Say I trade energies, fixed interest, currencies and stock indices. How do I manage my risk? Like I might be short usd/yen and long euro/usd but I am still overall short USD vs yen and euro so I might have double the risk rather than 2 separate trades. Or if I am long the ES but short ZN and they tend to be inversely correlated. So I guess really I would like help in understanding how I can spread risk accross multiple markets I guess.... without accidentally doubling down.
as far as fx goes you can spread. For instance, in your example short usd.yen long eur.usd makes you net short usd, so you would put on two other positions which are long usd, like short aud.usd and long usd.cad. And that just creates the cross long eur.cad and eur.aud short aud.jpy and short cad.jpy and the leaves you for the most part net flat usd if that is what you want if you wanted to get long eur but didn't necessarily want to get short usd you could just go long eur against all other majors x usd
Select markets for your portfolio that are uncorrelated with each other (correlation as near to zero as you can find). Keep in mind that two markets with zero correlation will still move in the same direction 50% of the time (most people don't realize this) and they might be more likely to move together when there is a systemic market shock. If you want to lower the risk during a market shock event, you need to look at the correlations during similar periods when these events happened in the past.
the foot bone is connected to the ankle bone and CAD is connected to CL and the ankle bone is connected to the knee bone and CL is connected to SPY and the knee bone is connected to the thigh bone and SPY is connected to USD Now hear the word of the Lord
Is there a fancier way of accounting for increasing correlations? I do not know of manual adjustments to VAR. Do you just watch a table of correlations?
well historical var just runs over historical data and you look at worst cases... when shit hits the fan and correlations all go to 1. i dont know about changing correlations; maybe that can be done in a montecarlo approach when you bump a variable and rerun but thats a guess