Secret of Mana

2015-07-22 05:17:51 pm

never a frown

So I had enough interest in me to calculate a rough estimate as to how much faster the Japanese version would be for "Magic%" than the English version and it seems that excluding possible differences such as transition speed, lag, et cetera (not sure that there should be any difference?) that the Japanese version should be around ten seconds faster than English for the category; according to my analysis (comparing J on higan (I do have it on VC but used this for ease) versus my SDA recording which is VC-U) it loses 3-4 seconds at Neko's house and about half a second from the Soldier's dialogue of all things but it appears to save 4-6 seconds from the intro/waterfall cutscene, and around 3 seconds (each) in the sword cutscene, pre-Mantis cutscene, and post-Mantis cutscene. The Potos banishment meeting speed seemed about the same.

Here's the interesting part: The barrel animation (to open the glitched menu) is about half a second faster than on English primarily because the Japanese version's 'barreled!' message disappears in ~65 frames instead of ~99; however, not only does it appear that the Japanese version has slightly different shaking animations (will need to analyze this further later on to make sure this is the case), but the inputs for calling "The End" are not the same, and I do not know what those are, assuming those exist, so I mainly wanted to make this post to see if Crow (or HHS, or anyone else) would know, or would know how to find them. Even though it is possible that these inputs could be a bit faster for the Japanese version, it could be possible that it's an even longer sequence, potentially long enough to nullify the gains that are primarily the cause of faster text speed.

So I had enough interest in me to calculate a rough estimate as to how much faster the Japanese version would be for "Magic%" than the English version and it seems that excluding possible differences such as transition speed, lag, et cetera (not sure that there should be any difference?) that the Japanese version should be around ten seconds faster than English for the category; according to my analysis (comparing J on higan (I do have it on VC but used this for ease) versus my SDA recording which is VC-U) it loses 3-4 seconds at Neko's house and about half a second from the Soldier's dialogue of all things but it appears to save 4-6 seconds from the intro/waterfall cutscene, and around 3 seconds (each) in the sword cutscene, pre-Mantis cutscene, and post-Mantis cutscene. The Potos banishment meeting speed seemed about the same.

Here's the interesting part: The barrel animation (to open the glitched menu) is about half a second faster than on English primarily because the Japanese version's 'barreled!' message disappears in ~65 frames instead of ~99; however, not only does it appear that the Japanese version has slightly different shaking animations (will need to analyze this further later on to make sure this is the case), but the inputs for calling "The End" are not the same, and I do not know what those are, assuming those exist, so I mainly wanted to make this post to see if Crow (or HHS, or anyone else) would know, or would know how to find them. Even though it is possible that these inputs could be a bit faster for the Japanese version, it could be possible that it's an even longer sequence, potentially long enough to nullify the gains that are primarily the cause of faster text speed.

Crow!

Edit history:

Crow!: 2015-07-23 11:41:14 am

Crow!: 2015-07-23 11:40:49 am

Crow!: 2015-07-23 11:38:50 am

2015-07-23 11:38:13 am

What's that gemma?

The reason the sequence for calling the The End screen is different is because each event in the game is not of a fixed size - with more or less dialogue, you'll have to advance the game's dialogue choice event pointer to a different place to reach the position of the same event number.

What you'll need to do is to make a file similar to the game event script that Yagamoth linked me to one time (I'm not sure who actually made that file... HHS seems likely) for the JP version of the game, then figure out where the glitched barrel / cannon travel menu starts you, then plot a course where you jump from dialogue character to dialogue character while avoiding memory addresses of real commands, until eventually you reach the "summon the The End screen" command, and you let the game execute that.

As for how you plot a course, the basic idea is each time you press B, you go forward 1 byte, while each time you press left then B, you go forward 256 bytes (which is "100" bytes in Hexadecimal, which is how the memory addresses will probably be listed in a game event script printout). If necessary, pressing left (N+1) times then pressing B will move you forward (256-N) bytes, but using that method is a serious pain, and should be avoided if possible.

