Quote from ExplodingCabbage:
Yeah, as far as I can see route planning software only starts to even potentially become worthwhile when you're dealing with extremely complex RPGs, and then it becomes impossible. There may be a tiny minority of games for which the computer planning approach works but I cannot think of any.
For one thing the order in which you do tasks will almost always affect the time they take (due to having different equipment/abilities), so for the software to just brute force all possible routes you'd need to have tested every fight in the game at every possible level of experience and combination of equipment and tell the computer how fast it can be done, you'd need to tell the computer the experience and item reward for every possible action in the route, plus have it know all the different consumable items you have access to (including all the different possibilities for randomised loot) and the time they can save in every fight. And you'd need to have an algorithm capable of processing all this information.
Clearly this is unworkable, since just to have the computer brute force the problem you're effectively testing every possible route in the game yourself.
Of course, you could skip some of this work and put in guesses/approximations, but then you're still doing a whole lot of work to acheive something that could've been done better with some thought, testing and intuition.
For one thing the order in which you do tasks will almost always affect the time they take (due to having different equipment/abilities), so for the software to just brute force all possible routes you'd need to have tested every fight in the game at every possible level of experience and combination of equipment and tell the computer how fast it can be done, you'd need to tell the computer the experience and item reward for every possible action in the route, plus have it know all the different consumable items you have access to (including all the different possibilities for randomised loot) and the time they can save in every fight. And you'd need to have an algorithm capable of processing all this information.
Clearly this is unworkable, since just to have the computer brute force the problem you're effectively testing every possible route in the game yourself.
Of course, you could skip some of this work and put in guesses/approximations, but then you're still doing a whole lot of work to acheive something that could've been done better with some thought, testing and intuition.
The thing you'd need to do is reverse-engineer the game in order to find out all map lay-outs, experience algorithms, drop chances and go on. I think it's best for optimizing movement in smaller areas, though it'll probably be impossible to follow the given route close enough for the difference to be noticeable. Unless you're making a TAS obviously.
It's useful for games of which the algorithms for the physics are known. Because then you can calculate whether that jump that SEEMS like a good shortcut is actually at all possible (for NES/SNES/GameBoy/Stuff: use an emulator and frame advance).
I feel that the most useful tools for speedrunning are not so much tools that can feed you a route to follow, but stuff like emulators and memory readers. Because, whilst strafing might not seem faster, it may just get you .1 pixel per second further. In a game where you have to walk a few megazillion pixels.