SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
greek2
Topic Author
Posts: 23
Joined: November 22nd, 2003, 2:25 am

VAR backtesting

August 14th, 2009, 2:10 am

Doing some backtesting whereby we compare Value at Risk vs. the trading PnL. We use a 95% VAR so we would expect to see some data points each year whereby the PnL would fall outside the VAR boundaries. However, we are seeing no such exceptions (too good to be true).Any ideas? has anyone else come across this?thanks
 
User avatar
jomni
Posts: 999
Joined: January 26th, 2005, 11:36 pm

VAR backtesting

August 14th, 2009, 2:36 am

Traders are so good that they don't incur spectacular losses?How about a hypothetical test where you compare against hypothetical PnL (keeping portfolio the same but using next day's prices).This method is more in line with VaR method's limitations (static portfolio).
Last edited by jomni on August 13th, 2009, 10:00 pm, edited 1 time in total.
 
User avatar
rmax
Posts: 6080
Joined: December 8th, 2005, 9:31 am

VAR backtesting

August 14th, 2009, 6:47 am

Are you looking at both upside and downside - you should break your limit as well with gains.Are you also comparing the correct clean trading PL. You can only backtest if you strip out new trades, terminations etc - this bit is pretty obvious however you also need to understand what else VaR includes / does not include. Theta is one example.
 
User avatar
freddiemac
Posts: 557
Joined: July 17th, 2006, 8:29 am

VAR backtesting

August 14th, 2009, 2:30 pm

Perhaps your VaR model is bad and it over estimates the risk. It is not necessarily good to have no returns that exceeds VaR, it might suggest that you have a model that does not correspond to the actual risks. Another explanation might be that the traders are not "doing their job" - ie playing it safer than management wants.
 
User avatar
riskguru
Posts: 79
Joined: August 11th, 2004, 4:24 pm

VAR backtesting

August 14th, 2009, 3:25 pm

If the traders are not taking enough risk that would presumable get picked up in the VaR right??
 
User avatar
pcg
Posts: 194
Joined: September 13th, 2004, 11:11 am

VAR backtesting

August 14th, 2009, 3:32 pm

Mos probably what is happening is that the last 500 days volatility at 95% is way above current market movements and hence your VaR is overstating the P&L .Try using an EWMA modified VaR.
 
User avatar
freddiemac
Posts: 557
Joined: July 17th, 2006, 8:29 am

VAR backtesting

August 15th, 2009, 8:31 am

QuoteIf the traders are not taking enough risk that would presumable get picked up in the VaR right??Correct, my bad.
 
User avatar
pcg
Posts: 194
Joined: September 13th, 2004, 11:11 am

VAR backtesting

August 19th, 2009, 2:51 am

I am sure traders have short memories so the problem won't last long . With some market stability positions will build and VaR will go down and you will start gettting backtest failures.
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On