Topic 8 – Term Structure, Seasonality, and Shape Risk (Power & Commodities)

1. What is a term structure here?

In commodities we almost never talk about “one price”. We talk about a forward curve:

• For each delivery period $T$ (day, month, quarter, year)
• There is a forward price $F(0, T)$.

Plot $T \mapsto F(0,T)$ → term structure.

Examples:

• Gas: winter months more expensive than summer.
• Power: strong seasonality (winter demand, summer cooling), plus weekday/weekend and hourly structure.

Key point: Risk is not only “level of the curve”, but also its shape over time.

2. Seasonality: level changes over the year

Seasonality = systematic pattern by calendar:

• Winter vs summer: heating/cooling demand.
• Weekday vs weekend: industrial load.
• Peak vs off-peak hours: daily demand shape.

In forwards, that shows up as:

• Cal-27 (year) price averaged over all hours.
• Winter-27 strip > Summer-27 strip.
• Certain months (e.g. Jan, Feb, Aug) systematically trade richer.

Practitioner lens:

“Even if overall price expectations are unchanged, the market constantly re-prices when energy is valuable inside the year.”

3. Shape risk: level vs shape moves

Imagine a simple yearly forward curve:

• Today:
  • Summer months ~80
  • Winter months ~120
• Tomorrow:
  • Everything moves to 100 (flat).

Two different risk moves:

1. Level move: Everything up/down by ~10 EUR/MWh.
2. Shape move: Winters up, summers down, but average stays the same.

Shape risk = P&L driven by relative moves between buckets, not by the overall level.

Examples:

• Long winter, short summer (classic calendar spread):
  • Flat level move → almost no P&L.
  • Shape move (winter up, summer down) → big P&L.
• Supply contract with flat customer tariff but seasonal hedge:
  • Mis-hedging shape gives you residual exposure even if annual volume is correct.

4. Why term structure and seasonality matter for a quant dev

From a system / model point of view:

• Pricing engine must understand:
  • Contracts that deliver over periods (month, quarter, year).
  • How to integrate the hourly/period forward curve over delivery.
  • How to report risk by buckets (months, seasons, shapes).
• Risk layer must distinguish:
  • Level factor: parallel shift of all tenors.
  • Slope / curve factors: steepening, flattening.
  • Seasonal factors: winter vs summer, base vs peak.

For an interview:

“In power, P&L often comes more from shape changes than from parallel moves. A good risk system decomposes exposures into level and shape factors so traders can see where they are long or short seasonality.”

5. Simple modelling intuition

A toy seasonal forward curve:

$$F(0, t) = L + A \cdot \sin(2\pi t) + \text{noise}$$

• $L$: overall level.
• $A$: amplitude of seasonality.
• “Shape risk” = sensitivity to changes in $A$ and to local deformations of that seasonal pattern.

You can imagine:

• Scenario 1: bump $L$ by +10 → parallel shift.
• Scenario 2: bump $A$ by +10 → winters richer, summers cheaper.

Different portfolios will respond very differently.

6. How to talk about it in an interview (short)

You can summarise like this:

“In commodities we price not just a point, but a whole forward curve. Seasonality makes winter and summer, weekdays and weekends, base and peak blocks trade at different levels. Shape risk is the risk that the relative prices of these buckets move, even if the overall level stays the same.

A structured portfolio may be roughly flat to a parallel move but very long or short certain seasonal shapes. The analytics stack must therefore expose risk by time buckets and factors like level, slope, and seasonal spreads, not just a single ‘price’.”