Also, Anyone who thinks they are able to manipulate the RNG in this game well, I would love your help in the future. Otherwise I'd say I'm doing fairly optimally, disregarding acknowledged laziness sacrificing a few frames here and there, in my test run.
I'm working on a Lua script at the moment that should facilitate luck manipulation and general strategery.
I'm not sure if I should bother with decoding the algos/RNG, since:
1) knowing what is going to happen doesn't mean being able to change anything, and
2) the bot could just brute force (with some degree of intelligence) the overworld instead; test a slew of different battles to see which has the overall best outcome.
If there's interest (i.e., if someone wants to deal with the strategy and route-planning end) I'll get right on it.
I would be willing to put more effort into this run in the future. I'm currently looking to finish my Bad Dudes TAS over Christmas Break.
If someone could decode a way for me to set my stats to values I choose, it would make finding levels I could feasibly beat bosses at much easier. It would help me to move on a bit more. I assume there would be memory addresses I can set to set these stats for testing purposes?
My previous strategy issue was finding an efficient way to level to an unknown level required to beat bosses.
I believe the first segment of the run is "How do we get to and beat Red Pison as fast as possible?" Lux seemed to be the best option. We have B Powers and B Protects at our disposal, but Lux needs to be fairly high leveled in order to damage Red Pison a fair amount.
I don't know if this next point ever got answered previously. This point is very important to implement the run appropriately, and I don't know the answer. If we recruit a new character and kill off Lux, does the new character gain full experience and function like a single character switch? What if we wanted to recruit a new character after that while Lux is still dead? Would that character's level be based on Lux's, the previous recruits, or a combination thereof? The most efficient way to level would be with a single character so experience doesn't get split in half.
Is there a character we could recruit that is good for killing Red Pison at a low level? We could actually kill off this character right before Red Pison is dead so that Lux gets full experience and we can get a new recruit later at a higher level. We would need to gain a substantial amount of experience before this recruit would consider joining us.
What character can beat S Brains at the lowest level? This will almost surely be our method of grinding levels for most of the run.
Can we get to S Brains and start killing them without having to level up at all after Red Pison? The castle boss for the Third Rune is actually supposed to be easier if you are at a low level.
What character combination would be best for beating all bosses thereafter including the final boss?
This is all based off of memory from about 6 months ago, so I may not recall some things correctly.
That's stuff I need to help me move on with this run in the future with regards to route planning. Dromiceius stated in the past that he could help to implement the run with regards to luck manipulation if I was able to lay out the realization.
We have a lot of resources available. If I could get some help with the issues and questions above, I would be glad to take another look at this run after I hopefully finish Bad Dudes in December.
My responses are based on memory from a long time ago too, but I hope they help.
Kirkq wrote:
If someone could decode a way for me to set my stats to values I choose, it would make finding levels I could feasibly beat bosses at much easier. It would help me to move on a bit more. I assume there would be memory addresses I can set to set these stats for testing purposes?
Good thing I made some notes.
Current HP: 7E1DC4 / 7E1E24 (2 bytes)
Maximum HP: 7E1DC6 / 7E1E26 (2 bytes)
Current MP: 7E1DC8 / 7E1E28 (2 bytes)
Maximum MP: 7E1DCA / 7E1E2A (2 bytes)
Power: 7E1DCC / 7E1E2C (2 bytes)
Magic: 7E1DCE / 7E1E2E
Speed: 7E1DCF / 7E1E2F
Defense: 7E1DDC / 7E1E3C (2 bytes)
I don't know if this next point ever got answered previously. This point is very important to implement the run appropriately, and I don't know the answer. If we recruit a new character and kill off Lux, does the new character gain full experience and function like a single character switch? What if we wanted to recruit a new character after that while Lux is still dead? Would that character's level be based on Lux's, the previous recruits, or a combination thereof? The most efficient way to level would be with a single character so experience doesn't get split in half.
If only one character is alive, that character will receive full experience. Both characters need to be alive to recruit a new character. The new character will replace the old partner, and his/her/its level will be based solely on the main character's level.
