Intraday Risk Measurement for Trader PnL-to-VaR Evaluation
Summary
The document asks how proprietary trading firms account for large intraday exposures when assessing annual trader performance with a PnL-to-VaR ratio. It distinguishes end-of-day ex-ante VaR, based on a trader’s positions, from risk estimates based on realized PnL. The answer argues that historical VaR is still an ex-ante measure: past market-variable changes are applied to current instruments, the resulting revaluations are ranked, and a tail percentile is used as the estimate. The respondent describes calculating this daily and annualizing it as their own experience.
For intraday measurement, the answer mentions using five-minute returns and scaling the result to a daily horizon. It cautions that very short intervals can be affected by market microstructure. The response is anecdotal rather than a survey of firm practice, and it does not specify a scaling model, confidence level, or validation method. Its main distinction is between revaluing positions under historical market moves and treating a trader’s own PnL distribution as VaR.
Key ideas
- Historical VaR can revalue current instruments using historical changes in market variables.
- The revalued outcomes can be ranked to estimate a tail loss percentile.
- The respondent describes daily historical VaR that is annualized for performance assessment.
- Five-minute returns are suggested for intraday risk measurement, with a caveat about microstructure effects.
- The document does not establish a universal industry standard or a specific scaling method.
Tags
Full text
# How to handle intraday risk when evaluating performance with PnL/VaR? # How to handle intraday risk when evaluating performance with PnL/VaR? When evaluating a proprietary trader's performance at the end of the year, it's common to use the PnL/VaR ratio as a key metric. From what I understand, the VaR used as the denominator should typically be the ex-ante estimate, often calculated at the end of each day based on the trader's open positions. However, I’m curious about how to handle significant intraday risk exposures. Specifically, do firms evaluate risk at a more granular level (e.g., hourly) and then scale this to a daily VaR? Or is it common practice to use a realized PnL distribution, such as taking the trader’s historical PnL and computing quantiles to assess risk? I would greatly appreciate insights on the common industry practices, particularly in the context of proprietary trading firms. How do practitioners address these considerations when assessing performance metrics? ## Answer by KaiSqDist (score 2) https://quant.stackexchange.com/a/81072 From my experience, even historical VaR is on an ex-ante basis, > Or is it common practice to use a realized PnL distribution, such as taking the trader’s historical PnL and computing quantiles to assess risk? this is not used. What historical VaR is - is using changes in market variables to revalue instruments in their own manner, and the returns generated through the revaluation are ranked and the appropriate percentile is taken to be the historical VaR value. This was done on a daily basis and annualized (in my experience). If we are talking about > Specifically, do firms evaluate risk at a more granular level (e.g., hourly) and then scale this to a daily VaR? I usually hear people use 5-minute returns, but this could be subject to microstructure effects, and I am not exactly an expert on it. But if I had to, I would use 5-minute returns and then scale it up to a day.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.