CoolHandMike
He/Him
Editor, Skilled player (1863)
Joined: 3/9/2019
Posts: 1022
Lord_Bahamut wrote:
Gecko closing out 2025 with another banger. Link to video
Wow a 100% ff7 tas!
discord: CoolHandMike#0352
Post subject: Random Encounter Pattern Optimisation
Joined: 3/18/2006
Posts: 972
I've been really enjoying Lil_Gecko's any% TAS - great work. A couple of questions on the routing:
  1. Has the random encounter pattern been optimised? i.e. the decision tree of whether to stutter-walk, menu-glitch, or just take the battle. IIRC, this was brutal to optimise because each encounter decision changed the future RNG and danger state, making the search space explode.
  2. If so, what approach did you use? A full optimisation pass, or a heuristic? Did BrutalAl's REPCalc come into play, or did you roll something custom?
If nothing's been done on this front, I'd be up for attempting to build an optimiser app for it.
Lil_Gecko
He/Him
Player (196)
Joined: 4/7/2011
Posts: 529
antd wrote:
I've been really enjoying Lil_Gecko's any% TAS - great work. A couple of questions on the routing:
  1. Has the random encounter pattern been optimised? i.e. the decision tree of whether to stutter-walk, menu-glitch, or just take the battle. IIRC, this was brutal to optimise because each encounter decision changed the future RNG and danger state, making the search space explode.
  2. If so, what approach did you use? A full optimisation pass, or a heuristic? Did BrutalAl's REPCalc come into play, or did you roll something custom?
If nothing's been done on this front, I'd be up for attempting to build an optimiser app for it.
I have been using AceZephyr's tool called Big_Shoes that produces a stepgraph in real time while TASing and lets you know about enemy formations you can encounter on each field. On top of trying to go fast, there are also a lot of specific formations you need to encounter (Stealing the Hardedge, the Circlet,...), so sometimes you need to menu-glitch to increase the formation value even if not optimal for speed. (Cannot remember if I had to manipulate them but that's something to consider) However every decision I took, I took by watching at the graph only and testing, there hasn't been any program calculating the best path. There are already a few improvements I know about and there's very little doubt in my mind that it can be optimised further though. I think an optimiser app would need to include specific random encounters that ultimately speed things up in other fights but if you're willing to work on it, by all means do, it can only be beneficial for anybody who'd like to take a shot at TASing.