First part of the run up to first encounter with Tritorch :
http://www.step.polymtl.ca/~jyzero/ff3smv1.zip
Rom : Final Fantasy III (U) (V1.0) [!].smc
Comments and advices are welcome.
Notes :
-Delays are here for randomness manipulation. It's why I got a lot of preemptive battle
-Lobo and Vomammoth prevents party from escaping battle. It's why I didn't escape in 3rd battle and hadto battle a bit in 4th and 5th battle
-I tried very hard to beat Whelk before it hides. With current levels, I needed 2 Tekmissiles + 3 Bolt beam to kill it (Lv 6 tekmissible does 620 damage and Lv1 Bolt Beam does 120 for a total of 1600 damage). Under the best case, I could only send 2 Techmissiles and 2 bolt beams before it hides. So I had to let it hides and killed it with 3 Techmissiles.
but then you take my 75 perchance chance of winning, if we was to go one-on-one, and then add 66 and two-thirds ch...percents...i got a 141 and two-thirds chance of winning at sacrifice
Joined: 5/20/2004
Posts: 118
Location: Over there!
Look at the graph again, this time with your eyes open. The periodicity is 15, not 16.
There were no Side / Back attacks is because the group of enemies used for testing disables them. The first groups that allow all types of attack are inside the mine.
Now I've got that out of the way, onwards to the technical crap!
There is a frame counter at 0x7E021E that cycles from 1 to 60 (60 / 4 = 15, hurrah!). The battle timing is generated from the value of this counter when the screen has completely faded to black.
The blurring effect that precedes every battle can only occur every 4 frames. This causes the 4 frame delay between changes in the battle timing.
(Big Edit: Removed value/attack list and an entire paragraph because there are unknown factors that affect it)
Thank you for the Info SprintGod. If odds of getting side or preemptive battles are indeed rigged, a back guard might be considerable (can get one in banquet). Where did you get this information ?
ventuz wrote:
anyone find that it desync at start of Narshe entrance? im using FF3 1.0 (U) (!)
MD5SUM is 4cdfc79b44cd6d532dc6631a5c762094
Make sure you don't have any cheat or ips file applied.
I'm not sure if having a previous saved game could affect too.
Joined: 5/20/2004
Posts: 118
Location: Over there!
The source is me. I'm good at digging around in games' memory :)
Preemptive attacks shouldn't be blocked by the formation you tested on, but since there were only 15 ways for the battle to start out, it's possible that none of them would give pre-emptive attacks.
The timer pauses for a moment whenever the screen fades out (entering a battle, new area, menu, etc.). This is the only way I currently know of that allows you to manipulate the possibilities of the next battle.
The 'unknown factors' mentioned in my last post seem to have something to do with the number of characters/enemies in the battle.
For example, when the timer is 16 (PAR code: 7E021E10), you will get Preemptive attacks against Guard/Guard, Lobo/Lobo and Were-Rat/Repo Man, but not against Guard/Guard/Vomammoth/Vomammoth or Were-Rat/Were-Rat/Were-Rat.
Progress report : none
Actually I've got something new to play with : 7E00BE
This is the random number generator index.
Understanding 00BE is the key to master FF3 ramdomness.
For the marshall battle to be optimal, I need two consecutives critical hits. 00BE is what gives critical hits.
SprintGod : Your discoveries about battles formations are described here :
http://www14.brinkster.com/assassin17/http://assassin17.home.comcast.net/
edit:link updated
I love to see when a game get breaked down like this to be abused at most. That shows some skills and planning for sure. :)
As much as I want you to rush through the game so it will be completed I still want you to take your time to plan as much as possible and discover more useful stuff etc. :)
Also, I just remembered that the Algorithms FAQ used to give different increment values for each terrain, but these have inexplicably disappeared (despite being apparently correct...). Anyways... here they are:
Grass 96
Desert/Forest 192
Elsewhere 112
The first screen of the first cave seems to be an exception. It adds 64 instead of 112.
i had a role in their disappearance, and i'll gladly defend the new ones. :P
http://assassin17.home.comcast.net/c0bank.txt
see C0/C29F through C0/C2DE. yes, the data layout means you can punish somebody who's stupid enough to combine Moogle Charm with Charm Bangle. >:)
note all my talk of "FF3usME values" doesn't mean much, because:
- FF3usME doesn't retain changes to this field anyway.
- Lord J is bound to fix the names quite soon.
if somehow the code i've studied is highly deceptive and you're sure the old values are right, do speak up. free free to ask any followup questions regardless.
if you said something different since Page 3 that i missed while skimming this thread, apologies.
