I love reading and listening to people who make finance sound like playing with Lego.
Everything clicks together and is so self explanatory.
Meanwhile, I am wracked by anxieties, because nothing I 'know' about finance really works consistently.
Imagine buying a croissant at your local bakery every morning, and every other day you get a burnt or slightly stodgy croissant in return.
Not much consistency on offer.
So when I added the 'Yearly Probability of Loss' analytic to the Lazy Backtesting IDE on the one hand it alerts people to the possibility of ending up with a burnt croissant in your hands, on the other the measure is too finely detailed for our inconsistent field.
For example the Realised Steady Vol strategy reports a yearly probability loss of less than 0.1%, or one in a thousand.
That means one bad croissant in a thousand.
Furthermore, we don't have over a thousand years to test out the claim properly.
Not to mention, out of 55 years of market history that are available, only 7 in 10 years were actually profitable!
Sharpe's t-test for his ratio gets closer and closer to normality the longer our back test period goes on.
Why is normality an issue?
Normality implies quite consistent market returns; but obviously market returns rarely results in consistency. Bad results often cluster together.
Rather than throwing the baby out with the bath water though, it would be nice to modify this test a bit, and come up with something which reflects the lack of consistency better.
Watch this space.