Seven notions to read the market like a macro desk.

The regime places the economy on two axes: is growth accelerating or slowing, is inflation rising or ebbing. That gives four regimes — inflationary boom, stagflation, Goldilocks, deflation — and each favours certain assets. Prisma adds risk sentiment, and a question long taken for granted: can the central bank come to the rescue?

The regime is read before choosing an instrument: the same technical signal does not give the same result in a disinflationary expansion and in an inflationary slowdown. Identifying the backdrop avoids applying a good method in the wrong one.

For a long time a shock was handled one way: equities fall, bonds rise, the central bank comes to the rescue. That assumes a central bank free to react to growth, so inflation contained enough for it to cut rates without contradicting itself. When that is no longer the case, a supply shock pushes inflation up against a bank that cannot respond: equities and bonds fall together, and the bond hedge disappears.

The verdict combines three readings: the inflation gap to target, the tone of statements and the dispersion of the committee. Its limit is openly acknowledged: the vote count reads public remarks, not official votes, and a member can speak firmly and then vote to hold. When inflation is missing, the verdict stays undetermined rather than guessed: a missing figure must never be read as an available backstop.

Rate expectations are read in prices. Short-term rate futures let you infer the probability of a cut, a hold or a hike at each meeting: if the contract implies a rate 16 basis points above the current one and a hike would be worth 25, the market is pricing about a 64 % chance of a hike. Prisma also shows Polymarket, where each meeting is split into mutually exclusive contracts (“25 bp cut”, “hold”, “25 bp hike”…): the price of each, between 0 and 1, reads as a probability.

These are not opinions, they are money committed. And what informs is the move, not the level: seeing a probability go from 20 % to 60 % in three sessions says a release has changed the market’s mind.

The yield curve says the same over time. The central bank sets the short rate; the market sets the 10-year rate, which includes what it thinks about future rates. The slope between the two is a permanent vote on monetary policy: it flattens when the market doubts the current level will hold, and inverts when it prices cuts. Then there is language: a bank can change nothing and still move its currency with a mere hardening of wording, because the decision was expected and the tone was not.

Polymarket bets are those of individuals, sometimes for small amounts: a thin market can move on a single order, hence the value of comparing them with futures rather than reading them alone. Polymarket is banned for players in France: Prisma only shows its prices, collected by its server.

An inverted curve does not announce “a recession”: it says current policy is judged unsustainable, and the recession follows afterwards, when the bank cuts in the face of the slowdown. As for a statement’s tone score, it is an interpretation of text, not a measurement: it sheds light on the statement, it does not replace it.

The consensus is the average of economists’ forecasts before the release, the actual is the real figure, and the gap between them is the surprise. It is what moves the price: the level of the figure is already in the market. The CESI accumulates these surprises over a rolling window. It does not say whether the economy is doing well, but whether it is doing better or worse than economists expected.

Growth of 2 % is good news if 1.5 % was expected, bad news if 2.5 % was. The range of forecasts, from their minimum to their maximum, deserves a look: when it is wide, economists disagree and the reaction will be stronger whatever the result.

Since an exchange rate is a relative price, two economies are compared: the surprise differential isolates what matters. Two economies that disappoint equally produce no currency move; one that surprises while the other disappoints does.

The consensus adapts: a very high CESI announces its own fall, since economists raise their forecasts, and a CESI at its lowest prepares the opposite. Read it as a cycle, never as a trend. And many figures are republished corrected: a good release alongside a downward revision of the previous month is not a good release.

It remains to be seen whether the differential really explains the price. R² says so: 0 for no link, 1 for a perfect link, and in macro 0.3 is already notable. But a correlation is not a cause, and it is not stable: one that held for three years can vanish when the regime changes.

The PMSI is a single score per currency, computed from published macro data: inflation, growth, employment, confidence, trade, housing. It aggregates what is measured, not what is forecast. Two horizons coexist: Swing looks at the underlying trend over several months, moves slowly and gives the backdrop; Momentum looks at recent surprises, which fade with time, moves fast and gives the tempo.

To compare two indicators, or two currencies, you normalise. A z-score answers “compared with its habit, is this figure extreme?”: it counts in standard deviations, 0 for a normal value, ±2 for a value rarely seen. A 0.3 % move is big on EUR/USD and means nothing on bitcoin; the same scale makes the comparison legitimate.

On the Macro page this logic shows: bright colour is reserved for moves at least 1.5 times larger than usual.

A score is not an order. It says which way the macro leans; the price can lean the other way for weeks, and that gap is precisely what is interesting. Three explanations are possible: the market anticipates what the macro has not yet recorded, another factor dominates (flows, politics, risk aversion), or the price is wrong. The first two are the most frequent: a divergence is documented before it is traded.

The COT report, published every week by the CFTC, says how many futures contracts each category of player holds: leveraged funds, asset managers, dealers, commercials. Prisma derives an index from it that places the current position within its range of the last three years. On the other side, retail sentiment: the share of buyers and sellers that brokers publish (here, OANDA). And options say where protection is being bought: implied volatility, skew, GEX, put/call ratio.

What you look for is not the level but the extreme. When almost everyone is already positioned the same way, there are no buyers left to push the price higher: the position becomes a vulnerability, not a confirmation. Retail traders are read backwards: when 80 % of them are buyers, they have already bought, and their protective orders pile up just under the price.

On options, implied volatility is an insurance price, to be compared with its own past; skew says which side the market fears; GEX says whether market makers’ hedges dampen the move or amplify it; a very high put/call ratio signals fear already expressed, so already in the price.

Crowding measures fragility, not direction: it indicates the size of a possible reversal, never its date. The COT arrives late — data cut off on Tuesday, published on Friday, so at least three days old — and each category has its limit: an asset manager can be short dollars to hedge a portfolio of US equities, and a dealer does not bet on direction.

Retail sentiment is not an entry signal: it says a move the other way would be violent, not that it will happen. And for options on futures, the app reads a proxy ETF, which it flags on screen.

Geopolitical risk tracks the events likely to move capital towards safe havens: conflicts, elections, sanctions, trade tensions. In Prisma, World places them on a globe — news, prediction markets, the US fleet and Atlas notes — and its timeline strings the developments together.

What matters is not an event’s gravity but its surprise. A risk announced months ago is already in the price; it is the unexpected that moves the Swiss franc, the yen and gold.

The same event does not have the same sign everywhere: a safe-haven premium supports a haven currency and weighs on the indices of exporting countries. Reading a shock therefore means asking which asset feels it, and in which direction, before asking whether it is serious.

A currency’s bias aggregates positioning, macro, surprises, rates, geopolitics and retail sentiment: six families of signals. It reads by pairs: a bullish bias on the euro and a bearish one on the yen points to EUR/JPY, not to the euro alone.

The matrix shows one row per currency and one column per component; the conclusion draws a direction and a conviction from it. Conviction is not direction: it says how far the signals point the same way. The Visions then break the bias down by horizon: a pair can be bullish over three months and bearish over three days, and that is not a contradiction, they are two different questions.

A bias is not a recommendation: it is a starting point to check against the price, and the detail by horizon exists so that you can challenge the conclusion. A neutral currency is not without interest, it serves as a clean counterpart. And when a family of signals is missing, the score is recomputed on the others: the app says so, so that an incomplete gap is not read as a full one.