Forward Curve Construction for Power & Gas Trading

Executive summary

Forward curves are where market prices become tradable value. A curve, the complete set of expected prices across future delivery periods, is the foundation of mark-to-market, position valuation, VaR, P&L, hedge optimization, and structured pricing. Get the curve right and every valuation that references it is sound; get it wrong and the error propagates into every number the desk produces.

Forward curve construction is also one of the least written-about topics in ETRM, despite being central, because it sits at the intersection of market knowledge and quantitative technique. This guide is a deep, practitioner-level treatment of how prices become curves: the market inputs, the construction workflow, the specifics of power and gas, and the shaping, interpolation, and governance that turn sparse traded points into a complete, usable, auditable curve.

It builds on the market data platform guide, which covers ingestion and governance, and connects to the quant engine and valuation. Where the market data guide is about trusted inputs, this one is about the trusted derivation on top of them. The recurring theme: a curve is a model, and its value depends on sound method, honest assumptions, and full reproducibility.

What a forward curve is

A forward curve is the complete set of expected market prices for a commodity across future delivery periods. Where a spot price tells you the value of delivery now, a forward curve tells you the value of delivery next month, next quarter, next year, at whatever granularity the market and the desk require.

Forward curves are the foundation of a long list of calculations: mark-to-market and position valuation, VaR, P&L, hedge optimization, scheduling, structured pricing, and options valuation. In energy markets they are indispensable, because delivery period, seasonality, and operational constraints make a single spot price wholly inadequate to value a book. The curve is how a desk sees the whole term structure of price at once.

Why power and gas markets need forward curves

Power and gas make curve construction harder, and more essential, than in many financial markets, because their physical characteristics shape price in ways a simple curve cannot capture.

Electricity is non-storable, which ties price tightly to the moment of delivery and produces distinct peak and off-peak products, hourly granularity, congestion effects, renewable intermittency, and ancillary-service value. A power curve is not one number per month; it is a shaped structure that has to represent how price varies within the delivery period. See power for the market detail.

Natural gas is storable, which links prices across time through storage economics, and it is strongly seasonal, weather-dependent, and shaped by pipeline constraints and LNG imports. A gas curve has to reflect the summer-winter spread and the cost of carry that storage implies. See gas for more. In both cases, the physical reality is what makes curve construction a modelling problem rather than a lookup.

Market data inputs

A curve is only as good as its inputs, and construction draws on a wide set: exchange settlements, broker quotes, OTC transactions, bilateral trades, forward contracts, futures prices, and options-implied information for the price structure; FX and interest rates for conversion and discounting; and transportation tariffs, weather forecasts, and storage data for the physical shaping.

These come from the market data platform, validated and governed, and feed the construction engine as the raw material of the curve. The quality and coverage of the inputs largely determine the quality of the curve: liquid, well-observed periods produce confident curves, while illiquid periods force the modelling assumptions that the rest of this guide addresses.

The curve construction workflow

Turning inputs into a published curve follows a disciplined workflow, and each stage matters. Raw inputs, exchange prices, broker quotes, OTC deals, plus weather, FX, and storage, are first validated, then normalized into a consistent model. The curve builder assembles the observed points into a term structure; interpolation fills the gaps between traded points; shaping breaks coarse periods into finer granularity; the curve is approved; and finally it is published to the ETRM for consumption.

The reason to formalise this as a workflow, rather than a spreadsheet someone runs, is governance and reproducibility. Each stage, validation, normalization, construction, interpolation, shaping, approval, publication, is a controlled step with a record, so a published curve can be explained, audited, and reproduced. This is the difference between a curve a firm can price real risk against and one it merely hopes is right.

Power forward curves

Power curves are the most structurally complex, because electricity price varies within the delivery period, not just across periods. Construction produces base load, peak load, and off-peak curves; hourly curves; weekend products; and monthly, quarterly, and annual strips, all of which must be internally consistent (the shaped hourly curve must aggregate back to the traded block).

The central challenge is deriving hourly shape from the products that actually trade. The market trades blocks, base, peak, monthly strips, but valuation and scheduling need the hourly profile within those blocks. That shape is derived from traded peak/off-peak relationships and operational assumptions about load patterns, then calibrated so it reconstructs the traded prices. This shaping is where power-curve construction earns its complexity, and where a capable curve engine distinguishes itself.