The reason the sequence for calling the The End screen is different is because each event in the game is not of a fixed size - with more or less dialogue, you'll have to advance the game's dialogue choice event pointer to a different place to reach the position of the same event number.

What you'll need to do is to make a file similar to the game event script that Yagamoth linked me to one time (I'm not sure who actually made that file... HHS seems likely) for the JP version of the game, then figure out where the glitched barrel / cannon travel menu starts you, then plot a course where you jump from dialogue character to dialogue character while avoiding memory addresses of real commands, until eventually you reach the "summon the The End screen" command, and you let the game execute that.

As for how you plot a course, the basic idea is each time you press B, you go forward 1 byte, while each time you press left then B, you go forward 256 bytes (which is "100" bytes in Hexadecimal, which is how the memory addresses will probably be listed in a game event script printout).  If necessary, pressing left (N+1) times then pressing B will move you forward (256-N) bytes, but using that method is a serious pain, and should be avoided if possible.

Overfiendvip

2015-07-23 12:12:26 pm

never a frown

Thank you. I was interested in why these were different and now I know. I will probably look into this sometime after SGDQ, unless someone else finds it before I attempt to.

For reference, the L-20-L-7-L-1-L-1 series still works on Japanese in that it will lead you to that point in the menu without interruption (from things such as being told to "get out of here fast"), but once you reach that point, nothing happens (as if it is still expecting more inputs to occur, so to speak), so perhaps it could be a very similar series...

Crow!

2015-08-02 07:36:33 pm

What's that gemma?

While watching Stinger earlier today, I had a thought:

During the Fire Gigas fight, there's a pretty big risk of a low damage roll wrecking your day. What about swapping the Sword onto the Boy after the overcharge is underway? The sword is 4 more base damage, which gets multiplied a lot by the overcharge. You could then simply charge to level 2 "normally" during the Wall Face fight before switching back to the spear.

It would be one less kill for the Sprite's sword experience, which might mean having to kill an extra rabite or needing to have the Sprite initiate the Mech Rider fights. I'm also not sure whether charging using more in-game time during the Wall Face fight is a problem - I don't think it would be, but it could lead to missing an eye open/close cycle or something.

Yagamoth

2015-08-03 03:32:03 am

INTJ

I like that idea crow, I'll try that on my runs today Smiley

Yagamoth

Edit history:

Yagamoth: 2015-08-10 06:39:28 pm

2015-08-10 02:39:56 pm

INTJ

Completed my route optimization stuff today, and also managed to create a new savestate-pack alongside it:
http://bombch.us/BRxi

Here are the rough notes, I'll rewrite them tomorrow or after the stream Smiley

=====================================

After Water Palace: Kill Mushroom (special thanks to SNESWarrior for input)
- There can only ever be 2 enemies on that screen at once
- Mushroom is more reliable to kill than any goblin, and probably faster (can be hit in either damaged-animation)
- if you kill the mushroom, you can force both, the Rabite and the Flower to spawn after that (make sure to not move too fast to the Rabites spawn-position before the Mushbooms damage numbers fully disappear)
- Aim to kill all flowers and Rabites along the way, and use spawn manipulation to make all of them spawn
- Total kill count is 8

Kill Flower with charge before canon travel
- After Tropicallo on the flower before the canon
- You can walk the Sprite into the flower and wake it up, and then walk the Sprite towards the canon-entrance
- Just charge sword to level 1 and release

Earth palace pole-jump despawn strat
- Running diagonally down left far past the first goblin at the top will make the goblin on the right AND left side despawn

Small change in the ruins
- 3rd room, boop the wall a little after the first sword to despawn the 2nd sword for sure

2 Small changes pandora fields run
- Diagonal up left dash after the mushrooms
- Straight up walk and the up left dash at the start

