Var calculations ? Expousres

My VAR is the difference between my entry point and my stop. :)

In all seriousness, though, even the bankers who supposedly understand VAR obviously don't know how to use it as witnessed by what we've seen the last couple years.
 
Cross market correlation was one of the things that interested me most when I first started looking at charts. But with my limited brain capacity I decided that there are just too many variables for me to begin to comprehend the complexity of the equations involved. Different weather patterns throughout the world, cultures, geopolitics and government intervention... it all adds up.

Saying that, the weather man does tell you its going to rain sometimes... Not the best analogy given we're a small island on the edge of a giant ocean but still.
 
Utterly trite, but not without a grain of truth for all that. imho VaR only tells half the story. You can paint as accurate a picture as you like about what will happen to your balance sheet if the market does something out of the ordinary, but that's no good without knowing what the chances are of this actually occurring.

I'm in agreement with the basic premise of your post, and in your book recommendation, but to my mind the issue with VaR was more on the impact side of things than the probability side. VaR is basically "What kind of portfolio impact falls within the 95% (or whatever) confidence level?" Granted, the odds may be incorrectly determined (as they no doubt were), but the big problem was that they never accounted for the fact that the area beyond the 95% was potentially a "there be dragons" type of risk zone - meaning not just portfolio destruction, but total company wipeout.
 
FT-Maths and Markets

I liked this article. In particular the paragraph

The standard risk measure used by the industry from the mid 1990s, known as value-at-risk or Var, was criticised by mathematicians almost from the start for the way it drew inferences about forward-looking risk from past patterns of price movements. As a result, the risk of extreme bank-shattering events was greatly underestimated.

Yes Var was known to have limitations, but you have to do something right? I think mathematics can often give people a sense of security, which they should not have unless they really understand what the maths can and can't do. And for that you would also need an appreciation of trading too. At the highest level of banks, how many are there that have a strong understanding of trading and/or a strong understanding of the quantitative side? As we see from some of the banks, they put too much trust in something they likely didn't understand the limits of, or as has been said above, didn't really understand the probability of extreme events or the effect of such events. The same is true of LTCM, but they should have known better given the mathematicians involved at a high level in that. Maybe greed took over or maybe those mathematicians didn't understand the market well enough.
 
Last edited:
I dont think VAR is a bad tool at all, the examples you quote are largely because mathematicians do the calc and undertstand it , but management dont understand either the implications / assumptions nor the value of the calc (i.e. it is not the only tool). In my experience there is too much reliance on the computation and the models are not really subject to any stress testing.

The other thing you have to be aware of particularly using historical VAR is that it won't cover for example global flu epidemics, large wars or large scale frauds for example. It is as good as measure you are going to get about forward looking risk as you would using any other model. (all which are subject to assumptions)

This is also what happened to Metalgeshaft in the 80's through derivatives. Its not so that the technique/ calculation / instruments did not work........its just that management did not understand it.

I've been doing a var or potential var before I take on a position, it is based on historical data so is a historical var (more so because I don't know fully how to model the a Monte Carlo simulation). The problem I have is that this is a fairly painful task and I wish one of the index providers would provide a mechanism to compute this - even if it as simple as a historical analysis of outcomes. It may be crude but to be honest its a good as picture you will get even if you employ a oxbridge mathematician.

What Rhody has pointed out is quite important as you get an appreciation of the downside risk.

I don't employ stop loss as volatility is nasty stuff - you only need to look at the S &P over the last few days.

The var provides a mechanism to limit the risk position I'm taking, particularly as I increase the number of positions that are open. i.e. if the s**t hits the fan will I be sprayed in the process.

I am a but surprised by Gamma Jammas comments that traders a lot of these (private) day traders don't use var and I would conclude from there may not understand their risk positions in a probability sense before undertaking trades.......probably why the stats say that most of the day traders make losses and very few end up making gains in the long term.
 
I don't employ stop loss as volatility is nasty stuff
But you use a mental stop, right? If not, then how does VaR give you an accurate measure of the risk you're taking. Sure I understand that you might calculate your VaR and you arrive at a 1% chance you'll lose more than X amount in some time period, but that doesn't tell you whether in the event of that 1%, you lose just over X, 2X or 10X. I personally prefer stop losses, because it is clear exactly how much I can lose.

I am a but surprised by Gamma Jammas comments that traders a lot of these (private) day traders don't use var and I would conclude from there may not understand their risk positions in a probability sense before undertaking trades.......probably why the stats say that most of the day traders make losses and very few end up making gains in the long term.

