breaks.

The zones

0.00–0.15Heavy Accumulation
0.15–0.30Strong Accumulation
0.30–0.50Accumulation
0.50–0.65Light Accumulation
0.65–0.75Neutral
0.75–0.85Distribution
0.85–1.00Heavy Distribution

Readings are grouped into seven named zones for readability. Four accumulation bands, two distribution bands: the asymmetry is honest. Bitcoin spends most of its life accumulating — the index has never printed above 0.926. A symmetric scale would have meant inventing gradations where the market almost never goes.

Neutral is the hinge rather than a step — the band where accumulation conditions have ended and distribution conditions haven't begun.

The data, and the rhythm

The index is computed once a day from the previous day's confirmed close, using price and on-chain data from the CoinMetrics Community API, and published automatically.

WHY YESTERDAY, NOT NOW

The reading always reflects a settled, verified day rather than an unconfirmed tick. A number that changed every five minutes would invite exactly the reactive behaviour a cycle indicator exists to replace. If the data falls more than 48 hours behind, the site says so rather than showing a quietly outdated number.

What the index deliberately doesn't do

Predict. No forecast of price, direction, or timing, at any reading.
Call tops and bottoms. It describes conditions. Tops and bottoms are only identifiable afterwards — and the index has been publicly wrong about one, which stays on the record.
React to news, narrative, or sentiment. Only the seven measurements enter the calculation.
Adapt itself. No machine learning, no parameter that updates in response to recent performance, no discretionary override. The specification is frozen; changing it requires a version number and a published reason.
Depend on the four-year cycle — by choice, at a measured cost.
Know anything about you. It's a public number. It has no access to your holdings, accounts, or balances, and no part of this site asks for them.

Check it yourself

The test vectors

Nineteen dated readings the engine must reproduce exactly. Every daily run re-proves all nineteen before it may publish; a failed check blocks that day's number.

The daily record

Every reading since publication began — append-only, timestamped by version control, impossible to edit quietly after the fact.

This page describes a measurement. What we do with a measurement is a separate question, answered separately — see the strategies.

Informational and educational only — not investment advice.