Curve data model

Gathering and keeping up to data all master data needed to have a sound Commodities trading platform is a challenge. Commopedia aims to ease this effort.What has been gathered so far:

The approach is always to mention the source of the available data and trail you back to the organizations releasing them. Those data are public domain, but bringing them all together, proposing a consistent naming convention, and structuring their retrieval is our main purpose.

The naming convention for each item of the different objects tries to combine both readability by business practitioners and a sound definition of the reality the label covers. Following principles apply:


Commodity trades are often priced relative to a benchmark price (future price from an exchange or index quote released by an agency). Those benchmarks spot well defined delivery conditions. A same delivery combination can be traced by different markers (e.g. Platts, Argus). Because they mark the same reality, those indices are fungible or inter-changeable. It is the same risk (delivery), only the fixing (marker) may be slightly different.

If trading is purely financial, those delivery conditions are taken for granted or even ignored, traders just taking a position on the forward curve set forth by the exchange or agency. However when the delivery becomes 'wet', when effective physical delivery happens, compliance to each attribute implied in the benchmark is verified. Contractual agreements usually foresee premiums or penalties received or paid for being above or below par. Those apply for all attributes, i.e. quality specifications of the goods delivered, distance and facilities of the effective delivery location compared to benchmark one, etc.

Hence the importance to define well the delivery, which is the nucleus on which all pricing will be stemmed. The multiple indices or futures which may mark it, are just different wrappers around the same ‘stock’. The main delivery attributes may be resumed as detailed below. The granularity of those attributes or even the attribute itself will depend on the market which is scrutinized. What matters is to what extend each attribute value triggers price differentiation. If it does, agencies will define distinct indices to trace this. Describing the delivery resumes to identify the attributes of what, where and how: physical, geography and terms.

Delivery attributes



As explained in the price discovery section, there main sources for price dissemination are:

  • Specialized agencies (Platts, Argus, Opis, and others) interview buyers and sellers, identify done deals, even facilitate broadcasting of those, cross check the information, perform statistical checks, some cleansing if required and consolidate all this information. This thorough market polling exercise allows them to release on regular interval (often daily) high/low prices representative of what effectively happened in the market. Those prices are widely accepted by market participants, both physical and financial, to fix price formulae they may have agreed upon among each other.
  • For a limited range of deliveries (the most traded ones with enough liquidity), exchanges have crafted future contracts. Those are 'markets' where bids and offers from participants set the price.

For each market segment under scrutiny a marker is defined. A marker thus stands for a given delivery (physical, geography, terms) quoted by a given publisher (exchange or agency).

Marker attributes


The price agreed for a particular delivery can be outright (fix price) or relative to a benchmark (floating price, spread). How this benchmark is observed or fixed is contractually agreed upon. Each market, depending on its particular conditions, has developed typical 'fixing windows' (monthly average, expiry date of a future, etc.).

Averaging on a larger window than just the exact delivery day, can be motivated to spread out the risk, or by the fact that forward looking, the exact date of the event itself is unknown and contractually parties are bound by a period rather than a date. Monthly average is often applied, but each market may adopt particular windows. Those will be discussed in more detail for each asset. We call the combination of underlying marker and a particular observation pattern an index.

Observation attributes

Observation examples


One may be interested not just in the outright price of a single asset, but in compounding those prices. This stems from the physical reality of commodities markets, e.g. to identify price differences for the same product at different locations or with different time horizons, production blends or consumption baskets, or yield differences between products. Widening or narrowing values for those compound indices can give clear signals as to wether or not transport, store or adjust a production plan. Each market has developed its own terminology to identify particular spreads. There are no standards but habits, the same reality can be labelled differently in each organisation. Typical examples:



Various contract are traded on Exchanges. They can be of different nature:

  • Futures, for which the price discovery happens on the exchange itself. The exchange is the market and a future is altogether marker and contract.
  • Cleared swaps (called Futures by the exchanges) for which the exchange organize a forward market, but which settlement ultimately is exogenous to the exchange, and relies on quotes provided by agencies. Exchanges propose listed contracts mimicking structures (swaps, options) on markers which are actively traded in OTC mode.

Contract attributes


The 'primary' markers are those publishing information directly on the observation of what happens in the market, such as:

  • The index values published by price reporting agencies having polled market activity.
  • Future prices which are a market of their own.

Each marker has a term structure associated to it. Exchanges define for each contract very strict sets of maturities which can be traded. In OTC mode habits prevail depending on how a market is used to quote and release data. Each bucket or pillar in the structure corresponds to a delivery period. Buckets can be of different tenor (intra-day, weeks, balance of month, month, quarter, season, etc.).
With the available discrete quotes (each corresponding to a pillar), there are several ways to build a continuous curve (daily or intra-day), based on curve construction rules.

Curve attributes