Hi,For historical var, there are lots of literature. However, few things are still unclear to me. To simulate historical scenarios, one can have few choices to generate scenarios.a. relative returns (simple, log, et al)b. absolute changesc. mixture of aboveSome of problems with a,b methods are outlined here Displaced Historical SimulationI wonder if anyone have opinion, maybe additional links? I am specifically interested in energy commodities.Thank you.Mephist.

There are lots of choices, but in practice you tend to only see simple things done... unfortunately I have no ready at hand papers on this subject, just messy experience with calculations.The largest issue here IMO is how to deal with spreads/basis: relative shocks separately on location A and location B prices does not treat the spread very intelligently, and if you have a portfolio that does not have any outright price risk then you you can easily get nonsense. So often you define outright prices somewhere (HH usually) as a risk factor and (somehow) spreads on top of that as additive risk factors; data is a challenge.Also, be careful with seasonal commodities -- you need to avoid contract rollover nonsense by calculating shocks from the same contract, not the same offset contract.Err... and power has its own problems...