this is an idea that has most likely been played around with in the past, but what if lutris or by extension some other project to do with playing Windows games on Linux, used wine and a mixture of other projects to translate Windows system calls to any number of available analogs or combination of system calls available on linux, I can’t imagine this would be too difficult considering all the work that has been put towards the runtime, but at the same time I have not worked with any of the windows apis or any language that is incredibly low level, especially considering that most of the byte code for these games is most likely going to be operated on at the assembly level, but hey this is a suggestion and I encourage discussion
I just realized I probably didn’t properly explain this, basically you provide a directory structure that contains a game in some form and this new tool will recursively translate all of those binaries into another directory, and all of the binaries in this new directory will be capable of running natively on your system in specific, using any number of libraries on the system during translation, you could even opt in or opt out of certain features similarly to how Portage works on Gentoo
Even if you could automatically generate native wrappers that will work out of the box, so to speak, it doesn’t really help you that much because you’re still doing that API translation. There are crappy native ports of games that have done exactly that. There are frameworks that can be hooked up to do that, like eON for example (and one example off the top of my head using eON is The Witcher 2 port.)
If you’re proposing to automatically generate a linux native program from a binary of a complex Windows program, without having source code to port and compile, you’re barking up the wrong tree. You might be able to START trying to figure something out by disassembling a program but it’s going to take a lot more than just reassembling that into another program. (It’s not going to be practical)