Core Path · L6.2.12

Building an FRTB internal model

14 minFRTBInternal models approachRFETNMRFSESIMCCModel change

By the end you will be able to

  1. 1Decide whether a trading desk may use an internal model, from its backtesting exceptions and its P&L attribution zone together.
  2. 2Run the risk factor eligibility test on real observation records and see which factors become non-modellable, and why.
  3. 3Compute stress scenario capital for non-modellable factors and apply liquidity horizons through the cascade.
  4. 4Assemble the IMCC and the capital, max(latest, m_c × average), for the sleeve and set it beside the booked figures.
  5. 5Judge what a model change does to capital, and why the answer is often smaller than expected.

Why it matters

Every piece of this module ends up here. The sleeve's historical simulation passes its desk backtests and its P&L attribution test, so it may use the internal model. Two of its 26 risk factors fail the data test: the sterling OIS curve, because this course holds only a month-end panel, and the option volatility, because its quotes stop in June. They come out of the model and are charged by stress scenarios. Put together, the sleeve's illustrative FRTB capital is £20.50m. Buy daily data for both factors and it falls by only £0.33m.