Joined: 5/20/2004
Posts: 118
Location: Over there!
Ah, I see how it works. Now if only that section of the Algorithms FAQ wasn't even more screwed up than it used to be:
Assassin wrote:
i had a role in their disappearance, and i'll gladly defend the new ones. :P
Algorithms FAQ wrote:
The game keeps a counter that increases each time you take a step. Walking on the overworld map adds 192 to the counter for each step taken; walking in caves, towns, etc. where monsters can be encountered adds 112 to the counter for each step.
At IRC, TNSe suggested (or theorized of) making a luck predictor version of snes9x.
That is, a patched snes9x for this specific purpose, which displays additional information of the game's state while making the movie, aimed for making it as easy as possible to manipulate the luck.
In my opinion this is a very useful-sounding idea.
Ah, I see how it works. Now if only that section of the Algorithms FAQ wasn't even more screwed up than it used to be:
Assassin wrote:
i had a role in their disappearance, and i'll gladly defend the new ones. :P
Algorithms FAQ wrote:
The game keeps a counter that increases each time you take a step. Walking on the overworld map adds 192 to the counter for each step taken; walking in caves, towns, etc. where monsters can be encountered adds 112 to the counter for each step.
What new ones? :)
How is that more screwed up? The older version wrongly suggested that Grass had a different encounter rate than Desert/Forest, undoubtedly based on tests in a "low encounter" grass area.
As explained in that link I gave, there are just TWO categories:
- overworld (the terrain type -- grass, dry land, desert, forest -- does not matter)
- dungeons (that includes caves, towns, etc)
Why Terii didn't give the four different encounter rates within each of those categories, I'm not sure. (Instead, he just gives the "normal encounter" rate, which is by far the most common, as you can see with FF3usME).
The bottom line is that while both versions are incomplete, the new one is more accurate and far less misleading.
Joined: 5/20/2004
Posts: 118
Location: Over there!
Assassin wrote:
As explained in that link I gave, there are just TWO categories:
- overworld (the terrain type -- grass, dry land, desert, forest -- does not matter)
- dungeons (that includes caves, towns, etc)
There is one place where terrain type does matter, and that's around Narshe in the WoB. The forest uses the first overworld value (192), everything else uses the second one (96).
Technically, terrain type matters everywhere else as well. But this is made irrelevant because the terrain types within each section of the overworld have been given the same encounter rates.
note that those links are now dead. this documentation is now with the rest of the guides on my sites:
http://www14.brinkster.com/assassin17/http://assassin17.home.comcast.net/
also, it's been updated to account for the 1-60 frame counter, which means even more thwarted attack formations are listed.
New section up :
http://www.step.polymtl.ca/~jyzero/ff3smv2.zip
up to South Figaro
I included a savestate if you want to see only the new stuff (Open the movie readonly and loadstate)
Here's some comments:
-Naming : I used default names even if selecting a shorter name would save some seconds on the whole movie. Having carachters with one letter name wouldn't be interesting
-Moogle battle: I choosed the Mog party to fight Marshall because this way I can unequip Mog during fight and let him learn a new dance.
-Marshall fight : I spent around 1000 rerecords on this fight. Here's why:
The optimal Marshall battle would require 2 consecutives critical hits. Critical hits hapens when 7e00be has a value of 22,54,68,75,7F,8D,81 or FD at the moment the critical hit calculation takes place (cheat 7e00befd makes you critical hit everytime). 7e00be doesn't increase when there is an attack animation on screen. This leaves a tight window of manipulation. The best I could do was 1 critical hit. I couldn't get the second critical hit without waiting too long (and getting my party killed). However I might be possible with battle set to slow. (Setting it to fast was a big mistake as the party is unaffected by this setting.)
-Road to Figaro Castle : I make an extra little step to avoid a fight in the desert.
For the next part I have a big dilemma. I can get the runningshoes now but it takes 45 seconds. Vargas and Ultros will be easier and faster with this relic but I doubt that I'll save the whole 45 seconds.
For the next part I have a big dilemma. I can get the runningshoes now but it takes 45 seconds. Vargas and Ultros will be easier and faster with this relic but I doubt that I'll save the whole 45 seconds.
Running shoes will probably benefit you in many more boss battles than just those two.. It depends, of course, on who you put the shoes on and how many battles they're in... but I think that ultimately the shoes will "pay for themselves", so I'd suggest getting'em.
