The CRS Arabica Report

The line above the matrix

Four figures that set the scene. None of them is ranked — they are there so the ranked numbers have somewhere to stand.

Close date

The last market close included in this issue. The issue date is when it was published; the close date is what the numbers are actually made of. They are usually a day apart.

KC front

Coffee C, ICE's arabica futures contract, at its nearest actively traded delivery month, in US cents per pound. This is the number most people mean by 'the coffee price.'

It is a paper price. Nothing in the matrix is ranked against it — it appears in the context line so you can see where physical sits relative to the market everyone quotes.

Basis

Whole-market arabica minus KC front, in cents per pound. Positive basis means real coffee is trading above the futures market; negative means below.

Shown as a level only, never ranked. Ranking it honestly needs a back-adjusted continuous KC series aligned to the physical dates, which is the top item under Not yet carried.

USD/EUR

The daily exchange rate used to express the European ex-dock figure in euros. Sourced fresh each issue rather than averaged, so the desk arithmetic reproduces exactly.

Sample

The stretch of history every rank on the page is drawn against, given as a count of daily observations and a date range.

This is the single most important limit on the page. A rank of 99 against fourteen months of history is not the same claim as a rank of 99 against twenty years, and the matrix cannot make it one.

What each row measures

Fourteen metrics. Levels are prices. Spreads are gaps between prices. Volatility is how hard the price has been moving. One row — managed money — is not a price at all; it is who is holding the risk. Every metric is ranked against its own history and no other row's.

Whole-market arabica ↑ matrix

CRS's arabica-only index: the three ICO arabica group indicators re-weighted to sum to 100%, keeping the ICO's original trade-weight proportions between them. It strips robusta out of the composite without changing the relationship between the arabica groups.

The headline physical arabica level. Most other rows on this page are either a component of it or a spread measured against it.

Washed-weighted arabica ↑ matrix

The same three arabica groups, but with natural-process Brazilian coffee deliberately down-weighted so the index sits closer to the washed arabica that actually settles a KC contract. This is an editorial adjustment by CRS, not official ICO methodology.

When this runs above whole-market arabica, washed coffee is carrying a premium. The gap between the two is published on its own line as Washed tilt.

Colombian milds ↑ matrix

An ICO group indicator: a daily trade-weighted average of physical prices for washed arabica from Colombia. A real transacted average, not a futures price and not an estimate.

Other milds ↑ matrix

The ICO group indicator covering washed arabica from origins other than Colombia — Central America, Peru, and the East African washed origins among them. A broad basket, so it moves less sharply than its components.

Brazilian naturals ↑ matrix

The ICO group indicator for natural- and pulped-natural-process arabica from Brazil.

The largest single volume in the arabica world, and the group most directly exposed to Brazilian weather and harvest timing. When this row moves alone, the cause is usually Brazilian; when it moves with the others, the cause usually is not.

Robustas ↑ matrix

The ICO group indicator for robusta — a different species, not a grade of arabica.

It is on an arabica page for one reason: the arabica − robusta spread needs a denominator. It is excluded from every arabica index above it.

Arabica − robusta ↑ matrix

Whole-market arabica minus the robusta indicator, in cents per pound. A widening spread means arabica is getting more expensive relative to robusta.

The substitution gauge. Commercial roasters revisit blend ratios when this gap reaches an extreme, and that decision feeds back into demand for both sides.

Washed tilt ↑ matrix

Washed-weighted arabica minus whole-market arabica, in cents per pound. Positive means the market is bidding washed coffee above the arabica average; negative means naturals are leading.

A process-and-quality signal rather than a price-level one. It can run to an extreme rank while the underlying price level sits mid-range, and often does.

Colombian − Brazilian ↑ matrix

Colombian Milds minus Brazilian Naturals, in cents per pound. The most direct read on washed-versus-natural demand available from the group indicators.

This spread and Washed tilt have moved together closely — see Correlated cluster. When both are marked, treat it as one observation rather than two.

Colombian − other milds ↑ matrix

Colombian Milds minus Other Milds, in cents per pound. Both sides are washed arabica, so process is held constant and what is left is the premium the market pays for Colombian origin specifically.

Realized vol, 20d ↑ matrix

The annualized standard deviation of daily returns in the whole-market index over the last 20 trading days. Realized means it is measured from what happened, not inferred from option prices.

A description of the recent past, not a forecast. High realized vol says the market has been moving violently; it says nothing about direction and nothing about tomorrow.

Max run-up, 20d ↑ matrix

The largest trough-to-peak rise inside the last 20 trading days — the biggest upward swing that was actually available in the window.

Its change columns show an em dash. A window statistic already contains its own look-back, so asking how much it changed over a further horizon would be measuring the same days twice.

Max drawdown, 20d ↑ matrix

The largest peak-to-trough fall inside the last 20 trading days. Same window as Max run-up, opposite direction.

Read the two together. A window with a large run-up and a large drawdown was choppy, not directional.

Managed money net, % of OI ↑ matrix

Every week the US futures regulator publishes a breakdown of who is holding coffee contracts. This row isolates one group — funds and speculators, money that will never take delivery of a single bag — and reports how far they lean in one direction, as a share of the whole market. Read a value of 20% as: about a fifth of the open market is a bet on the price rather than a claim on coffee.