Gas forward curves

Gas curves are shaped by storage and seasonality rather than intraday profile. Construction covers daily and weekend products, monthly contracts, seasonal strips (summer and winter), and annual strips, along with hub differentials and basis adjustments that locate the curve at a specific point on the network.

Storage economics are the defining feature. Because gas can be stored, the price relationship between summer and winter reflects the cost of storing gas from the low-demand season to the high-demand one, and a gas curve has to embed that seasonal spread coherently. Transportation constraints and hub basis then adjust the curve for location. The result is a curve that reflects not just when gas is delivered but the physical economics of getting it there and then.

Curve shaping

Shaping is the craft of turning coarse traded products into the fine granularity valuation and scheduling need, while preserving consistency with what actually traded. The common transformations are monthly to daily, daily to hourly, peak to hourly, off-peak allocation, weekend adjustments, holiday-calendar handling, and the application of load profiles.

The discipline in shaping is that the fine-grained result must aggregate back to the traded coarse price. If the market traded a peak block at a given price, the hourly shape derived within it must average back to that block. Good shaping uses standardized, documented profiles and calendars so the process is consistent and reproducible rather than a series of ad-hoc adjustments. It is where operational knowledge, when demand actually peaks, how holidays behave, meets quantitative method.

Interpolation and extrapolation

Markets do not trade every delivery period, so curves must fill the gaps, and the method matters. Common interpolation approaches include linear interpolation, cubic spline, monotonic methods, piecewise techniques, and flat-forward assumptions; beyond the last liquid point, extrapolation extends the curve into periods the market does not yet price.

Each method has trade-offs. Linear is simple and robust but can produce kinks; splines are smooth but can overshoot; flat-forward is stable but flat. The right choice depends on the commodity, the liquidity, and what the curve is used for, and a mature curve engine makes the method explicit and configurable rather than hidden. The honest handling of illiquid and extrapolated periods, clearly flagging where the curve is model rather than market, is a mark of a trustworthy construction process.

Curve versioning and governance

Because a curve is a model built on assumptions, governance and reproducibility are non-negotiable. A serious curve engine provides effective dating, multiple versions, a distinction between draft and approved curves, audit history, data lineage, maker-checker workflows, and approval processes.

Reproducibility is the goal that ties these together. For audit, valuation, and regulatory reporting, a firm must be able to retrieve the exact curve version used for any past valuation and explain how it was built from its inputs. A curve published today and a curve published tomorrow are different versions, both retained, so a valuation can always be rerun against the curve that was actually used at the time. Without this, historical valuations cannot be defended; with it, every number is auditable back to its source.

Integration with the pricing engine

Curves exist to be consumed, and the way they connect to valuation is what makes them useful. The pricing engine retrieves the appropriate curve, selected by valuation date, delivery period, commodity, and location, and uses it to value trades, which flows into P&L, VaR, stress tests, and reports.

The critical detail is version selection. The engine must pick the correct curve version for the valuation context, the current approved curve for today’s mark, or a specific historical version for a past-dated revaluation, so that valuations are consistent and reproducible. This tight, version-aware integration between the curve engine and the pricing engine is what ensures the whole valuation and risk stack rests on the same governed curves.

Multi-commodity curve management

A capable curve engine manages curves across the commodity complex, electricity, natural gas, LNG, crude oil, refined products, coal, carbon credits, and renewable energy certificates, under common governance while respecting each commodity’s modelling requirements.

The common thread is governance, versioning, lineage, and reproducibility apply to every commodity, while the shaping method is specific to each. A single curve engine that handles the whole complex under one governance model is what lets a multi-commodity desk value its entire book consistently. See all commodities.

Forward curve KPIs

Curve quality can be measured, and tracking a few KPIs keeps the process honest and reliable.

These targets make curve reliability observable: publication time and availability ensure curves are ready when the desk needs them; validation and missing-data rates ensure they are complete and sound; version traceability and replay accuracy ensure they are reproducible. A curve engine meeting them gives valuation and risk a foundation they can trust.

Why the Gravitas curve engine is different

Gravitas provides a dedicated, governed curve engine built on the principles this guide describes.

Because curves are versioned and integrated tightly with valuation, every mark, P&L, and VaR number is reproducible back to the exact curve used. And it is delivered at economics that suit desks the incumbents priced out. See the quant engine or request a demo to see curve construction on your own markets.