Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
I wasn't sure how far back to go. Here are some earlier ones.
0B (get hit against either the second or third one)
16 (get hit against the second one)
19
1B (get hit against either the second or third one)
1E (get hit against the third one)
Note that getting hit against the first one has the same effect as letting the RNG advance once before the battle. (Whether the high cadet is real or not is determined when you hit him.)
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
I still have my spreadsheet from when I helped with the IoG TAS, so it didn't take too long to find other values in that area which give 3 real High Cadets , allowing for some hits to be taken during the battle.
24
2C (get hit against either the second or third one)
2F (get hit against both the second and third ones)
32 (get hit against the second one)
35 (the one you know about)
37
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
The RNG has no source of entropy, so it can only be manipulated by using it. The value at 7E0417 increments every time the RNG is used, so you can watch that to keep track of whether the RNG changed. (If you want technical details, the RNG is a 128-bit big endian LCG, where 7E0417 is the least significant byte.)
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
Tub wrote:
CyberShadow wrote:
One state takes 10 bytes: 2 bytes (bitfield) to store the action required to get to that state from its parent state
So you're storing a separate state for each step? That may be more than you need.
The puzzle isn't solved by walking, but by interacting with the objects (boxes, spinners, level exit, ..), so I propose to store a state only for every interaction done.
There are 5 possible steps (including "switch character"), so 2 bytes would be wasteful if he were storing a separate state for each step. Conversely, storing a location would take 9 or 10 bits, and storing an action and direction with it would still easily fit within 2 bytes. Hence, I assume he's already doing as you proposed.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
Good luck on this project. I've done computer search for a couple things myself (one simple room in Eggerland:MnF and one boss in Lolo 3), but I didn't think it would be tractable for anything as complicated as a typical Kwirk (or any puzzle game) level.
I figured my Kwirk TAS might be suboptimal, mainly because it's the first TAS of the game (and was done by a human). I also found an improvement to 3-10 (426 steps, 1:14 vs 434 steps, 1:15) the day the run was published, but I'm sure a computer would do even better if it could stomach such a long level.
This probably won't affect the optimal solution too much, but times are measured in frames here instead of steps. Here are the frame counts for various actions:
Move one step normally: 9 frames
Move one step while pushing block: 10 frames
Move one step while pushing spinner: 12 frames (whether moving one space or two)
Filling in a hole: 26 frames (not including the time taken to push the block)
Switching character: 30 frames (32 if you immediately switch again)
Switching character due to finishing with one character: irrelevant, so I didn't check.
I remember briefly looking at this, and it's not an overflow. The same address within a block stores experience gained by defeat in blocks corresponding to enemies, and something else (I forgot, and I can't access my notes) in blocks corresponding to party members. The game doesn't expect you to be able to kill your own party, so it assumes the number in that address is the experience you're supposed to gain, and adds that to the survivors' experience totals.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
By Pick's theorem, the area of any polygon whose vertices have integer coordinates must be half an integer. The area of an equilateral triangle is s^2 * sqrt(3)/4, which is irrational (since s^2 is an integer). Hence, it is impossible to have an equilateral triangle in R^2 with integer coordinates.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
Good luck with the run. I recommend uploading your WIPs to http://dehacked.2y.net/microstorage.php so the rest of us can watch and maybe provide some useful feedback.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
The RNG is manipulated by spending a certain amount of frames with a cursor visible. Moving the cursor around has exactly the same effect as just sitting on the intended menu option and doing nothing for the same amount of time. Hence, I could freely make the menu movements as entertaining as possible (well, at least better than nothing) without losing any time.
Mother 2 / Earthbound works exactly the opposite way, so that's probably what you were thinking of.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
I think I'm qualified to answer those questions.
1) It was probably faster than manipulating it away. Hamburgers are a 1/8 drop from cops, so it's not surprising that one of the cops dropped one.
2) Selecting the weapon from the menu adds more text (and thus uses the RNG more) between selecting the cop's attack+target and checking for the SMAAAASH!. This presumably made it possible to reach a SMAAAASH! in a reasonable amount of time.
3) When you are confused, your target is randomly selected from all characters (enemy, ally, self, and NPC) in battle.
4) After Jeff crashes the Sky Runner, Ness and Paula (re)join the party. The teddy bear is in Paula's starting inventory, so it appears when Paula joins.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
When the ball is released, its speed is sqrt(2*r*g*cos(pi/9)). That can be converted to horizontal and vertical velocity components easily enough.
In order to calculate the time taken to reach the release point, I'd need to integrate sqrt(sin(x)) or sqrt(cos(x)), and I'm not sure how to do this.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
Don't forget that the tune-up also gives +20 power, guard, magic, and speed.
The hit rate formula in Nati's FAQ is wrong. With 142-171 speed, we have a 50% hit rate against Gorsia. Thus, 142 speed is all we need, since there's no way we'll level up enough to get a higher hit rate.
Also, I should mention that we do have several Recvrys, so Lejes's MP will not be a problem.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
Maybe I'm using a different definition of sequence breaking. The game is very open-ended after Magicant, so nothing that I skipped is actually "required". However, I did skip several events that are very much expected in normal play, including recruiting Ana and Teddy.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
Tristal wrote:
I wish I had a bit more experience with the game to know all the sequence breaking being done, but I was entertained watching it. Yes vote.
There aren't really any sequence breaks in this run. I merely skip random encounters and defeat enemies at low levels. Going to Rosemary Manor before Duncan's factory could be considered a sequence break, and I skip several optional things, but that's about it.
pirohiko wrote:
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
The remaining LastStarman battles didn't go quite as smoothly. I hope to get the R7038XX fight done in 2 turns, but that's unlikely.
Insert witty comment here.
(Note: I only get QuickUp because the RNG was uncooperative.)
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
Since pirohiko just submitted a glitched run of the Japanese version, I decided to stop being lazy and make some progress on this run. I still have more random encounters to do, but I'll post what I have for now.
Level 3 to 16 in under 4 minutes.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
That is correct. It is normally impossible to target enemies with restorative items. Somehow, the GreenDream makes the game think everyone in your party is dead or something, so it automatically targets the enemies instead.
Experienced Forum User, Published Author, Active player
(283)
Joined: 3/4/2006
Posts: 341
eternaljwh wrote:
Use Revive on enemy (special circumstances required, using aforementioned re-raise I think) to, as it sets HP of target to something based on a stat = 0 on enemies, so zero their HP = instakill.
You're thinking of FF4 (and that only works on dead enemies there anyway). The CT glitch uses an elixir to overflow the enemy's HP to a negative number, thus killing it.