Logit
The logit (short for "logarithm unit") is the logarith of p/(1-p), called the "odds", which is intuitive because the odds are better when p is larger than (1-p).
I used the logit as a transformation in Bayesian Networks . When comparing two sets of probabilities, it may occur that most probabilities are either close to 0 or 1. In this scenario, it may help to apply the logit transform on all probabilities (base 2, e, 10 are common) to get a clearer visual indication of similarity, plotted against a diagonal line.
E.g. in R:
plot( log10(p.lr/(1-p.lr)), log10(p.optbayes/(1-p.optbayes)),
bty="l", pch=19, xlim=c(-5,5), ylim=c(-5,5) )
abline( a=0, b=1 )