Consonant and Vowel masks

[mahoganyduck]

So I've been playing with some masks based on consonant/vowel patterns extracted from the top 1M words (according to google) and was seeing some surprising hashrate drops vs just using the built in mask place holders.

(hashrates are with -O & ntlm hashes)

EG:
-a 3  -1 aeiouy -2 bcdfghjklmnpqrstvwxyz -3 AEIOUY -4 BCDFGHJKLMNPQRSTVWXYZ ?3?2?2?1?2?2?1?2?2?d?d
-a 3  -1 aeiouy -2 bcdfghjklmnpqrstvwxyz -3 AEIOUY -4 BCDFGHJKLMNPQRSTVWXYZ ?3?2?2?1?2?2?1?2?2?d
-a 3  -1 aeiouy -2 bcdfghjklmnpqrstvwxyz -3 AEIOUY -4 BCDFGHJKLMNPQRSTVWXYZ ?4?1?2?2?1?2?1?2?2?1?d?d?s
...etc

They crack at 42563.9 MH/s.
Comparably:
?l?l?l?l?l?l?l?l?l?l?l?l?l    63469.2 MH/s
?d?d?d?d?d?d?d?d?d?d?d?d  36738.2 MH/s

At the longer password lengths, the decrease in hashrate vs target space is still a lot faster in completing than using ?l and ?u, so it's still very much worth it.

I've also noticed something similar with masks containing static pieces in certain spots:

Ilove?a?a?a?a?a 572.7 MH/s   <-wut

?sIlove?a?a?a?a?a 16587.4 MH/s

I imagine it might have something to do with specific places having fewer options to iterate through, but whats going on that would cause that?