Username:
B
I
U
S
"
url
img
#
code
sup
sub
font
size
color
smiley
embarassed
thumbsup
happy
Huh?
Angry
Roll Eyes
Undecided
Lips Sealed
Kiss
Cry
Grin
Wink
Tongue
Shocked
Cheesy
Smiley
Sad
<- 1  -   of 34 ->
--
--
List results:
Search options:
Use \ before commas in usernames
Well, that doesn't sound TOO bad (though we'll see when I actually try it). Thanks! Smiley

Another silly question, what do the people in charge of choosing game lists for the marathons think about Japanese versions of RPGs?  Do they feel that it wouldn't be as fun to watch for most viewers as the English version (since the language would be incomprehensible for most, and the glitches are essentially the same)?  If this isn't the right place to ask, I'll take the question to the Marathon board.
D:
You'd be best off asking the marathon organizers for an official answer.  They might be pretty tired for the next few days, though. Tongue

Speaking unofficially, there have been games in languages other than English at past GDQ marathons, but I think they've been games where you don't need to read the text to be entertained by what's going on.  Based on that, I don't think a glitchless run of Mother 2 would work, but a glitched run might be possible, especially if it showed off new strategies vs. the run from AGDQ 2012.  The English version would probably be preferred, though.

Again, just my opinion.
If I can get the RNG down, a stairs glitch run of Mother 2 would almost certainly be faster than the same run in Earthbound, so I'm hoping that might be a compelling argument for it.  I could always learn the RNG tricks for both versions, just in case.  And yeah, I'm sure the poor guys don't even want to hear the word "marathon" in the near future. Wink
D:
I made a Lua script for detecting enemy spawns.  It is attached to this post.  The way the script works is pretty simple:
  • Places where the game can randomly spawn enemies are highlighted by blue squares.
  • If you trigger a potential spawn, but the game chooses not to spawn an enemy there, the script draws a white label indicating what type of enemy group was triggered.
  • If you actually do spawn an enemy, the script draws a red label instead.

My intent in writing the script was to help find paths that prevent the game from even trying to spawn enemies, thus removing some randomness from the game.  Unfortunately, I spent about three hours playing around with the script and didn't find much.  There's probably stuff I didn't find, though.  I plan to go through that highlight tomorrow and post timestamps for anything that seems interesting.

You can also use the script to learn where enemy spawns are likely to be triggered and where the enemies usually appear.  I'm probably never going to put that much effort into running the game, but it's an option. Tongue

The script requires BizHawk 1.6.0a.  Due to a bug in that release, though, you need to do some configuration within the script to make the blue squares show up.  Please see the comments at the top of the script for information.  The bug should be fixed in the next BizHawk release; assuming it is, I'll post another version of the script that doesn't require the extra config when that release is available.

I also recommend using at least 2x zoom so the screen isn't completely covered in text.

Please report any issues with the script here (or PM me).
D:
I went back through the video and located a few parts where things actually happen.  All timestamps are for the llanfair timer rather than the Twitch seekbar because I was too dumb to sync the two when I made the highlight.  I think I started the timer at 2:16 in the video, if that helps.

4:00 - 4:10 = Skip one early spawn in PRV.  Not super helpful.

5:05 - 5:50 = When approaching the pencil statue the first time, you can spawn enemies behind you if you get too close to the statue.  There's a pretty obvious visual cue you can use to keep this from happening.  I think you're unlikely to get spawns anyway if you're careful about watching for the statue, though.

5:50 - 6:15 = You can accidentally spawn an oak behind the pencil statue if you take a really weird path getting there.

11:05 - 12:35 = It might be possible to prevent a few spawns just before the second butterfly tile in Peaceful Rest Valley.

14:40 - 15:50 = It also seems like you can skip a few potential spawns just northwest of the PRV cabin.

18:00 - 18:30 = Recap of butterfly and cottage skips.

22:20 - 24:05 = How to properly despawn the troll flies north of the Threed graveyard.

32:20 - 33:20 = There seems to be a safe path around the Cave Boys if you walk a little out of your way.  It's probably unnecessary given that the Cave Boys are easy to dodge.

