!!!!
Taking what Spark taught me about frame data I was able to do some testing with d cancels after her c series, meaning we have more frame data.
Was never confirmed before, but she has a 4 frame jump startup just like the most of the cast.
ON NORMAL BLOCK
Ch state ends on the frame after the first number, and Mu can chicken block on the frame after the second number
5d: -4/-8
2d: -6/-10
4d: -6/-10
6d: -6/-10
5c>5d>7AB is -(=/<)4, throw immune on frame 5, air blocking on frame 9
5c>2d>7AB is -6, throw immune on frame 7, air blocking on frame 11
5c>4d>7AB is -6, identical to 2d
6d identical to 2d
I'm not sure if 5d is completely accurate, I'm only so good at parse-frame imputs, but I did run it a good number of tests.
Bascially, if all that info is completely correct, you can cut a frame off the recovery of 5c AND get a stein with 5d. Going to go run tests on her other C's to see how drastically stein cancelling fixes those ugly -20's.
Just confirmed, all above data is also true for 3c, which is the same level as 5c.
2c is a level above 3/5c, therefore is -7 on 5d7 cancel, making 2c her safest C on block. Weird.
6c is also level 5, therefore is also safer than 5c.
Data looks odd, but it's the results, I'll keep running them looking for errors.
Not sure how I feel about 6c being safer than 5c, but 8 frames off the recovery of 3c is pretty nice.