The only guaranteed way to build combos is to test them on Normal speed, then try to convert them to Turbo speed and see if they're possible with frameskip. On average, this method should save time over the long term. But it's too much of a hassle for me (plus i'm stubborn about these things).
Turbo frameskip does make some combos completely impossible, but it took me like three years before i ran into one. Or rather, i ran into a bunch of problems but always maneuvered around them. While it seems more likely for very long combos (aka Marvel) to run into impossible frameskip configurations, in my experience it's actually easier work around them with all the flexibility you get in the Marvel series.
Basically, if one frameskip pattern doesn't work for you, try another one. Anytime you have a 2-frame window in the middle of your combo, you have two frameskip configurations to choose from for the rest of the combo. (Obviously this won't apply if you're trying to make the world's first 99 hit 1-frame-only Marvel combo, but there aren't too many of those floating around.)
More commonly, Turbo frameskip really kills you in tiny, compact sequences that require two or three 1-frame timings in quick succession. Keiko and i were trying to figure out this zerokoubou SFZ2A Ken combo back in February, and it turned out to be completely impossible on default speed (Turbo 2).
It looks simple, but i spent hours trying to work around frameskip and got nowhere. Meaty c.HP to c.MP is a 1-frame link, and the j.HK to c.HP transition requires 1-frame timing for c.HP to connect on the last frame. Text from an old email:
So yeah, beware prime numbers!Alright so i tested it on default speed (Turbo 2) from W60 to W70, optimizing each one, and none of them worked. In fact, the pattern repeated every three frames, so i'm guessing that Turbo 2 skips every fourth frame. Anyway i'm convinced that this combo is simply impossible on Turbo 2 speed.
Here's what the Normal speed script looks like:
W10,-_D.W60,+U.W36,6.W5,-3.W19,+D3.W42,D2.W60!
Rolento's c.HP is flexible - he can do it a couple of frames early or later and it won't affect anything. So we can take that out:
W60,U.W36,6.W25,D3.W42,D2.W60!
That leaves three critical wait times: W36, W25, and W42 - with inputs falling on the 37th frame, 63rd frame, and 106th frame after the jump.
Now every fourth frame is disregarded on Turbo 2 speed. That means we have four possible frameskip configurations, starting from the U input:
U frame skips = automatic failure
U frame registers, 1st frame skips, frames 2-4 register, 5th frame skips, frames 6-8 register, 9th frame skips, frames 10-12 register, 13th frame skips, frames 14-16 register, 17th frame skips, frames 18-20 register, 21st frame skips, frames 22-24 register, 25th frame skips, frames 26-28 register, 29th frame skips, frames 30-32 register, 33rd frame skips, frames 34-36 register, 37th frame skips = j.HK fails
U frame registers, 1st frame registers, 2nd frame skips, frames 3-5 register, 6th frame skips, frames 7-9 register, 10th frame skips, frames 11-13 register, 14th frame skips, frames 15-17 register, 18th frame skips, frames 19-21 register, 22nd frame skips, frames 23-25 register, 26th frame skips, frames 27-29 register, 30th frame skips, frames 31-33 register, 34th frame skips, frames 35-37 register (j.HK registers), 38th frame skips, frames 39-41 register, 42nd frame skips, frames 43-45 register, 46th frame skips, frames 47-49 register, 50th frame skips, frames 51-53 register, 54th frame skips, frames 55-57 register, 58th frame skips, frames 59-61 register, 62nd frame skips, frames 63-65 register (c.HP registers), 66th frame skips, frames 67-69 register, 70th frame skips, frames 71-73 register, 74th frame skips, frames 75-77 register, 78th frame skips, frames 79-81 register, 82nd frame skips, frames 83-85 register, 86th frame skips, frames 87-89 register, 90th frame skips, frames 91-93 register, 94th frame skips, frames 95-97 register, 98th frame skips, frames 99-101 register, 102nd frame skips, frames 103-105 register, 106th frame skips = c.MP fails
U frame registers, frames 1-2 register, 3rd frame skips, frames 4-6 register, 7th frame skips, frames 8-10 register, 11th frame skips, frames 12-14 register, 15th frame skips, frames 16-18 register, 19th frame skips, frames 20-22 register, 23rd frame skips, frames 24-26 register, 27th frame skips, frames 28-30 register, 31st frame skips, frames 32-34 register, 35th frame skips, frames 36-38 register (j.HK registers), 39th frame skips, frames 40-42 register, 43rd frame skips, frames 44-46 register, 47th frame skips, frames 48-50 register, 51st frame skips, frames 52-54 register, 55th frame skips, frames 56-58 register, 59th frame skips, frames 60-62 register, 63rd frame skips = c.HP fails
So we have four different cases and in each of the cases something fails, meaning that the sequence is completely impossible in Turbo 2.
What i don't understand is why c.LP doesn't work either, since c.HP to c.LP is a 2-frame link. It should be able to connect on both the 106th and 107th frames. That doesn't make sense at all. Maybe there's another restriction that i'm overlooking? Or maybe there is a way to make the c.LP version connect?
Anyway, it's very rare to have so many 1-frame inputs so close together, with such strange wait times inbetween them. None of them share any factors - 37 and 43 are both prime numbers and 26 is only divisible by 2 and 13, another prime number.