33:55 - 35:25 = There's a path that might reduce the risk of spawns in one part of Threed on the way from the prison to Boogey Tent.  It's not a high-risk area, but I thought it was interesting.

46:03 - 46:17 = If you can see the spawns soon enough, you can sometimes avoid getting trapped on the way into Belch's base.  Unfortunately, you usually can't see the spawns soon enough.

1:12:25 - 1:13:30 = There's a fairly clear path from the sailor guy to the next warp guy in Moonside.

1:14:35 - 1:17:05 = The invisible man text skip apparently disables spawns in Moonside.  Someone needs to test this on console...

1:26:40 - 1:27:15 = You can skip one tile in Dungeon Man by walking funny.  Not that helpful.

1:38:45 - 1:41:00 = You can avoid both enemy spawns after Tessie when you return to Winters.  This saves you from having to teleport around the enemies behind the pencil statue.  This might actually be pretty cool because I think it's faster than teleporting and it should hopefully be foolproof (i.e., me-proof).

1:50:15 - 1:51:15 = You can skip one tile in Stonehenge by walking funny.  Still not that helpful.

2:15:00 - 2:17:45 = It might be possible to have the first enemy in Fire Spring despawn itself if you wait in the right place for a couple seconds.

2:24:30 - 2:25:45 = You can skip one tile in Milky Well, but you have to walk pretty far out of your way to do it.

I was definitely hoping to find more than this, but maybe some of it will be helpful.
Edit history:
nemi: 2014-07-03 01:23:53 am
nemi: 2014-05-04 02:01:05 am
Debug menu in Onett:



There are basically 2 steps to this glitch:

1. Get an empty dialog box: this empty box should be preceded by a relatively quick sequence of garbage text that contains a pattern similar to "HOHO" somewhere in it.

2. On the empty dialog box, press or release any button on the second player controller. Exactly 11 or 13 frames later (either timing works), advance the dialog box by pressing Down-Right-Y-B-A on the first player controller. Since Down, Right, and Y don't do anything in a dialog box, you can be holding those buttons down before you press A & B on either of the correct frames.

While the video was recorded off of a saved file, this glitch should work without a safety save.

As far as I can tell, step 1 is completely random, and I'm not sure if anything affects the rate at which the empty dialog box appears. For what it's worth, each time I got the empty dialog box, I was holding the R button on the 2P controller while pressing and holding the L button (also 2P) to check the spot. I've also had success while holding the Y button instead of the R button. I'd estimate I got the empty box once in every 10 or 15 resets. (This is a very rough estimate, since I didn't actually count my attempts.)

Whenever I saw the empty box, it was the first glitched message I got. The most common glitched message I saw was a 10 or 15-second pause followed by the text "cOa QA." Another common result was an empty menu rather than an empty box, the difference being the empty menu contains a cursor pointing at nothing. I usually reset as soon as I saw a wrong glitched message.

I've also seen an empty box that didn't lead to the debug menu. IIRC, the garbage text preceding the wrong empty box didn't contain the "HOHO" pattern, so that pattern seems to be key.

I've tested this glitch on console and on snes9x 1.51, and as far as I can tell, the glitch behaves the same on both platforms. I've done a little testing on VC as well, but the most common glitched message I saw on VC was a long string of q's, which I never saw on console or snes9x. I suspect the string of q's corresponds to the "cOa QA" message on the SNES because the letters "Oa QA" appear once the q's stop.

EDIT: After several full-run attempts, I saw a few cases where the empty dialog box appeared after the 1st glitched message--in the first case, the first couple of glitched messages were extremely short, lasting less than a second each, and the empty box was the 3rd glitched message.

In the second case, the first few messages were also extremely short. After a while, I got an empty menu, and then pressing B on this menu led to the empty box. I failed to open the debug menu on the first empty box I got, but after checking the corner for a little while, I got the menu that led to the empty box again. I was able to repeat this several times before I finally succeeded in accessing the debug menu.

At this point, it looks like sub-12:00 (Nico timing) is possible, but only if the glitch produces the empty box relatively quickly. Otherwise, a low 12:xx is probably more likely.

EDIT 2: I experimented a bit more on getting the empty box, and I got it to appear by pressing L on the 1P controller while holding Y on the 2P controller. As far as I can tell, the rate of success is similar to the old method I was using (L and R both on 2P).
D:
Oh, man.  Good stuff, nemi!

