Macro and news

Crypto trades around the clock, but the market has a schedule that does not depend on it: economic data releases and central bank decisions. The loudest are US data: the Fed rate decision (eight times a year), inflation (CPI and PCE), employment (NFP — the change in non-farm payrolls — and unemployment), producer prices (PPI), GDP. In those minutes every market moves at once: the dollar, bonds, stocks — and crypto with them.

The calendar on the “News” tab takes its schedule from ForexFactory, for six countries. An event row has the time, the country flag, the name, the actual, right after it the deviation from forecast (▲ above, ▼ below, ● matched), and in small print on the right the forecast and the previous value. Importance stars are ForexFactory's rating, not ours. Past days of the week stay in place, paler; past weeks are paged with arrows.

• Forecast — the consensus of analysts polled before the release. It is what the market already expects and has largely priced in.
• Previous value — the figure from the last release. It is sometimes revised after the fact, and the revision can itself be news.
• Actual — what came out. EVA takes it not from a retelling but from the primary source: the statistics agency's series — BLS for employment and prices, BEA for GDP and PCE — or the FRED database these series land in. The calendar names the series the actual came from.

On the chart, US macro releases are marked with a circle bearing “$”. It shows on any timeframe while the news toggle is on and does not depend on the model's rating. The icon is hollow until the actual arrives and filled once it does; its card holds the actual, forecast, deviation and previous value.

The main idea of the lesson: the market is moved not by the number itself but by the difference between actual and forecast. The forecast is already in prices. If exactly what was expected came out, the news adds little. If something else came out, prices have to adjust.

The deviation is simple: actual − forecast, in the indicator's own units.

• Inflation in percent: 3.0% − 3.2% = −0.2 percentage points. Not “−6.25%”: the relative difference (3.0 − 3.2) / 3.2 says nothing here; the market thinks in points.
• Employment in thousands of jobs: 162K − 55K = +107K.
• The rate: 4.00% against a 4.00% forecast — a deviation of 0, even though against the previous 3.75% the rate was raised. The hike was expected and already priced in.

EVA draws the deviation's sign in one colour and does not paint it “good” or “bad”. The reason: for different indicators “above forecast” means opposite things for the market — unemployment above expectations signals a weak economy, payrolls above expectations a strong one. Besides, one release often carries several numbers pointing different ways.

How EVA shows the reaction. Next to the actual there is, say, “×5.1” — BTC's range over five minutes after the release, that many times the usual for that minute of the day. “Usual” is the median range at the same minute on past weekdays without events: in the UTC daytime, when the US trading day begins, the market is livelier than at night even with no event at all. Next to it is the kind of observation:

• “reaction” — the window is clean and the range large;
• “quiet” — the window is clean, the range usual;
• “shared window” — another event within an hour, the range is shared;
• “market moving before” — the market was already lively fifteen minutes before the release.

Even “reaction” only means a clean window and a large range, not that the event made the range. Coincidence in time is not cause. Clicking an event opens a card with the range over 5, 15 and 60 minutes, neighbouring events and past occurrences of the same event.

In the minutes before an important release, participants holding limit orders take more risk than usual: if the number is a surprise, their orders get filled at an already stale price. So many pull their orders or move them away from the price. The book thins, the spread widens (section 3).

What this means in practice. In a thin book a market order walks through more levels, and slippage grows (section 2). A stop order triggered in the first seconds after the release fills worse than in an ordinary hour. The first seconds can be sharp in both directions: the price is first thrown to where there are fewer orders, and only then does the market decide what to make of the number. It is a property of a thin book, not “the market learned the truth”.

A thin book sets no direction. It tells you about speed and fill price, not where the price goes. And the reaction in the first minute and the move an hour later often point different ways.

The “Macro” tab is the broad market: nine cards answering “is money leaving risk or rotating inside the market”. All are computed over the top 125 coins minus tokenised real-world assets, every five minutes; the dashed line on the left of the charts is a reconstructed estimate, not our measurement.

• TOTAL — the whole market's cap in dollars. TOTAL2 — without bitcoin. TOTAL3 — without bitcoin and ether, i.e. the market's tail.
• BTC.D and ETH.D — bitcoin's and ether's share of TOTAL, in percent. USDT.D — tether's share, STABLE.D — all stablecoins together.
• ISSUANCE — the cap of all stablecoins in money: how many dollars have been issued into crypto at all.
• ETH/BTC — the pair's rate, computed from prices, not caps.

The key point about any dominance: it is a ratio, and its denominator is alive. BTC.D rises both when bitcoin gains faster than the rest and when it falls more slowly than the rest. Without TOTAL next to it, the share does not tell you whether there is more bitcoin or less of everything else.

How to read combinations. This describes rotation; it is not a signal.

• USDT.D up, TOTAL down — inside the market money is moving from coins into dollars. But tether's share also grows simply because everything else gets cheaper while tether stays the same: a rising share is not necessarily an inflow.
• BTC.D up, ETH.D down — flight to quality inside the market. Both up — the tail is being sold, also visible in TOTAL3.
• TOTAL2 rising faster than TOTAL — money flowing into alts. TOTAL3 rising — a late stage of risk appetite, money has reached the riskiest; TOTAL3 falling is an early sign of exit, the tail is sold first.
• TOTAL up, ISSUANCE flat — the market got pricier through repricing; no new dollars came. ISSUANCE rising — new money really did come into crypto.

The tab's own caveats. TOTAL2 and TOTAL3 are computed by subtraction from TOTAL using our own dominance, so the parts always add up to the whole — and that is not an independent check. There will be no exact match with TradingView: their index composition is not published, and our bitcoin dominance is usually half a point lower. No card answers “when”: this is a monthly scale, not intraday. The “What agrees with what” block names cases where the metrics disagree; six values from one denominator agreeing is a sign of a coherent picture, not a correct one.

The “News” tab feed is a log of crypto and general-market headlines over a hundred days: TreeNews and the MarketTwits and WatcherGuru Telegram channels. English headlines are translated, the original on hover. The number on a row is the model's rating of the headline: 2 — concerns the crypto market as a whole, 1 — a single coin, company, country or sector, 0 — not about markets. “As one story” merges messages from different sources about one event into one row — the one that reached us first. Silent sources are shown above the feed so “little news” is not confused with “the source is not responding”.

On the chart important news is marked with a lightning bolt over the candle. The model decides importance: will the event move the whole crypto market noticeably more than usual within the next hour. It is the model's forecast from the headline, not a fact that the price moved. The bolt sits on the candle of the event's own time, not of the moment the news reached us; if more than a minute separates them, the card says “learned N min later”.

Two rules for reading news.

News is not the cause of a move. The price often moves before the headline, and the headline explains what has already happened. The bolt on the candle lets you see this: did the move start before the icon or after.

First — who said it and when. Most of the feed is retelling. A channel retells an agency, the agency retells a government office, and something is lost at each step: a number is rounded, monthly and annual change get mixed up, “above forecast” turns into “record surge”. A retelling of a retelling most often diverges from the primary source in the most important part — the number. Three channels with one story are not three confirmations but often one retelling spread across three channels. For US data the primary source is the agency, and EVA takes the calendar's actual from there.