Is there a character we could recruit that is good for killing Red Pison at a low level? We could actually kill off this character right before Red Pison is dead so that Lux gets full experience and we can get a new recruit later at a higher level. We would need to gain a substantial amount of experience before this recruit would consider joining us.
The amount of experience you have does not affect whether a recruit will consider joining you. Winning some number of battles will simply re-randomize everyone's reaction to you.
What character can beat S Brains at the lowest level? This will almost surely be our method of grinding levels for most of the run.
I'm not sure who would be best at that, but that would probably be the only method of grinding levels after defeating the Patrof apprentice. At that point, you'd just grind until your level is high enough to defeat every remaining boss in the entire game.
Can we get to S Brains and start killing them without having to level up at all after Red Pison? The castle boss for the Third Rune is actually supposed to be easier if you are at a low level.
It may be difficult to defeat S.Brains at a low level, but otherwise I don't see why not. The Patrof apprentice's level is actually set as soon as the sage in Eygus moves out of the way, so any levels gained after that would not be wasted.
My previous strategy issue was finding an efficient way to level to an unknown level required to beat bosses.
Well, on the overworld, you can fairly quick brute-force manipulate a pretty good start to a battle. If I remember correctly, there was a 4 frame timespan on which you could start pressing a direction to walk wherein if your movements after beginning to move were identical, the corresponding patterns of enemy movement were also identical. Past that, changing the path you walk (particularly easy if you need to move diagonally), led to additional groups of 4 potential battle scenarios. I'm thinking that enough experience can be gained from battles on the overworld to handle Romus, Pison, Red Pison, and the Patrof apprentice without backtracking, standing around, or otherwise sacrificing many frames outside of battle.
Kirkq wrote:
The most efficient way to level would be with a single character so experience doesn't get split in half.
I agree. As such, my theory has been that we should use Lux solo until after the Star Rune and after a significant level grinding session outside Luze. We then should recruit Olvan, who can use some really good equipment (that we won't spend money on for Lux), has Fire 1 (for that uhh... "Dogun" I think in Gariso's castle), and has some healing spells (which for all we know, may be necessary for Gorsia). In addition, waiting until a higher level to recruit Olvan will mean he comes equipped with better armor.
Kirkq wrote:
Is there a character we could recruit that is good for killing Red Pison at a low level?
That's a good question, but to the best of my knowledge, no.
Kirkq wrote:
What character can beat S Brains at the lowest level? This will almost surely be our method of grinding levels for most of the run.
I would say almost definitely Lux, then Wilme second.
Kirkq wrote:
Can we get to S Brains and start killing them without having to level up at all after Red Pison? The castle boss for the Third Rune is actually supposed to be easier if you are at a low level.
I'm thinking "yes, we can". There are three reasons I can come up for why the Patrof apprentice is more easily defeated at a lower level, and I think we may be able to make use of them all:
(1) The Elnard -> 7th Saga leveling stat bonus was removed for "The 7th Saga", but the stat calculations for an opposing apprentice include this bonus. This bonus doesn't really come into play until level 10, but it becomes an increasingly large discrepancy that becomes increasingly larger faster every additional 10 levels. Low levels would minimize this severe disadvantage.
(2) Depending on the apprentice and the level, some offensive spells might not yet be learned. A Wilme or Kamil battle is significantly easier if they cannot use Fire 2 on you.
(3) The Water Rune heals a greater percentage of your HP when you have lower maximum HP, so in a fight between lower levels, the Water Rune advantage will be greater.
Regardless, we will hopefully be able to go: B Power, Attack, Defend, Attack and win in 4 rounds.
Kirkq wrote:
What character combination would be best for beating all bosses thereafter including the final boss?
In my console experience: Lux and Olvan. This does have a lot to do with Olvan's raw power, equipment use, and as assumption that he will reach a high enough level that he gets his final spells (including Heal 3 and Revive1). Additionally, Metal Pison and the dragon-clone (Dogun?) is Gariso's castle that has high defense and resists except for Fire are concerns that Lux and Olvan would be better equipped to handle as a duo than any others I can think of (I'm discounting Wilme here since I find him too fragile and his fire magic too weak to be really useful).