I think everyone suspected the trick would work in the NA version of the game, but seeing that you've actually gotten it to happen is reallllly cool.
This is the closest I've gotten to a complete check area glitch run. I was lucky enough to get the empty box, but I failed to bring up the debug menu.



By Nico timing, which starts on file select, the empty box appeared at about 10:46.
D:
So very close. >_<  Great work as always.  You even got the wall clip so fast...

I guess sub-12 is pretty much guaranteed if you can get the glitch to happen?  I think the final bit with Picky takes less than a minute, if I remember right...
Yeah, the time between the empty box and "The End...?" is about 54 seconds. The glitch can sometimes take a while to bring up the debug menu, and the wall clip can sometimes be slow as well, so 12:xx is also a pretty likely time.
Finally got it...11:57.



The glitch took a while to bring up the debug menu (20 checks vs. 6 in the WIP video), and there are a few movement and menu flubs, but I think I'm OK with this.
I wasn't able to find anything on it with the search function, so I'm not sure if this is new info or not. It seems like it could be useful for the check area and  (maybe) any% stairs glitch categories. It turns out that event trigger skips are not so TAS-only, and you can menu buffer to get past them:



Entering the menu a frame after pushing a direction will open the menu will interrupt Ness's movement, which seems to interrupt the trigger like stutter stepping does. It's still frame perfect, but it's a lot easier to pull off in real-time. Buffering works with both the HP display and the normal menu. The HP display is faster to close, but it completely wrecks any RNG manipulation that was going on.

For Starman Jr., I use the cliff/grass tuff as a visual cue to get myself lined up right below the trigger. It takes 2-4 buffers for me to line up the trick, and 3 moves to cross the trigger. The  fight takes about 50 seconds if you do it perfectly, so obviously this way saves time if you can deal with not getting the experience/money from that fight. Skipping this leaves you $8 short of buying the standard Onett equipment, and it makes the initial shark fight very difficult. and At the very least it skips the dialogue around the fight, so it may still save time even if you need to route in a random encounter on the way to Onett.

It isn't possible to skip Buzz-Buzz joining, as you'll softlock during the Minch cutscene, but you can skip Buzz-Buzz's death (and getting the sound stone) using this method. That one is a bit harder to line up for. I try to use what pixels of the light blue tiles are showing. The amount of steps you need to take to skip this trigger is variable for some reason. It's possible to do in 3 steps, but sometimes it needs 4. I think it has something to do with whether the NPCs are moving or not, but I haven't spent too much time investigating so far.
For me, this makes the skip doable but still very difficult (vs. impossible with just button mashing). I've been simply trying to press Up and A at the same time and hoping human error will cause my presses to have slightly different timing, but I've been successful only once so far. Maybe there is a better button pressing method I'm missing.
Yeah, it's still difficult. It's going to take me a lot of practice to be comfortable trying it in a run. I don't use any special method, unfortunately, other than "push A just slightly after pushing up"

For any% stairs glitch, skipping Starman while adding a dog fight (hah) to the route gives you just enough XP to get level 2. You can then sell the bread roll to make enough cash for the standard starting gear. Much more manageable than selling the hamburger. I made it through the sharks without too much trouble today, but I got lucky with smashes. I'm still not sure if it's faster overall, so I'll try to find a good way to time this tomorrow.

The dialogue that Pokey gives you if you try to go south to Onett before bringing Picky back to the house seems unskippable. That trigger is different than the Starman one in that you don't really move on to it. It's more like an invisible wall that you push up against. King leaving when you go up to the meteor seems to be the same way.
Interesting findings CacheLine. How tall is the trigger area for the starman? From your video it looks like it's only a couple pixels tall, but that may be because you started recording in the middle of the skip. If it's just a couple pixels tall, this could save a bunch of time.

