1. What P&L explain tries to answer

Daily question on any trading floor:

“We made/lost X today. Why? Was it markets, positions, model, or booking?”

P&L explain is the bridge between:

  • Risk view (sensitivities, scenarios),
  • Trading view (positions and trades),
  • Finance view (booked P&L).

If this bridge is weak, nobody trusts the numbers.

2. A simple decomposition

For one day (t → t+1), for a portfolio:

$$ _{}

  • _{}
  • _{}
  • _{} $$

Interpretation:

  • Risk P&L: What yesterday’s sensitivities predict from today’s market moves.
  • New trades P&L: Impact of entering / exiting trades.
  • Model / data P&L: Curve rebuilds, calibration changes, fixing corrections, methodology changes.
  • Residual / unexplained: Everything else – should be small and monitored.

The ideal world:

  • Risk P&L + new trades + model/data ≈ actual P&L, residual ≈ noise.

3. How to compute the “risk P&L” term

Given:

  • Yesterday’s factors $ f^{(t)} $,
  • Today’s factors $ f^{(t+1)} $,
  • Sensitivities $ s_i^{(t)} $ to each factor $ f_i $.

Approximate:

ΔPVrisk≈∑isi(t)Δfi=s(t)⊤⋅(f(t+1)−f(t)) \Delta \text{PV}^{\text{risk}} \approx \sum_i s_i^{(t)} \, \Delta f_i = s^{(t)\top} \cdot (f^{(t+1)} - f^{(t)})

On a power/commodities book, factors might be:

  • Forward buckets (Cal-27 base, Winter-28 peak…),
  • Basis (zone vs hub),
  • Vol buckets,
  • FX, CO₂, fuel curves.

Refinements:

  • For bigger moves, include gamma / curvature.
  • For optionality, use revaluation at shocked factors rather than pure linear.

4. P&L explain at portfolio level

You compute this per trade, then aggregate:

  • By book / desk / trader,
  • By product type (vanilla, storage, swing, plants),
  • By factor set (level, shape, basis, vol).

Outputs:

  • How much P&L came from:
    • Curve level moves,
    • Shape moves (winter–summer, base–peak),
    • Basis,
    • Vol,
    • New trades / restructurings,
    • Model / data changes,
    • Residual.

Risk and desk then ask:

  • Do big moves match the risk reports and positions?
  • Are residuals small and noisy, or biased and systematic?

5. Why residual P&L is important

Persistent residuals can signal:

  • Wrong or stale sensitivities (bad mapping, wrong bumping),
  • Missing risk factors (e.g. unmodelled basis or shape component),
  • Booking errors (mis-booked trade legs, wrong notionals),
  • Model issues (non-linear products with poor linear approximation).

So residual ≈ model + data + booking problem detector.

On a good desk:

“We treat residual P&L as model risk P&L, not alpha. If residual grows, we investigate.”

6. Power/commodities flavour

Special features:

  • P&L driven by curve shapes and spreads, not just levels.
  • Optionality (plants, storage, swing) means:
    • Exercise decisions can change day to day,
    • Some P&L is from changed optimal exercise, not just factor moves.
  • You may have dedicated components like:
    • “Intrinsic vs extrinsic” change,
    • “Exercise / dispatch decision change” P&L.

But the core decomposition idea stays the same.

7. Interview soundbite

“P&L explain decomposes daily P&L into risk-factor moves, new trades, and model/data effects, with a residual bucket. We start from yesterday’s sensitivities to a stable set of factors, apply today’s market moves, and get a predicted risk P&L. New trades and calibration/curve changes form separate components. The residual is monitored as a model and booking control signal – on a power/commodities book it’s especially important to capture curve shape and spread risk, otherwise those residuals explode.”