The paper provides a methodological contribution to the multi‐step Value‐at‐Risk (VaR) and Expected Shortfall (ES) forecasting through a new adaptation of the Monte Carlo simulation approach for forecasting multi‐period volatility to a Fractionally Integrated Generalized Autoregressive Conditional Heteroscedasticity (FIGARCH) framework for leptokurtic and asymmetrically distributed portfolio returns. Accounting for long memory within the conditional variance process with skewed Student‐t (skT) conditionally distributed innovations, accurate 95 per cent and 99 per cent VaR and ES forecasts are calculated for multi‐period time horizons. The results show that the FIGARCH‐skT model has a superior multi‐period VaR and ES forecasting performance.