Small change in woods after Sylph orb
- Going to the left in the screen leading towards the cave gives an easier route

Small change in mushroom field
- No need to account for the nonexistant 3rd character route
- (I think Incro already does the newer route, and Stinger probably too, I was probably just too lazy to pick that up before)

Ice Country small changes (~3:40 in my last broadcast) (Special thanks to Andyman for input)
- Minor thing, first screen: Snowman-ing the first two wolves and pushing the along to despawn wolves in the top left
- 2nd screen has a complete idea of the route now including a chance to despawn the turtle (not 100%)
- 3rd screen now goes up again to spawn the first cloud and walks to the left to spawn the 2nd one in the middle. From there either hit the remaining wolf or run straight up

Gnome weapon on Ice Canon
- Fire Palace, entrance strat ... I'm not sure how reliable that is. I go to hit the left bow rat with the boy first and then immediately with the Girl afterwards. has a high chance up straight up killing the left bow rat with Gnome weapon (uses 2 controllers - keep in mind I can now handle all buttons and both D-Pads at the same time :P)
- The general idea is to handle the rats by casting an elemental weapon during canontravel to the desert now

Potential balloon skip in the sewers
- Cast Gnome weapon on canon
- hit the blue drops with gnome weapons and pray
- Seriously though, I had around a 50% success chance... It's not terrible, I just couldn't come up with anything reliable. But it saves around 8-13 seconds
- Petrified are like snowmen - ignored by your AI partners, so they just walk past

=====================================

I forgot to credit a lot of people, since I started way too late to realize, that I probably should credit people Tongue

Completed my route optimization stuff today, and also managed to create a new savestate-pack alongside it:
http://bombch.us/BRxi

Here are the rough notes, I'll rewrite them tomorrow or after the stream :)
=====================================

[b]After Water Palace: Kill Mushroom (special thanks to SNESWarrior for input)[/b]
- There can only ever be 2 enemies on that screen at once
- Mushroom is more reliable to kill than any goblin, and probably faster (can be hit in either damaged-animation)
- if you kill the mushroom, you can force both, the Rabite and the Flower to spawn after that (make sure to not move too fast to the Rabites spawn-position before the Mushbooms damage numbers fully disappear)
- Aim to kill all flowers and Rabites along the way, and use spawn manipulation to make all of them spawn
- Total kill count is 8

[b]Kill Flower with charge before canon travel[/b]
- After Tropicallo on the flower before the canon
- You can walk the Sprite into the flower and wake it up, and then walk the Sprite towards the canon-entrance
- Just charge sword to level 1 and release

[b]Earth palace pole-jump despawn strat[/b]
- Running diagonally down left far past the first goblin at the top will make the goblin on the right AND left side despawn

[b]Small change in the ruins[/b]
- 3rd room, boop the wall a little after the first sword to despawn the 2nd sword for sure

[b]2 Small changes pandora fields run[/b]
- Diagonal up left dash after the mushrooms
- Straight up walk and the up left dash at the start

[b]Small change in woods after Sylph orb[/b]
- Going to the left in the screen leading towards the cave gives an easier route

[b]Small change in mushroom field[/b]
- No need to account for the nonexistant 3rd character route
- (I think Incro already does the newer route, and Stinger probably too, I was probably just too lazy to pick that up before)

[b]Ice Country small changes (~3:40 in my last broadcast) (Special thanks to Andyman for input) [/b]
- Minor thing, first screen: Snowman-ing the first two wolves and pushing the along to despawn wolves in the top left
- 2nd screen has a complete idea of the route now including a chance to despawn the turtle  (not 100%)
- 3rd screen now goes up again to spawn the first cloud and walks to the left to spawn the 2nd one in the middle. From there either hit the remaining wolf or run straight up