Nitrodon wrote:
The Patrof apprentice's level is actually set as soon as the sage in Eygus moves out of the way
Are we absolutely certain about this? I don't think this is true, but I would be glad to be wrong.
Unfortunately, grad school has been keeping me from toying with 7th Saga, but I think Animand has been working with it somewhat. Maybe he'll drop in with anything he's found.
I'd love to do work on this movie, but I'm holding back because I'm unsure on what sort of power we'll have to manipulate critical boss battles. R.Pison is critical, but I'd contend that the needed S.Brain levelling for the final boss (or the previous boss series) would eliminate the worrying about levels for ALL* the other ones save Dragon.
Naturally, I concur that the best levelling method is a single*, condensed block of killing S.Brains for two reasons: while levelling later (in the past) can be faster, having the levels earlier can likely save more time, and the one levelling block avoids multiple spots of viewer boredom.
Points of interest (*):
---I never fought any battles between R.Pison and the Patrof apprentice. I avoided them all, and always managed to win that fight *on console* with Lux and minimal trouble. I'm not at all worried about that fight.
---There are things that may detract from the "single" part of this. For example, I suggest killing some Brains around Guntz in order to manipulate some Rcvry drops.
---I've had the Sky Rune apprentice want to join my party. Having Olvan hold the Sky Rune and getting him to join you would be insanely useful. This, of course, would require not having him before (should be discussed with Lux solo or with another partner) and knowing how to manipulate/check it. *raises an eyebrow at Nitrodon*
I'm not really here. Pressing Up+Down too many times glitched your graphics. And audio. Yeah.
The Patrof apprentice's level is actually set as soon as the sage in Eygus moves out of the way
Are we absolutely certain about this? I don't think this is true, but I would be glad to be wrong.
I am completely sure of this. His level will be 5 greater than yours at that point, and will never increase afterward.
Animand wrote:
I'd love to do work on this movie, but I'm holding back because I'm unsure on what sort of power we'll have to manipulate critical boss battles. R.Pison is critical, but I'd contend that the needed S.Brain levelling for the final boss (or the previous boss series) would eliminate the worrying about levels for ALL* the other ones save Dragon.
In my experience, enemy actions are completely impossible to manipulate (except for healing magic, since that's only used at low HP). In caves, castles, and similar, manipulating the RNG (for dodges, criticals, etc.) in the middle of battle is only possible by changing the actions done by each character. On the overworld, the RNG will advance in time due to waves crashing against the shore.
Naturally, I concur that the best levelling method is a single*, condensed block of killing S.Brains for two reasons: while levelling later (in the past) can be faster, having the levels earlier can likely save more time, and the one levelling block avoids multiple spots of viewer boredom.
Actually, S.Brains give more experience than any other non-boss enemy in the game, so I don't see how leveling later would be faster anyway.
Points of interest (*):
---I never fought any battles between R.Pison and the Patrof apprentice. I avoided them all, and always managed to win that fight *on console* with Lux and minimal trouble. I'm not at all worried about that fight.
Noted. In a TAS, it might be useful to manipulate Lejes there due to his low defense.
---There are things that may detract from the "single" part of this. For example, I suggest killing some Brains around Guntz in order to manipulate some Rcvry drops.
The experience may also make it easier to start killing S.Brains.
---I've had the Sky Rune apprentice want to join my party. Having Olvan hold the Sky Rune and getting him to join you would be insanely useful. This, of course, would require not having him before (should be discussed with Lux solo or with another partner) and knowing how to manipulate/check it. *raises an eyebrow at Nitrodon*
Olvan's response to you = 7E5022. This number ranges from 0 to 99, and changes randomly whenever 7E5120 reaches 0. If his response value is 40 or greater, Olvan will offer to join Lux if he has the Sky Rune. (If he doen't have the Sky Rune, the number needs to be 60 or greater instead.)