On a side note for anyone interested, I've been spending a ton of time recently with the any% stairs glitch route. I've got my time down to around 1:32:xx and I've been playing with some new techniques. The one I'm focusing on is skipping the Kraken in the Sea of Eden and going straight to Ness's nightmare to use RNG manipulation for that fight instead of hoping for something good. I'm in the process of writing a Lua script to interface with BizHawk that will brute force the fight and find the fastest way for me to kill Ness's nightmare. Ness is level 10 and Paula is level 1 so I have to get rounds where I'm getting smashes from the flying men and Nightmare uses his flash move but doesn't kill anyone.

There are Pros and Cons to this strategy: It should make sea of eden faster by a few minutes since you're gaining fewer levels and watching less text scroll by (and you have one less fight of course). BUT, You no longer have shield Sigma for the Giygas fight. I'm going to implement heavy RNG manipulation for the giygas fight as well going for smash hits and First round paralysis + Rockin Alpha.

There are a few more places I want to try using RNG manipulation in the run (Frank, Giant's Step, Winters, etc.) but I'll let ya'll know how things go. I'll also post the Lua script here when I'm done if any of you guys are interested.
Quote from andyperfect:
How tall is the trigger area for the starman? From your video it looks like it's only a couple pixels tall, but that may be because you started recording in the middle of the skip. If it's just a couple pixels tall, this could save a bunch of time.


I started right on the edge in the video - from there I could move left or right and not hit it. The trigger is only 3 pixels tall, but lining up on the exact edge of it is tricky.
Some thoughts regarding manipulation of Ness's Nightmare:

Manipulating Ness's nightmare would probably make the manipulation of Giygas necessary because the loss of PSI Shield Σ will likely make the Giygas fight longer, negating some of the time gained off of Ness's Nightmare. The good thing is that manipulating Ness's Nightmare makes manipulating Giygas more doable because the Flying Men will always be in the same state.

The remaining obstacle to manipulating Giygas is that the RNG is extremely active in front of the entrance to Giygas's lair, which makes the RNG inconsistent when taking a direct path through the Cave of the Past. However, I noticed a while ago that walking up to the left edge of the Cave of the Past settles the RNG down. Here is a possible route that makes the RNG more consistent:



The only precise part of this route is at the very end before entering Giygas's lair. Lining up Ness with a texture on the wall and then walking straight right seems to work well.

If manipulation of Giygas is possible, then shopping in Saturn Valley will be unnecessary because everything bought in Saturn Valley (Night Pendant, Horn of Life, Secret Herb) are contingencies for the Giygas fight. Skipping this shopping should save about a minute.
So I had what I thought was a perfect RNG route all the way through Sea of Eden and until the end of Ness's nightmare, but after spending several hours with it last night, Sea of Eden is proving to be crazy difficult to map out consistent RNG. I am able to get the first kraken trapped against the pillar at the exact same position down to the pixel 100% of the time, but RNG does funky things after that, even with identical actions.

For example, if I make a save state with the kraken caught against the pillar and swim up, the kraken will swim down and despawn himself, but sometimes RNG shifts forward one more unit than I'm expecting and sometimes it doesn't. Either result seems to happen equally as likely. Then sometimes the RNG will just continute to shift forward on its own. I'm going to be spending a lot more time with Sea of Eden to narrow this down. I was not expecting the Krakens to be so difficult to manipulate, but this is easily the most difficult area I've tried manipulating (and I even tried manipulating Winters with Bubble Monkey following behind you, whose movements act as an NPC, shifting RNG forward everytime he stops following you whenever he tries to catch back up to you).
Edit history:
MUGG: 2014-10-03 02:45:09 pm
MUGG: 2014-10-03 02:40:36 pm
MUGG: 2014-10-01 04:29:33 pm
MUGG: 2014-10-01 04:26:43 pm
MUGG: 2014-10-01 04:26:19 pm
TASer
Maybe there are people that remember when I said, some years ago, a Japanese runner named Ukyo got a 1:35:55 time (although with the use of autofire...). I've seen his video and shared it with very few people back then. Now, I looked at this page again and there are new interesting times... 1:16:03 and 3:56:50 by あくえり ...

Just bringing this up; maybe I will try asking a Japanese to help find video proof or any details.

