Why FRTB exists
During the 2007-08 financial crisis, trading-book losses dwarfed the capital banks were holding against them. The interim 'Basel 2.5' fixes bolted stressed VaR and an incremental risk charge onto a framework that was never designed for that stress, producing capital numbers that were inconsistent across banks and easy to game.
The Basel Committee on Banking Supervision responded with a ground-up rewrite, first published in 2016 and finalised in 2019. The goal was a market-risk framework that is more risk-sensitive, harder to arbitrage, and more comparable from one bank to the next. Implementation has been staggered by jurisdiction, with the EU, UK, Canada, Japan and others live from 2025 and the US still finalising its approach.
Expected Shortfall replaces Value at Risk
The headline change is the move from 99% VaR to 97.5% Expected Shortfall. VaR answers 'what is the most I expect to lose on 99 of 100 days?' but says nothing about how bad the worst 1% gets. Expected Shortfall instead averages the losses beyond the 97.5% threshold, so it is sensitive to the depth of the tail — exactly where crises live.
FRTB also calibrates ES to a period of significant financial stress and applies liquidity horizons of 10 to 120 days depending on how quickly a risk factor can realistically be hedged or exited. That makes the capital number reflect how long a bank would actually be exposed, not an idealised one-day unwind.
Standardised vs Internal Models Approach
Every bank must be able to compute the Standardised Approach (SA). It is a sensitivities-based method: delta, vega and curvature charges across defined risk classes, plus a default risk charge and a residual risk add-on. The SA is now risk-sensitive enough to be a credible primary measure, not just a fallback, and it doubles as a transparent floor and reporting benchmark.
Banks that want capital relief can apply for the Internal Models Approach (IMA), but the bar is high. Approval is granted desk by desk and depends on passing two ongoing tests: a profit-and-loss attribution (PLA) test that checks the risk model's P&L tracks the front-office P&L, and backtesting. Risk factors that lack sufficient real price observations become Non-Modellable Risk Factors (NMRFs) and attract a separate stressed capital add-on, which is often a decisive part of the total charge.
The trading-book / banking-book boundary
FRTB hardens the boundary between the trading book and the banking book with prescriptive assignment rules and strict limits on re-designating instruments after initial classification. The intent is to stop firms parking positions on whichever side of the boundary carries the lower capital charge.
In practice this forces governance: documented desk structures, clear policies for what sits where, and approval workflows for any internal risk transfer. The boundary is as much an operating-model question as a quant one.
What FRTB means operationally
FRTB is a data and computation problem before it is a capital problem. The IMA needs years of clean risk-factor price history to classify factors as modellable, daily PLA reconciliation between risk and finance, and the ability to run Expected Shortfall across multiple liquidity horizons and stress windows. The SA needs a complete, governed sensitivities feed.
The firms that cope best treat FRTB as a unified market-risk operating loop: one trusted source of positions and sensitivities, reproducible model runs with full lineage, and a clear audit trail from a regulatory capital number back to the trades and risk factors that produced it. That is exactly the discipline regulators expect to see during model approval and review.