Over twenty years ago I recall having a discussion with fellow dealers about how big our risk limits should be. We concluded that it was straight forward. The risk limits ought to be based on how much the bank could afford to lose. Whilst 14 standard deviation events were known about, for limit purposes they were ignored. Why? Largely because if you based your business on extreme events, then 99.99% of the time you would be taking less risk than you considered appropriate. This view was reinforced by the fact that even if it was a very bad day the size of the risk would not jeopardize the bank because trading was only a very small proportion of the bank's activities. Fast forward twenty years and two things have altered. First, the exposure banks have to fluctuating market prices is now far greater. This is caused by a number of factors including the type of business conducted, leverage and accounting rules. Second, "fat-tails" are an accepted part of market behaviour and must be assessed and quantified by using stress testing. Put these together and on a bad day you lose a lot more money than anticipated. A fact not lost on regulators who are significantly increasing the amount of capital and liquidity banks are being asked to hold. So how do you look at risk? Do you see it as what happens 99.99% or 0.01% of the time? If you move from the former to the latter it makes you much more aware that your risk limits should be moving down rather than up. The natural corollary of this is that profits from risk taking will fall. This puts greater value on earnings from the franchise.