This is the only row here that is not a price. It tells you who is holding the risk, and that separates two situations a price chart cannot tell apart: coffee is cheap because there is too much of it, or coffee is cheap because this money is leaving. The first tends to persist. The second can reverse before your next contract is signed. When the share sits at an extreme, expect larger moves in both directions — this money can leave far faster than physical demand can replace it.

Read it against the price rows, not on its own. This row high and a price row low means the cheapness is more likely financial than physical — unstable, and liable to snap back. This row mid-range and a price row low is more likely a real surplus, and more likely to hold. The same reading tells anyone placing a hedge whether they are placing it into a crowded book.

Open interest (OI)

The number of coffee contracts currently live — every position that has been opened and not yet closed out or delivered against. It is the size of the market at that moment, and it is what the managed money row divides by.

How to read the marks

The page is monochrome and carries no colour coding. Direction, severity, and evidence strength are each stated by a different device, and they run independently of one another.

Xtop quintile top quintile Obottom quintile bottom quintile Xtop five percent top 5% Obottom five percent bottom 5%

Signed rank

Every value in the matrix is placed against that metric's own history and expressed from 0 to 100. A rank of 96 means the current reading sits above 96% of the readings in the sample.

Ranks are never comparable across rows. Each row is measured against itself, and the histories are not all the same length — a row carrying an asterisk is drawn from a different sample entirely.

Quintile

The top or bottom 20% of a metric's own history. Marked with a circled X for the top, a circled O for the bottom.

Tail

The top or bottom 5%. Same circle as a quintile mark, with a dagger added.

The dagger is the only severity signal on the page. A circle tells you a reading is unusual; the dagger tells you it is rare.

The circle and the dagger

The circle states direction — X for high, O for low. The dagger states severity. Nothing else on the page carries either meaning.

The page is monochrome by design. With the glyph carrying direction, colour has no honest job left, and a red-green scheme would only recode information already stated.

Horizon

A look-back length. On this page: 1w is 5 trading days, 2w is 10, 1m is 21, 3m is 63, and 12m is 252. The right-hand columns show how far each metric moved across each one.

The change is arithmetic and is always true. Whether that change is unusual is a separate question, and the sample cannot always answer it — see Withheld ranks.

Independent window

A stretch of history long enough to hold one non-overlapping instance of a horizon. A fourteen-month sample contains dozens of independent one-week windows but barely one twelve-month window.

Overlapping windows are not additional evidence — they share most of their days. Below the threshold stated in the footer, a rank at that horizon would be arithmetic dressed as a finding, so it is withheld.

Withheld ranks and lighter ink

A horizon marked with a middle dot has too few independent windows to rank. Its values are still shown, in lighter ink, and carry no marks.

Ink density is carrying evidence strength, and it runs independently of value. A large number in light ink is a real move that the sample cannot yet call unusual.

wks

The number of consecutive weeks a rank has held its current tier. Shown only where the rank carries a mark, and only from the second week onward.

This separates an entrenched condition from a one-week spike. A mark in its eighth week is a different fact from a mark in its second, and the matrix will not tell you which without this column.

Correlated cluster

A set of rows that have moved together across the sample, listed with their pairwise correlations.

Marks on clustered rows are not independent evidence. Three marks across a tight cluster is closer to one observation than to three, and the census count does not adjust for this.

Census

The number of marks in this issue and how many of them fall in the tail.

A fast read on whether this is an ordinary week or an extreme one, before you read any individual row.

Asterisk on a row label

The row's history is not the page's default sample — it runs on a different frequency or a different span, stated in the footer.

Without this, a rank drawn from fourteen months and a rank drawn from twenty years would sit in the same column looking equivalent.

The conversion desk and the forward figure

Everything below the matrix. The desk turns an index number into a price at your port; the forward figure sets the futures curve against what this market has actually done.

The conversion desk

A worked arithmetic path from the published index to a delivered price at a named port in a named currency, shown one step at a time.

The differentials are contract constants, so the same arithmetic runs on any starting figure — including one this report never published. It is a method, not a quote.

Origin differential

A fixed premium or discount applied on the basis of where the coffee was grown. Set by the contract itself, not by today's market, and quoted in points where 1,000 points is 10.00 cents per pound.

Port differential

A fixed discount applied on the basis of which licensed port takes delivery — European ports carry a standing deduction. Also a contract constant.

Ex-dock reference price

The physical index re-expressed at a region's contract port and currency. It deliberately excludes freight, insurance, and duty.

Not landed cost. Every figure in the desk traces to a published source, and the moment freight or duty enters, it stops being sourced and starts being estimated.

Forward move vs historical range

The size of the move the futures curve is pricing over the next twelve months, set against the largest twelve-month move in the physical record, as a percentage.

Below 100% means the curve is promising less movement than this market has actually delivered. It is a statement about the width of the forecast, not its direction, and it inherits every limit of the sample it is measured against.

Not yet carried

Metrics the report intends to publish and does not yet, each with its status and what is missing.

Published as a standing list rather than quietly omitted. A reader can only judge what the page leaves out if the page says what it leaves out.