First off, this should go into the Method Development thread, as it has not been discussed enough to be worthy of its own thread, and it isn't even a good method/idea.
And as for the method itself, trying to combine Roux and CFOP has been done again and again, and it's never better. OLLCP is a very underdeveloped algset, having nearly as many algs as ZBLL while being worse overall. It doesn't even solve that much, which just means another step is needed after the fact. OLLCP algs are by far worse than doing standard OLL PLL, while still having the same amount of looks, required pauses between steps. Nothing is realistically gained from not solving 2 cross pieces that roux isn't better at. In your example, you didn't even do anything special that makes use of the unsolved edges, so it didn't even save anything from a standard CFOP solution. OLLCP algs don't even save more than 1 move against ZBLL, being around 11 moves with ZBLL at 12, while ZBLL solves all of the last layer, and OLLCP requires you do to an EPLL, or in your method, LSE. Which doesn't even save anything compared to doing full ZB or regular OLL PLL.
Trying to combine Roux and CFOP just isn't viable for a speedsolving method, and claiming it can compete with ZB, which has been demonstrably 2-looked. While with this, it can at best be 3-looked, with full pseudo-F2L OLLCP LSE, which just can't compete with ZB. It might save more moves than CFOP, but the low quality of OLLCP algs, combined with the intuitive MU moves of LSE make it much less reliable, or even better than OLL PLL already is. Pinkie Pie as its own method is likely better than this, since it does the same thing, without rotating. This hardly even classifies as a different method, since it finishes the exact same as Pinkie Pie, except you're rotating for FB and SB and forcing the D center to be solved, which reduces efficiency.
You basically removed the best parts of each method, and replaced it with a worse version. FB and SB of Roux is insanely efficient, and allows for a lot of creative solutions to solve them in an ergonomic matter, so replacing it with CFOP F2L, which has rotations and forces you to break and repair, which is an artifact of cross. So trying to apply that to building FB and SB is just how a CFOP solver would approach roux if they had no experience. It's just bad, and defeats the entire point of building the first 2 blocks. Then OLL PLL is one of the best steps of CFOP, being very easy to recognise and execute quickly without being overloaded with algs, which you replaced with OLLCP, which has bad algs and recognition, combined with LSE, which is good, but gets slowed because of OLLCP.
It's just not a good way to "create" methods by smashing together existing steps of different methods, and calling it better than ZB. It's more efficient to solve 2 blocks and finish like roux, or to finish cross and finish like CFOP/ZB, you've found the worst of both methods and stuck them together thinking it was actually going to be good. And trying to pass off OLLCP as being just as good as the base OLL alg is just completely wrong, and comparing it to ZB, which doesn't even do the same thing, is just pointless. If you want actual feedback on this "method" then post it in the method development thread, because that's where undeveloped methods like this are supposed to go. Then get feedback from people with more experience, who can help make it better. And if it turns out to be something worthwhile, after a lot of effort working on it, not just personal bias, then it might be worthy of its own thread. But as it stands, this is way too underdeveloped to have its own thread.
