Societé Generale: "We use a short-term model based on a concept of quantile regression to identify the extremes that EUR/USD could reach. To understand what a quantile estimation is one should first look at a regression. A regression finds the mean estimate with a certain error assuming a normal world. This would be the same result found when minimizing the moment of the distribution. These moments define uniquely a normal distribution via the mean and variance.
A quantile model concentrates on only one part of the distribution or equivalently moment.This is useful
for considering how a model would behave under more extreme scenario, presuming there is something to learn from past history. The 95% quantile estimate is 1.3420 where major stops losses can be found as it also conveniently enough a key technical level.
Even a quantile model has its limits and in the past we have seen EUR/USD overshoot even the most extreme quantile estimates by 2.5% (1.3630) and 5% (1.3970).
This suggests that one should sell EUR/USD between 1.3420 and 1.3630, while selling the risk of a move above 1.40."