What do you have when you take a useless idea and make a slight change? Like starting with a 0 and then doubling it, or tripling it, or multiplying it by 100 – you’re still left with 0. They say there’s no stopping an idea whose time has come, but it seems like it’s awfully hard to kill ideas whose time has passed.

Speaking of which – remember when JP Morgan suffered that huge trading loss earlier this year? Remember how turned out that they were using a measure called Value at Risk (VaR) with a few changes the way they calculated it, but after their losses they decided to revert to the older model? It looks like the lesson still hasn’t been learned. FT Alphaville has the breakdown:

Buried in Morgan Stanley’s decent third-quarter results (excluding the absurdity that is DVA of course) is this intriguing footnote:

“Morgan Stanley’s average trading Value-at-Risk (VaR) measured at the 95% confidence level was $63 million compared with $76 million in second quarter of 2012 and $99 million in the third quarter of the prior year. The Firm modified its VaR model this quarter to make it more responsive to recent market conditions.”

The folks over at FT Alphaville have also assembled a handful of great charts to show how the new calculation of VaR “lowers” Morgan Stanley’s risk.  Or more correctly, lowers the amount of risk Morgan Stanley believes is inherent in their positions at the 95% confidence level.  But let’s be completely serious for a moment: does anyone really believe that their risk is that much lower now than it was? They’ve essentially made a tiny change to their calculations (changing how many years back it looked) and voila! The world is now safer! And if you believe that, we’ve got some land in Florida to sell you, too.

But setting aside the fact that these banks are creating a shifting target by constantly updating their models, we’re left with an even bigger question – is anyone convinced that the VaR models are actually a useful measure of risk in the first place?

If we look at VaR in a little more simplistic way (and not entirely correct statistically, but good enough for an example) –  there is a 95% chance they won’t lose more than $60 million dollars in a day, and a 5% chance they will lose more than that in a single day. Problem is – this sort of risk model says nothing about how much could be lost on the rare 5% of cases where things go wrong in a hurry. Does anyone find that reassuring? Or can we finally agree to put some more robust risk metrics in place?