IMO VaR is only applicable if you're running abook of multiple positions, and even then it has its limitations. Try hunting around for CrashMetrics or RiskMetrics on the web, I think there is some data you can download...?? Anyhow... building a proprietary model for they type of trading I guess we all do (i.e. simple long or short until out, not pairs trading or carry trades or model based trades) is a lot of work for little return. I guess the crux of the matter is that to develop a sound model of your risk, you need a sound model of how the asset behaves - A sound model of how the asset behaves is keep slots of outrageously bright people quiet, so it is safe to assume it isnt something like "buy it on a bounce off support, close it at XXXX, take profits at YYYY".
IF you wanted to, I guess you could use some bootstrappping ot Monte Carlo simulations to give an indication of liklihood price will reach your target - if you think it will be any use, which I doubt.
Also, I think there is an important clafification to make here: do you want a way to monitor your own performance, such as risk adjusted returns, or are you looking for something that will give you an idea of your "expected" risk for any given trade?
If the former, then look up Sharpe ratio's, Kelly Criterion and such...
If the latter... well, to be honest, i wouldn't bother. You might like to look at building a GARCH or EWMA model of each asset you track, and take it from there - though, I have to say, i don't think this will be any use either. Modelling asset prices as stochastics, Markov's etc.. is all good and well if you are continually in positions. Prices are driven by buying and selling, they are not random walks a priori, it is just that random walks do very well as a model. You are most concerbned about risk. volatility etc.. when you are in a trade, you will only be in a trade when you can determine the emerging pattern of buying or selling.
So, in short:
Sharpe Ratio's, Alpha if your trading stocks etc.. , Kelly Criterion.
You could use VaR, but unless you're running multiple positions regularly, its over the top by a country mile.
GARCH etc.. are useful if you are trading from models of asset behaviour, but it strikes me as a clear contradiction in terms: why would you model your risk on a random walk, when you are only at risk when prices aren't "randomly walking", they are being driven by excess demand or supply. Perhaps there is something in building a model that shows you when the RW model isn't applicable anymore?? Don't know, just thinking out loud.