Doesn't this depend on how you are trading? For someone who only trades GBPUSD for example, they have their stop loss and their position size, so they should be well aware of what they are risking. That is what the stop loss does for them, and they don't really need to know about VaR. If you have a portfolio of assets, it might be different, but stop loss can still let you know what you're risking. And I wouldn't think that VaR is the reason why most people lose. Just my opinion.
 
Why would technical traders want to use VaR in the first place? Underling VaR is the assumption that asset prices are Generalised Wiener processes and the returns are lognormally distributed; technical traders modus operandi is that there are times when they are not.
 
I'm not really a technical trader, my main strategy is to buy / sell equity and hedge out market risk hence the need for var as multiple positions open / closed over a longer duration.

I also write options an again understanding potential exposure to these highly risky and volatile creatures is necessary.

Aaronmalins var is measure of risk not return. You effectively trying to gauge the probability of bankruptcy should things go really badly. Your not trying to figure out your chances of being a millionaire
 
probably why the stats say that most of the day traders make losses and very few end up making gains in the long term.

There are a million reasons why most day traders don't make money, and the understanding of VAR doesn't even make the top 2/3 of those. Why would someone who makes trading decisions in seconds rather than hours even consider this? Try developing an understanding of what is happening intraday; that way you won't depend on models making your decisions for you and ultimately letting you down.
 
VAR or some variant thereof is a useful tool, if treated with caution and understanding. I don't see any problem with using techniques like VAR for higher-frequency trading, either.

It's all a matter of understanding the tool and its flaws and limitations and remembering that nothing is perfect. In fact, VAR and other statistical models evolved precisely because humans are known to have systematic biases and to make predictably wrong decisions. The original premise, which is still very valid, in my view, was to try to augment 'human' risk management methodologies with quantitative ones, thus having the best of both worlds. As usual, the effort made the pendulum swing too far and the quantitative methods came to dominate. But what we're engaging in (and that's why I have a problem with Taleb) is no better, as we're just slipping into the opposite extreme.

My 2c...
 
If more than 50% of the market is making the wrong decision, that decision becomes the right one. I'd never use any indicator or quantitative measurement that told me when to enter or exit a trade.
 
If more than 50% of the market is making the wrong decision, that decision becomes the right one. I'd never use any indicator or quantitative measurement that told me when to enter or exit a trade.
Disagree with your point regarding wrong being made right. Right/wrong is only meaningful for a particular time horizon, which means that decisions can be right and wrong at the same time. That's the basis of all liquidity providing strategies. Your second statement suggests to me that you're either very good or very deluded (i.e. you're either able to overcome your biases or not aware of them)...

I use a whole lot of quantitative techniques, including indicators that suggest entering/exiting trades. I truly don't see what makes quantitative methods (assuming they're based on sensible logic) different from any other heuristic that you may use to make decisions.
 
I'm not saying they don't have merit, for certain types of trading. It's just my personal opinion that having a feel for price action and market sentiment will be infinitely more valuable than an MA cross, evening star, or any other 'signal' that so many traders spend their lives searching for.

Let me also add that I spend many hours a day staring at the ladders, which isn't possible for many traders - I should qualify my statement and say that I don't believe intraday traders should use any form of signals to enter or exit trades.
 
I'm not saying they don't have merit, for certain types of trading. It's just my personal opinion that having a feel for price action and market sentiment will be infinitely more valuable than an MA cross, evening star, or any other 'signal' that so many traders spend their lives searching for.
Ah, I see... Agreed on the price action/mkt sentiment. As to the dojis, morningstars and other TA indicators you're describing, those are not the methods I am referring to, since I cannot discern any sensible logic behind them. I am talking about statistical techniques that are used, for example, in relative value trading.
 
OK, so I believe that quantitative analysis of historical tendencies definitely has merit, because then you're talking about the analysis of previous human behaviour. I still see it only as a background to what you're seeing happening in front of you. With regards to analysis of value, I'm of the personal belief that the fair value of anything is what someone is willing to pay for it.
 
Last edited:
An example: There's an oldish stat that says something along the lines of '85% of the time the Bund will make its daily high/low in the first 15 minutes'. OK, so if you had infinitely deep pockets you might be able to make a living exploiting something like this; I still believe you'd be better off just knowing this in the very back of your head, and trading what you see happening in front of you.
 
Top