摘要: The evaluation of market risk corresponds essentially to a forecast for the lower tails return distribution (i.e., losses). available methodologies are broadly classified into return-based and innovation-based methodologies. limitations main methodology—the historical method—are discussed. more powerful, since they disentangle better components price dynamics, connection with processes allow one compute required forecasts. Depending ARCH process used model financial time series, different can be constructed, various trade-off between simplicity accuracy. difficulties related practical evaluations risks discussed target portfolio sizes order 10.000 complex contracts. Finally, should validated by proper backtesting procedure. A simple powerful test is proposed applied century Dow-Jones index. Surprisingly, procedure works perfectly over such long span history.