That's stuff I need to help me move on with this run in the future with regards to route planning. Dromiceius stated in the past that he could help to implement the run with regards to luck manipulation if I was able to lay out the realization.
I started on a bot yesterday. I uploaded what I've come up with thus far because I'm done working on it for now.
The finished design, as I currently envision it, will have two primary tasks: 1) find an optimal battle on the world map, and 2) find the optimal boss fight within a dungeon. Lofty goals, but check my pseudocode to see if I'm not on the right track. :D
It first generates input on the world map. The idea is to have the bot either find and kill S.Brains, or walk into a dungeon, and then kill the boss. It depends on when you execute the script, and whether you specify "targetcoords" that the bot should try to approach. If you were to execute the script standing next to Romus' castle, for instance, the bot might walk in there, and then, if you had specified a path for the bot to follow through the dungeon, try to kill Romus.
The bot will also manage the battles themselves, testing the use of defend, as well as a subset of items and magic, so as to defeat bosses as fast as possible.
When the battle is done, the script examines a set of memory addresses, and uses them to calculate the "fitness" of that outcome. You'll need to modify the script on a situation-by-situation basis to indicate what kind of outcome is most desirable.
function comparestats(init, current)
damagetaken = init.playerhp - current.playerhp
--[[for i,v in ipairs(init.enemy) do
damageinflicted = damageinflicted + (v - current.enemy[i])
endbørk]]
return -framecount
end
Here, for example, the user only cares about "negative framecount"; the higher the framecount is, the lower the return value will be.
You'll also be able to test for Lux' hp/mp/statistics as well as item drops and item usage vis-a-vis comparing the inventory chunks of the initial state vs the state of the current attempt.
It should be pretty thorough, if it all works in practice- most randomness within the bounds of reasonable testing should be accounted for, leaving almost nothing to chance... besides the user's choice of strategy, that is. ;)
Comments and suggestions on the design and its purported functionality are welcome; the code is fresh and untested (unfinished, in fact) so I may well have made obvious oversights.
Incidentally, has anyone any experience with the "corruption" this (Game)FAQ speaks of?
IFerrel wrote:
*Now I write the offsets of the items, I put 50 offsets,the game can have up
to 99 item spaces,but you have a high risk of corrupting the state,
Pokemon and Chrono Trigger have shown that a little deliberate memory corruption can go a long way.
Wow. I didn't realize so many people had so much information on this run =p.
Nitrodon, thanks for answering almost every one of my questions.
Honestly, I think we may need to recruit someone to kill Red Pison at a lower level. Just having two turns to his one is a huge advantage. We could feasibly kill them off at the end of the fight so Lux gets all the experience. We can revive them later before recruiting Olvan. Actually we could possibly recruit Olvan here and kill off Lux during the fight.
So I'm hearing that we need to recruit Olvan, which sounds logical. Olvan has the highest attack power right? One of Olvan or Lux is probably better than the other for S Brain killing. Perhaps we will need both for the first few. We will probably get enough money during S Brain grinding to purchase suitable gear, so having Olvan start with optimal gear may not be the best method. What I'm hearing is that at S Brains we should kill off Lux for the rest of the game.
If what Nitrodon says is true about the apprentice level, we can walk into Patrof, leave, level up on S Brains, return with the wind rune, and probably one shot the apprentice later.
Animand wrote:
---I've had the Sky Rune apprentice want to join my party. Having Olvan hold the Sky Rune and getting him to join you would be insanely useful.
This is possible a solution to the above speculation worth considering, but I think since we could just one shot the apprentice later the choice we make should focus on the other variables. (Olvan's equipment and level upon joining us.) If having Olvan at a low level helps us kill S Brains fastest, this is probably a very good option.
Dromiceius wrote:
Incidentally, has anyone any experience with the "corruption" this (Game)FAQ speaks of?
I believe what he is saying is that the maximum number of items you SHOULD put in is 50. I seem to recall that there is a limit on items below 50 even, so this would be unfeasible.
