Market regime & the daily summary
The hub answers "what is the market doing right now" twice: once as a classification — risk-on, neutral or risk-off — and once as five written sentences. This page is the definition of both.
Every figure here describes what has already happened. Nothing on this page is a forecast, and nothing on it is a recommendation to buy or sell anything.
Nothing in the summary is written by a language model
The five sentences are picked by a rule table, not generated. The server sends three things per point — an id, a tone and the raw numbers — and the app turns those into a sentence in your language and your number format. Two consequences, and they are the reason it is built this way:
- A sentence cannot state a number no module produced. The figures are read back out of the modules that own them, so the market cap in the summary is the same computation as the market cap card above it, to the decimal.
- A sentence cannot be missing a number. If any figure a point needs is unavailable, the whole point is dropped rather than printed with a gap in it. The panel says how many of the five it could write.
The regime score
A single 0–100 reading, made of four drivers. Every driver is normalised so that higher always means more risk-on, which means two of them are inverted:
| Driver | What it measures | Direction |
|---|---|---|
| Breadth | The share of measured assets above their 50-day EMA | As measured — broader participation is a broader advance |
| BTC dominance | Bitcoin's share of total market cap, over 30 days | Inverted — money rotating into bitcoin is money leaving everything else |
| Volatility | The universe's median volatility percentile | Inverted — a market at the top of its own volatility range is not a calm one |
| ETF flows | Five-day net spot ETF flow, against its own 20-day activity | As measured — net creations are new money arriving |
The score is a weighted mean of the drivers that are available. Breadth carries the most weight because it is the only one measured across the whole universe rather than off a single series, and it is the one you can check directly against the breadth page.
The three states
| State | Score |
|---|---|
| Risk-off | below 36 |
| Neutral | 36 up to 65 |
| Risk-on | 65 and above |
The band edges are published with the reading, so the coloured scale on the card is drawn from the same numbers the classification used. They cannot disagree.
A missing driver is not a zero
If a driver cannot be measured, the remaining weights are renormalised — the absent driver is not scored zero. On this scale zero means "maximum risk-off", so scoring an unbuilt feed as zero would turn a plumbing gap into a market signal.
The card states how much of the intended evidence the score was made of, and names the missing drivers. Two of the four are genuinely absent early in the product's life: the dominance driver needs thirty days of recorded history (nothing is back-filled — see the dominance page), and the flows driver needs an ETF import.
Below half the total weight the score is withheld entirely. A "market regime" resting on breadth alone is breadth wearing a different name, and it would read as a second, corroborating signal when it is the same one.
"Neutral since 14 February"
The regime state is recorded once a day, and since is the first day of the
current unbroken run of that state. Days on which no regime could be computed do
not break a run — they are days nobody measured, not days the market changed.
There are only three answers and none of them is an estimate:
- the first day of the current run;
- today, when the state recorded this morning differs from the live one — which is a fact rather than a guess, because the change can only have happened since;
- "—", when nothing has been recorded yet, or when the newest record disagrees and is older than today. In that case the change happened somewhere between that day and now, and the page will not pick a date it cannot support.
The arrows
Each driver carries an arrow: how it has moved against the last recorded day that had a reading for it. A move of one point or less on the driver's 0–100 scale reads as flat — below that the arrow would be drawing noise. A driver with nothing to compare against gets no arrow at all, which is a different statement from "it has not moved".
The five summary points
| Point | What it states | What decides its tone |
|---|---|---|
| Market cap | The change in total market cap over the window it was measured against, and the cap- vs equal-weighted return under it | A move of more than 0.1% |
| ETF flows | The last settlement day's net flow for Bitcoin and Ethereum, and the 5- and 20-day cumulatives | The sign of the Bitcoin figure |
| Sentiment | The Fear & Greed reading, its band, and where it sat yesterday and a month ago | The move against yesterday |
| Dominance | BTC dominance and its change in percentage points, plus the stablecoin share | A move of more than 0.1 points |
| Breadth | How many assets hold above their 200-day EMA, and the day's advance/decline split | Whether half or more hold above it |
Three details that are easy to misread and are stated on the card itself:
- The market-cap point names its own window. The change is measured against the most recent recorded day, which is normally yesterday — but after a missed daily recording it is longer, and the sentence says how long rather than claiming a clean 24 hours.
- The breadth point carries two denominators, and both are correct. The advance/decline split covers every asset in the universe; the 200-day EMA count covers only the assets with two hundred days of price history. "50 of 95" and "57 of 100" in one sentence are two true statements.
- The cumulative ETF figures are counted in settlement days, not calendar days, and both windows must be complete. A "20-day cumulative" summed over eleven recorded days would read as a quiet month rather than as a short history, so it is withheld instead.
The flow point is dropped entirely once the newest settlement day is more than a week old: a summary of today's market should not be headlined by a figure from last month.