For dad 1 I did 5 runs with Smath strat and 5 runs with GMP's strat. Likewise I did the same for dad 2 from an NMG save (I assume from an any% save the results would be similar to dad 1 so I didn't bother to do that).
Although the sample size is small, if there is a significant difference we should still be able to see it. I was also quite careful to make sure I don't cherry-pick data, so I tested exactly 5 attempts from the moment I decided I'll start testing. The only exception was when one time I fucked up an attempt for dad 1 Smath strat, so I scrapped all data I had collected up to that point and started again - this was to prevent more difficult scenarios from being underrepresented.
Here is the raw data (measured in seconds and frames at 60fps):
19:43
13:20
15:18
15:49
13:18
mean = 15:30
17:17
19:43
17:17
22:23
19:58
mean = 19:20
16:57
16:35
14:29
16:59
14:28
mean = 15:54
16:43
22:00
16:39
24:21
19:17
mean = 19:24
To my surprise, Smath strat appears to be better for both dad fights 1 and 2 - even with the worse RNG of NMG for dad 2. The sample size is small, but given the effort I put in to not cherry-pick the data and how significant the results are, I think it's quite clear. Even if we ignore the mean values and look at individual data points, it's clear that with Smath-strat the data tend to be closer together (bar a singular anomaly for dad 1), whereas for GMP-strat the data tend to be further apart, which suggests that Smath strat is better on average but is also more consistent.