EDIT:

According to pirohiko, "He seemed to patternize luck manipulation by established operation from some reset."
There exists a very small video showing 1 minute of him speedrunning and resetting. When I know more I will edit again.
D:
I've updated the glitchless route on the knowledgebase.  It should now be closer to the route most people are using anyway.  The main changes are using Super bombs on Clumsy Robot (faster even with inventory manipulation as long as you skip the Hyper beam and Mr. Baseball bat) and using three Super bombs on Kraken (much more likely to get you a one-round even if buying the Super bomb costs a few seconds).  I also added some other strategies I've seen people using and updated some of the inventory routing to make junk items less likely to end up on Ness if things don't go exactly according to plan.

Please let me know here or on the talk page or in my stream or wherever if you have comments or suggestions regarding the route.

Sorry for not updating it sooner. >__>
Edit history:
andyperfect: 2014-10-13 02:19:25 pm
andyperfect: 2014-10-13 02:18:56 pm
andyperfect: 2014-10-13 02:17:23 pm
andyperfect: 2014-10-13 02:16:58 pm
andyperfect: 2014-10-13 02:16:19 pm
Thanks for updating that sdfg:

To MUGG and anyone else interested:

Yeah, heavy RNG manipulation is pretty much required if you're aiming for any time under about 1:45. I'm working heavily on the route and my goal is a sub 1:20 English run.  I've started working on a new route that includes killing Ness's Nightmare in 2 rounds (similar to the TAS)...

Video of me doing RNG manipularion through  Sea of Eden and Ness's Nightmare fight: http://www.twitch.tv/andyperfect/c/5262629

...and killing Giygas using RNG manipulation as well. The thing that worries me about that, though, is that if I want to manipulate smash attacks, it relies on the guts stat. BUT, there are areas during the run that are not being RNG manipulated where Ness is gaining levels. This means that I'm not 100% sure Ness's guts stat will always be the same at the time of the Giygas fight.

I have some questions about the mechanics of how SMASH attacks are determined, if I can pick ya'lls brain.

Imagine I have two different Ness's starting the Giygas fight. All stats, inventories, etc. are the same between the two situations except for one thing: Guts stat. Ness A has 42 guts and Ness B has 43 guts. Now let's imagine that the RNG values are all the same, and Ness is about to a normal attack. How exactly is the Smash attack determined? I know what the odds are (From Starmen.net)

guts/500 or 1/20, whichever is greater

So because my guts is greater than 25, the former is used, guts/500, or 42/500 and 43/500. Now for my question. How exactly is that guts value used in the calculation? Does the game generate a random value and then compare against Ness's stat? Essentially what I'm asking is the following:

If Ness B (43 guts) gets a smash, did the game just generate a random number in the range of 1-500 and then it fell under 43, or did it use 43 as some kind of seed to determine the calculation? If it's the former, I can figure out what Ness's lowest possible guts is at the Giygas fight and then manipulate the fight so that no matter what his guts is during the fight, I'll have consistent smash hits.

Hopefully that question makes sense. If you have any ideas on a better way to tackle the problem of inconsistent guts values, I'd love to listen Smiley The other thing I've considered is making sure every area where I gain levels (Frank, Punks, Starman Jr., Giant's Step, the Cop fight, Ness's Nightmare) is RNG manipulated for consistent level-ups.) but that is a daunting task that I'd like to tackle over time.
D:
I grabbed an assembly trace during attacking and eventually found what looks like the relevant code.  It pulls a value from the normal RNG, multiplies it by 500, divides it by 256, and then compares it to your guts (or 25, if guts < 25).  You smash if the random value is less then or equal to your (effective) guts.  So, any value that translates to less than whatever your minimum guts is should do the job.

I'm not sure if inconsistent stats could affect other parts of the battle text, though (e.g., taking 2 digits of damage instead of 3 somewhere, or the turn order changing).
__sdfg holy cow, you managed to trace the assembly? My goodness I didn't even know that was possible. That is awesome -- thanks for doing that Smiley

Awesome information to know, by the way. I've managed to consistently manipulate from the beginning of Cave of the Past to the end of Phase 1 Giygas with Ness's guts at 42. I don't believe anything else will affect my end damage. At the end, I've done about 2050 damage after 7 rounds. That should be plenty and even if my offense is 3-4 less I'll still be fine. If I find out that I need to do more damage, it's trivial to change what I've done. I don't believe anything will change the amount of text being printed. Ness's smashes do 305 damage (after his offense gets downed on the first round by Pokey) and Paula's freeze betas do 130+ every time.

