Talk:Tips and Tricks/@comment-187.189.173.144-20140814174124/@comment-137.207.75.220-20140912001507

Well this makes some good sense if you think about it

Let's make an assumption of crystal invoke giving us a B/A/S&SS distribution of 60%/30%/10% just for easy math, which seems roughly what I'm getting anyway. This distribution doesn't really matter though, the end result will hold true regardless.

A mega-invoke would then give us:

5.4 B (9*0.60)

3.7 A (1+ 9*0.30)

0.9 S/SS (9*0.10)

rerolling this won't change the average distribution, but you may always get lucky. It pretty much guarantees an A, but also that you'll be stuck with a lot of crap

10 single invokes, where you reinvoke any Bs you roll would give you

3.6 B (10*.60*0.60, the amount from rerolling)

4.8 A (10*0.3+6*0.3, 3 from original, 1.8 from rerolled Bs)

1.6 S/SS  (10*0.1+6*0.1, 1 from original, 0.6 from rerolled Bs)

End result; rerolling single invokes is statistically better, but only when single reinvocation is available