[b]Gnome weapon on Ice Canon[/b]
- Fire Palace, entrance strat ... I'm not sure how reliable that is. I go to hit the left bow rat with the boy first and then immediately with the Girl afterwards. has a high chance up straight up killing the left bow rat with Gnome weapon (uses 2 controllers - keep in mind I can now handle all buttons and both D-Pads at the same time :P)
- The general idea is to handle the rats by casting an elemental weapon during canontravel to the desert now

[b]Potential balloon skip in the sewers[/b]
- Cast Gnome weapon on canon
- hit the blue drops with gnome weapons and pray
- Seriously though, I had around a 50% success chance... It's not terrible, I just couldn't come up with anything reliable. But it saves around 8-13 seconds
- Petrified are like snowmen - ignored by your AI partners, so they just walk past

=====================================

I forgot to credit a lot of people, since I started way too late to realize, that I probably should credit people :P

Crow!

Edit history:

Crow!: 2015-08-15 01:27:55 am

Crow!: 2015-08-15 01:26:04 am

Crow!: 2015-08-14 03:35:12 pm

2015-08-14 03:34:34 pm

What's that gemma?

There's been some discussion lately about glitchless, and in particular about these formulae brought to us by the TAS guys:

Quote from HHS:

Mana Sword Power = Total weapon experience/100+Total weapon levels/8 (when the man is using the sword with Mana Magic)
Force = Attack*(1+(Charge level+Mana Sword Power)/4)

Problem is, this is completely wrong. If the game worked as written, a charge level 1 attack would deal 25% more damage than a charge level 0 attack. Instead, it deals 50% more, like the way spells do.

So, I decided I'd check whether the supposed Mana Sword Power is being factored in at all, and if so, how.

1p2c with 0.38 Mana Sword Power:
Unable to deal damage with charge level 2 mana sword.
With acid storm: 80-110 damage with charge level 2

so, I was dealing about 550 damage strikes at charge level 2.

1p2c with supposed 4.28 mana sword power:
160-190 damage at charge level 2 (740 full damage rolls)
70-100 damage at charge level 1 (650 full damage rolls)
no damage at charge level 0 (550 or so)

weapon screen lists 187 as the attack power for the boy, so +90-100 damage on full damage hits per charge level is consistent with a 50% bonus per charge level theory.

Anyway, that additional +3.9 mana sword power came out to an additional ~190 damage - almost equal to the attack power of the boy.

TL;DR: In other words, the mana sword power seems to fit the formula the TAS guys give, even though the charge level portion does not agree (each charge level is worth 2x what they claim it does).

(For future reference, it would have been much easier to test this by using cheat codes to set 7ee3cd=38 and 7ee5cd=07, which is how Mana Magic is added to your spell list at the end of the Thanatos fight, rather than having to deal with the Mana Beast).

-----------

To get max damage vs the Mana Beast with simple sword slashes (i.e. no charging), you would need a Mana Sword Power of about 10. As there are only 8 weapons, this is pretty much certain not to happen during a speed run.

There's been some discussion lately about glitchless, and in particular about these formulae brought to us by the TAS guys:

[quote=HHS]
Mana Sword Power = Total weapon experience/100+Total weapon levels/8 (when the man is using the sword with Mana Magic) 
Force = Attack*(1+(Charge level+Mana Sword Power)/4)
[/quote]

Problem is, this is completely wrong.  If the game worked as written, a charge level 1 attack would deal 25% more damage than a charge level 0 attack.  Instead, it deals 50% more, like the way spells do.

So, I decided I'd check whether the supposed Mana Sword Power is being factored in at all, and if so, how.

1p2c with 0.38 Mana Sword Power:
Unable to deal damage with charge level 2 mana sword.
With acid storm: 80-110 damage with charge level 2

so, I was dealing about 550 damage strikes at charge level 2.

1p2c with supposed 4.28 mana sword power:
160-190 damage at charge level 2 (740 full damage rolls)
70-100 damage at charge level 1 (650 full damage rolls)
no damage at charge level 0 (550 or so)

weapon screen lists 187 as the attack power for the boy, so +90-100 damage on full damage hits per charge level is consistent with a 50% bonus per charge level theory.