New point: I recall there is a point later in the game where having Olvan creates a slightly different scenario. Can anyone elaborate on this point? Does he have to be your main character for this to occur, or can he be your apprentice? Is the Olvan path faster (less walking), or is the Normal Route faster?
If what Nitrodon says is true about the apprentice level, we can walk into Patrof, leave, level up on S Brains, return with the wind rune, and probably one shot the apprentice later.
If I remember correctly, S.Brains are found near Luze. It's impossible to reach any towns past Bone (along any of the three possible routes) unless you have both the Water Rune and the Star Rune.
New point: I recall there is a point later in the game where having Olvan creates a slightly different scenario. Can anyone elaborate on this point? Does he have to be your main character for this to occur, or can he be your apprentice? Is the Olvan path faster (less walking), or is the Normal Route faster?
He needs to be the main character. If you talk to a certain person in Bone, he will give you the key to Grime Tower, and you can use that to warp to somewhere near Telaine. Note that he can still take the same "normal route" that anyone other than Esuna or Lux will take. In fact, without savestates, the enemies in the tower to the north are likely too difficult at that point in the game, so you'll want to take the normal route on console.
In caves, castles, and similar, manipulating the RNG (for dodges, criticals, etc.) in the middle of battle is only possible by changing the actions done by each character.
Given a high enough speed stat, choosing actions during fights in dungeons with the premise of dodging attacks and avoiding damage could make choices like Escape make sense in order to get better results from enemy actions. An important example would be keeping R.Pison from successfully one-hit KOing you, especially if you have a partner at that point.
Nitrodon wrote:
Actually, S.Brains give more experience than any other non-boss enemy in the game, so I don't see how leveling later would be faster anyway.
Battles with multiple high XP enemies could be fast enough to present a viable option. I doubt it, but I still intend to be sure.
Nitrodon wrote:
Olvan's response to you = 7E5022. This number ranges from 0 to 99, and changes randomly whenever 7E5120 reaches 0. If his response value is 40 or greater, Olvan will offer to join Lux if he has the Sky Rune. (If he doen't have the Sky Rune, the number needs to be 60 or greater instead.)
Fantastic.
I'm not really here. Pressing Up+Down too many times glitched your graphics. And audio. Yeah.
Work on the bot is progressing steadily. Rather than using brute force (which takes too damn long) it "evolves" the fastest possible sequence for an individual battle by means of a genetic algorithm. It's going well, in other words- the hard part is out of the way. :)
Just one thing I'd like some clarification on something before I move on to the next part:
Kirkq wrote:
On the other hand the sequence done by pison is only altered by the random number generator on the frame you hit A to pause, so I could run tests in real time through sequences.
To what extent can calling the menu affect the RNG? I thought that the only way to arbitrarily manipulate a dungeon/boss fight was to wait on the world map and wait for the tide to come in, then enter the dungeon. I know that using a seed advances the RNG, but are you saying that using it on different frames can give you more leverage?
If I recall correctly, standing still in front of pison for 3-4 frames advances the RNG each few frames. The RNG advances constantly while in dungeons, just not DURING a dungeon fight.
I tested 4 or 5 Pison fight sequences in my old Test Run and I believe to get a different sequence only required a few frames of waiting.
What I was saying is the RNG for the beginning of the fight is locked as soon as you open up the menu to talk to him.
Specifically, the RNG advances several times each frame outside battle if there are enemies in the area. I can only think of three bosses which are not fought in areas with random encounters, so you will be able to manipulate the RNG easily immediately before most boss fights (and in the last random encounter area before the boss in all other cases).
He needs to be the main character. If you talk to a certain person in Bone, he will give you the key to Grime Tower...
I'm 95% certain Olvan can get the key if he is the partner AND the leader is dead. Ditto on the hidden room in the hidden cave. However, I'm also 95% certain that both the boat to Pandam and the remote control submarine + overworld trekking are more effective routes. Both towers are too far out of the way and too convoluted to realistically save time. The treasures also suck, especially given how leveled one must be to survive the two of them.