The only oddity I found weird was the following. Using a save state, I found that rounds do not always have consistent RNG results given identical inputs. On one of my rounds, I found that after I have given all of my inputs and the attacks start, the RNG is always the same, but by the time it gets around to Ness attacking, he didn't always have the same attack as RNG wasn't always the same. One result led to a smash, the other led to him doing a weak 26 damage attack. Note that this was with allowing the attacks to do their thing and not mashing anything during the animations/text so it was something completely out of my control. There must be some internal race condition in certain situations that can lead to different results. This is the only round of attacks in all my time manipulating RNG that I have ever seen this happen. I found though, that if I just mash A during the attacks/text, the RNG manages to be the same.

Thanks again for taking the time to figure that out for me again! I appreciate it.
D:
Quote from andyperfect:
__sdfg holy cow, you managed to trace the assembly? My goodness I didn't even know that was possible. That is awesome -- thanks for doing that Smiley

That's a standard feature in TASing emulators.  You can do it in BizHawk using the Trace Logger window, but you need a version where it isn't completely broken. >_>  I used the latest one (1.8.4) since it happened to work.

Interpreting the traces takes a bit longer since you don't really get a source file, you get a list of instructions as they are executed and the register values.  So, you have to identify all the loops, subroutines, etc., yourself...  I'm not very good at it. >__>

Quote:
Awesome information to know, by the way. I've managed to consistently manipulate from the beginning of Cave of the Past to the end of Phase 1 Giygas with Ness's guts at 42. I don't believe anything else will affect my end damage. At the end, I've done about 2050 damage after 7 rounds. That should be plenty and even if my offense is 3-4 less I'll still be fine. If I find out that I need to do more damage, it's trivial to change what I've done. I don't believe anything will change the amount of text being printed. Ness's smashes do 305 damage (after his offense gets downed on the first round by Pokey) and Paula's freeze betas do 130+ every time.

Nice!  Hopefully, there aren't any oddities to the battle system that will cause your stats to mess with stuff.

Quote:
The only oddity

Uh-oh.

Quote:
I found weird was the following. Using a save state, I found that rounds do not always have consistent RNG results given identical inputs. On one of my rounds, I found that after I have given all of my inputs and the attacks start, the RNG is always the same, but by the time it gets around to Ness attacking, he didn't always have the same attack as RNG wasn't always the same. One result led to a smash, the other led to him doing a weak 26 damage attack. Note that this was with allowing the attacks to do their thing and not mashing anything during the animations/text so it was something completely out of my control. There must be some internal race condition in certain situations that can lead to different results. This is the only round of attacks in all my time manipulating RNG that I have ever seen this happen. I found though, that if I just mash A during the attacks/text, the RNG manages to be the same.

What?  That's...  what?

Were other characters dying during this time, maybe?  That's the only thing I can think of that would do that. :/

Quote:
Thanks again for taking the time to figure that out for me again! I appreciate it.

Sure, no problem! ^_^
Edit history:
andyperfect: 2014-10-16 02:55:40 am
So Frank is stupid. And his prayer is stupid. All of the prayer sequences have scripted NPC movement EXCEPT the Frank prayer where there's two NPCs in the burger joint screwing things up, The thing is, they move so much that it's impossible to get consistent RNG. What I'm having to do is watch the damage that Giygas takes after the prayer. From there, I can advance the RNG to a common spot beyond any resulting RNG and then have consistent values. It took me a couple hours to narrow down and get right, but we should be good to go now for the entire Giygas fight from start to finish being manipulated. I did it all in a practice session this evening and from beginning in Cave of the Past to the final prayer on Giygas ends up being just shy of 13 minutes, about 5 minutes faster than my current PB. If you factor in the fact that we no longer have to shop to buy secret herbs, the night pendant, etc., it's a total time save of a little over 6 minutes. Together with the Ness's Nightmare manipulation I think we're looking at a possible sub 1:20:00 run.

I'll be grinding out attempts and improving my execution of the manipulation in the coming weeks. I'm excited to see this all coming together Smiley