Anyway, that additional +3.9 mana sword power came out to an additional ~190 damage - almost equal to the attack power of the boy.

[b]TL;DR[/b]: In other words, the mana sword power seems to fit the formula the TAS guys give, even though the charge level portion does not agree (each charge level is worth 2x what they claim it does).

(For future reference, it would have been much easier to test this by using cheat codes to set 7ee3cd=38 and 7ee5cd=07, which is how Mana Magic is added to your spell list at the end of the Thanatos fight, rather than having to deal with the Mana Beast).

-----------

To get max damage vs the Mana Beast with simple sword slashes (i.e. no charging), you would need a Mana Sword Power of about 10.  As there are only 8 weapons, this is pretty much certain not to happen during a speed run.

Crow!

Edit history:

Crow!: 2015-08-16 01:56:03 pm

Crow!: 2015-08-16 12:50:52 pm

2015-08-16 12:50:07 pm

What's that gemma?

In my glitchless race yesterday, I achieved a Mana Sword Power of 4.61. That, combined with a Moon Energy, was enough to deal about 860 damage to the Mana Beast per normal slash. Adding in Lunar Boost brought the damage to 999.

If you managed to get all weapons but the sword to 0:90, and just left the Sword at 2:00, that would be worth 6.55 Mana Sword Power, which should be more than enough to make Lunar Boost not even needed. So, there's a fair amount of leeway in optimizing the Mana Sword Power route.

Mana Magic's animation is very slow and needs to be canceled with something. The ideal in the 3-5 Mana Sword Power range is probably to cast Moon Energy between cycles, then cast Mana Magic as the Mana Beast gets into position and cast Lunar Boost ASAP afterward. Somewhere above 5, you probably want the Girl to be facing the Boy as Mana Magic is cast, then cast Moon Energy.

Moon Energy gives X+1 crits, where X is the level of the spell. I suspect level 2 is ideal. (During the race, I erroneously remembered it as X+2 crits).

------

Here's what HHS says about attack animations:

Quote from HHS:

Player attack action is taken from D10480[F*40+D10000[A*144+B*16+C*4+D*2+E]], and enemy attack action is taken from D105C0[G*32+H*16+D10240[A*144+B*16+C*4+D*2+E]], where:
A = 0:Less than 16 pixels away, 1:Less than 32 pixels away, 2:Less than 48 pixels away, 3:48 or more pixels away
B = Charge level
C = 0:Nothing, 1:Hurt or opening chest, 2:Hurt badly, 3:Sleeping
D = Floating
E = 1 every other frame
F = Weapon type
G = Enemy type
H = Current weapon (each enemy can have 2)

So, apparently there are 4 zones for your attack animation depending on how close/far you are from whatever the game thinks is your target (which I assume is the closest enemy to your position).

This explains some things we've noticed before:
- If you're too close to the chick/spring/whatever used for the Mana Fortress barrier skip, you get a version of the sword jump that doesn't work.
- When you're too close to a Robin Foot, the punch animation from the glove moves a pathetically short distance up the ramp.
- You have to be as far away from Joch as you can go for the stalagmite jump to travel far enough.

The other notable factor is that one of two animations happen every other frame (so if you pause buffer an overcharge that can change animations, frame rules probably mean that you'll always get the same attack from the same distance from the enemy). Apart from pause buffering, this part is random as far as a human player is concerned.

I tried to put together a table documenting the behavior of the sword at specific ranges, but I'm finding it to be difficult since I don't know where the "center" of you and your target are for the distance measurements. One thing that is certain, though, is that if you are far enough away from all enemies and NPCs, you will always do a horizontal swing with the sword. (For example, in Gaia's Navel, once you've killed the Goblins, you will never miss the skull.)

The level 0 sword jump does not happen against an enemy in the normal state.

