deletions | additions       diff --git a/untitled.tex b/untitled.tex  index 1da3ca1..f9ff823 100644  --- a/untitled.tex  +++ b/untitled.tex 
...
Combining these two phases gives us a total of 291 doublings, 123 additions and 23 negations.  If our Elliptic Curve representation makes addition as cheap as doubling (which some do), and we ignore the negations (which are comparatively cheap), then the base-32 method turns out to be 5.6\% faster than the base-48 method. In other words, from this perspective, we can implement the blinding on a special format prime, and be within 6% 6\%  of the performance of a random prime. These results are fairly stable if we tweak our assumptions (e.g. change the size of the group order or the size of $r$) \begin{itemize}  \item Generating exponents into non-power-of-2 base 
...