Grinding suggestion:
Once magic power reaches approximately 160, the most efficient method of collecting experience in the present becomes using a wide spread elemental spell against three Sword enemies, ending the battle in one move for a total of 3303 EXP. Lux with un-manipulated stats plus manipulated seeds will accomplish this at approximately Level 50. Esuna would be around 38.
Also, given the "Elnard boost", it may be beneficial to kill off the leader at a point and coast through the game solo with a boosted partner. Eventually the capability gap between the leader and a partner acquired at a high level is so large that the leader becomes a liability.
Hmm. I'm not sure if you're manipulating le'elup stats...and, unless it results in a different crit/miss setup, defend->attack should be faster than attack spam against Pison, as it means you take less damage(One round of high defense in two) while dealing roughly the same(~2x damage when you attack), so you have to heal less, meaning possibly a round of combat saved. Since you mixed them, I'm guessing that you (or the bot) already took this into account/it's not enough to get you an extra hit between heals.
I can say there is one glaring flaw that seems to pop out- in those last 1 on 3 battles, you don't end the battle facing the middle one(by killing it last), and it (seems to) take much longer for the game to turn you to the center after the battle than it would if you did in battle. Did you test that?
Otherwise, you can almost hear LUX saying...
"RANDOM ENCOUNTER? WHAT IS THIS RANDOM ENCOUNTER YOU SPEAK OF?" which is good.
"A"? ...hmm. Le'elup texts are one point for short-name.
Remember, dungeon battles cannot be manipulated once started. In those cases where i used attack instead of defend attack pison would have dodged.
I don't think this is anywhere near optimal. The very first fight is terrible. Seeds were probably used wrong. At the inn I should have just saved instead of resting.
However while I would love to take into account things like winning battles without extra scrolling I also have to take into account oftentimes what the game gives me as an option. If I have to choose between 'wait 100 frames' or 'lose a few frames on scrolling' obviously its a no brainer.
I also anticipated having Valsu as an option in Bonro. I need to do a little more research in figuring out when the Bonro apprentices are decided as well as memory addresses for their responses to you.
Shorter name is used because the character's name is mentioned in conversations/level up texts/ item texts. And I mean its not like lux is the most entertaining name in the first place.
The Bonro/Zellis apprentice locations are decided when you start the game. The memory addresses corresponding to these locations are as follows:
7E0737: Bonro item shop
7E0738: Bonro weapon/armor shop
7E0739: Bonro tavern
7E073A: Zellis tavern
7E073B: Zellis armor shop
7E073C: Zellis inn
The apprentice values are as follows:
Kamil: 0
Olvan: 1
Esuna: 2
Wilme: 3
Lux: 4
Valsu: 5
Lejes: 6
For future reference, the Patrof apprentice is at 076B.
Valsu's response to you is at 50A2. For him to offer to join Lux, this value needs to be less than 40 (preferably less than 20 due to text length). Apprentice responses change every 4-11 battles based on a counter at 5121.
My old test run: http://dehacked.2y.net/microstorage.php/info/1084734740/LuxPison.smv
I did not know that person in Lemele gave you a potion, that's a fast free potion.
Other than that, I was able to level from 3-4 in one fight, which I'm sure is fastest. (You fought a hermit on your way to Rablesk)
Getting past Red Pison efficiently is the difficult part.
There is a shield destroyer hidden in Bonro. (item that casts defense 2)
Pison defense is 50.
R pisons is 100 (using defense2 will bring him to 50). he also only barely has more hp than regular pison.
The only remaining issue is Pison's power is 40 while R Pison's is 120.
However we have some flexibility here if we choose Valsu as a partner. We can manipulate R-Pison to hit Lux while Valsu keeps him healed, then have Valsu take a fatal hit before Lux deals the final blow.
Nitrodon, do you know how the level of the apprentices is determined? Is it based on when you walk into the town? iIt would be nice to manipulate say, a level 7 Valsu early in the game, and a partner 2 levels higher than we are later on.