------

Here's another thing in HHS's post that I don't understand:

Quote:

Weapon effects accumulate if another weapon enhancing spell is cast before the time runs out.

After some testing, this definitely does not refer to the number of uses left for a saber spell, and it also does not refer to the damage boost of a saber spell, and it also does not refer to the status effects of a saber spell. It could be that this is referring specifically to Mana Magic.

In [url=http://www.twitch.tv/iicrowii/v/11211201]my glitchless race[/url] yesterday, I achieved a Mana Sword Power of 4.61.  That, combined with a Moon Energy, was enough to deal about 860 damage to the Mana Beast per normal slash.  Adding in Lunar Boost brought the damage to 999.

If you managed to get all weapons but the sword to 0:90, and just left the Sword at 2:00, that would be worth 6.55 Mana Sword Power, which should be more than enough to make Lunar Boost not even needed.  So, there's a fair amount of leeway in optimizing the Mana Sword Power route.

Mana Magic's animation is very slow and needs to be canceled with something.  The ideal in the 3-5 Mana Sword Power range is probably to cast Moon Energy between cycles, then cast Mana Magic as the Mana Beast gets into position and cast Lunar Boost ASAP afterward.  Somewhere above 5, you probably want the Girl to be facing the Boy as Mana Magic is cast, then cast Moon Energy.

Moon Energy gives X+1 crits, where X is the level of the spell.  I suspect level 2 is ideal.  (During the race, I erroneously remembered it as X+2 crits).

------

Here's what HHS says about attack animations:

[quote=HHS]Player attack action is taken from D10480[F*40+D10000[A*144+B*16+C*4+D*2+E]], and enemy attack action is taken from D105C0[G*32+H*16+D10240[A*144+B*16+C*4+D*2+E]], where: 
A = 0:Less than 16 pixels away, 1:Less than 32 pixels away, 2:Less than 48 pixels away, 3:48 or more pixels away 
B = Charge level 
C = 0:Nothing, 1:Hurt or opening chest, 2:Hurt badly, 3:Sleeping 
D = Floating 
E = 1 every other frame 
F = Weapon type 
G = Enemy type 
H = Current weapon (each enemy can have 2)[/quote]
So, apparently there are 4 zones for your attack animation depending on how close/far you are from whatever the game thinks is your target (which I assume is the closest enemy to your position).

This explains some things we've noticed before:
- If you're too close to the chick/spring/whatever used for the Mana Fortress barrier skip, you get a version of the sword jump that doesn't work.
- When you're too close to a Robin Foot, the punch animation from the glove moves a pathetically short distance up the ramp.
- You have to be as far away from Joch as you can go for the stalagmite jump to travel far enough.

The other notable factor is that one of two animations happen every other frame (so if you pause buffer an overcharge that can change animations, frame rules probably mean that you'll always get the same attack from the same distance from the enemy).  Apart from pause buffering, this part is random as far as a human player is concerned.

I tried to put together a table documenting the behavior of the sword at specific ranges, but I'm finding it to be difficult since I don't know where the "center" of you and your target are for the distance measurements.  One thing that is certain, though, is that if you are far enough away from all enemies and NPCs, you will always do a horizontal swing with the sword.  (For example, in Gaia's Navel, once you've killed the Goblins, you will never miss the skull.)

The level 0 sword jump does not happen against an enemy in the normal state.

------

Here's another thing in HHS's post that I don't understand:
[quote]Weapon effects accumulate if another weapon enhancing spell is cast before the time runs out. [/quote]

After some testing, this definitely does not refer to the number of uses left for a saber spell, and it also does not refer to the damage boost of a saber spell, and it also does not refer to the status effects of a saber spell.  It could be that this is referring specifically to Mana Magic.

Yagamoth

2015-08-16 07:46:30 pm

INTJ

So... If I understand this correctly, this mana magic power should also affect 1p1c, right?

