It's his day off.

CDF shows the upper or lower expectation of getting the value in question in a Poisson distribution.

Note there's 2 commonly used - by ooc and me at least

There's the straight CDF of a single event

and the CDF[Erlang] of a set of events.

https://kano.is/index.php?k=pblocksSo like on my pool blocks page, the CDF on the right of each block shows the expectation that the Diff would occur better (lower) than the block's Diff

The most recent block has a Diff of 110.597% and a CDF of 0.669

i.e. ~2/3 of the time you'd expect each block to do better.

The top table shows the expectation that the number of blocks would get a lower Mean% Difficulty.

In this case it's a very different result coz you are looking at a set of results, rather than just one.

i.e. if the pool found 100 Blocks, what's the chance it would do better?

In my pool's case, the last 100 CDF[Erl] is currently 0.3382 (Mean% Diff 95.56%) which is super good and beats the crap out of shitty pools like Eligius

That 0.3382 means that you'd only expect that to be better ~1/3 of the time for a range of 100 blocks.

As the number of events increases, the CDF[Erl] will decrease for the same Mean% since it becomes rarer that you'd continue to get the same luck for more events if the Mean% Diff was below 100%