Sharpe Ratio for prop deriv desks

[proption]

Hi
Anyone know if there is an 'accepted' format for calculating the Sharpe Ratio if you work on a prop desk at an investment bank? It is not straight forward to know what the starting capital is, like it would be fora hedge fund. Does one relate the vol of returns to either

1. your VaR limit
2. your total stop loss (where you stop trading for the year)
3. a solvency figure (or amount of capital set aside with which to take risk, related to VaR)

Any hints greatly received....

 

[rufus_4000]

Quote from proption:

Hi
Anyone know if there is an 'accepted' format for calculating the Sharpe Ratio if you work on a prop desk at an investment bank? It is not straight forward to know what the starting capital is, like it would be fora hedge fund. Does one relate the vol of returns to either

1. your VaR limit
2. your total stop loss (where you stop trading for the year)
3. a solvency figure (or amount of capital set aside with which to take risk, related to VaR)

Any hints greatly received....

I have quite a few friends who are on prop desk at ibank these days. And it is hard for me (a professional independent trader) and them to compare notes such as RR, sharpe ratio, etc, since the concept of cap usage is a bit foreign to them (although we are good enough friends that we can talk about specific strategies in detail!). Your item (2), stop loss, doesn't apply to some strategies, (3) is also not commonly done for prop desks, risk managers don't always compute VaR at the level of detail, the tolerances are only adjusted so often.

The only thing that the prop desks and I have in common are the periodical gross P&L computations (daily, weekly, etc) (forget about commission calc for prop desks even, since their access to the markets are essentially free), and the relative absolute (including hedged) portfolio marked to market, so we can get some idea of the daily P&L volatility, which essentially what Sharpe ratio represent. So the ones that I know would compare their RR to a relevant benchmark, and "rough" calculate return vol from there.

Rufus