I mean, it seems relatively simply to gather a fair bit more Mana Magic along the way as the Boy oneshotting enemies, maybe even with Stone Saber?

Crow!

2015-08-16 10:42:58 pm

What's that gemma?

I would need to obtain, and level up, Luna magic for it to be useful in 1p1c. Also, I'm not sure whether Mana Sword Power is applied when the Infinite Sword Glitch is active... I'll have to check that.

Crow!

2015-08-23 11:52:25 pm

What's that gemma?

In Glitchless, leveling up Shade all the way to level 4 just for one fight totally sucks. Can we do better?

When you complete the Mana Tree cutscene, your sword's Orb level is set to 8. Having Watts forge the sword to 8 costs about as much money as you get from a single Purelands boss. So, upgrading the sword to level 8 costs some menuing and traveling time, but nothing more than that.

You can buff your attack with an elemental Saber (say, level 2 Moon Saber for +12.5% attack for the next 12 strikes), with Lunar Boost, and with automatic crits from Moon Energy. Dread Slime can have his defense reduced by 40 with Acid Storm - either from the Sprite, or reflected onto himself.

Trying it out, it seems that the damage range for a level 2 attack with all buffs and debuffs active is about 400-600. If you're controlling the character, level 1 attacks are probably the highest DPS, but for a 1 controller "glitchless" run, this seems unlikely.

There are a few issues. The party scattering is a concern, and the constant Acid Storms from the Dread Slime are problematic for an essentially no-armor run. A setup like in 1p1c, with the Boy set to charge via AI, is probably what would have to be done, so that you have access to both the Girl and the Sprite for menuing / damage canceling as necessary. The Girl should probably keep a Wall active, which does eat into the MP budget. The fight would be very technical, and would require a lot of practice (especially if you haven't learned the 1p1c Dread Slime fight).

By the way, AI characters get small stat bonuses and penalties based on their action grid position. The best one seems to be here, for +2 STR, +1 AGI:

oXoo
oooo
oooo
oooo

I'll do a stream sometime soon outlining one working strategy and coming up with an estimate of how much time could be saved if all goes well. I'll also give a tutorial for the 1c Fire Palace ramp climb, since I think I finally have a good setup for that.

In Glitchless, leveling up Shade all the way to level 4 just for one fight totally sucks.  Can we do better?

When you complete the Mana Tree cutscene, your sword's Orb level is set to 8.  Having Watts forge the sword to 8 costs about as much money as you get from a single Purelands boss.  So, upgrading the sword to level 8 costs some menuing and traveling time, but nothing more than that.

You can buff your attack with an elemental Saber (say, level 2 Moon Saber for +12.5% attack for the next 12 strikes), with Lunar Boost, and with automatic crits from Moon Energy.  Dread Slime can have his defense reduced by 40 with Acid Storm - either from the Sprite, or reflected onto himself.

Trying it out, it seems that the damage range for a level 2 attack with all buffs and debuffs active is about 400-600.  If you're controlling the character, level 1 attacks are probably the highest DPS, but for a 1 controller "glitchless" run, this seems unlikely.

There are a few issues.  The party scattering is a concern, and the constant Acid Storms from the Dread Slime are problematic for an essentially no-armor run.  A setup like in 1p1c, with the Boy set to charge via AI, is probably what would have to be done, so that you have access to both the Girl and the Sprite for menuing / damage canceling as necessary.  The Girl should probably keep a Wall active, which does eat into the MP budget.  The fight would be very technical, and would require a lot of practice (especially if you haven't learned the 1p1c Dread Slime fight).

By the way, AI characters get small stat bonuses and penalties based on their action grid position.  The best one seems to be here, for +2 STR, +1 AGI:

oXoo
oooo
oooo
oooo

I'll do a stream sometime soon outlining one working strategy and coming up with an estimate of how much time could be saved if all goes well.  I'll also give a tutorial for the 1c Fire Palace ramp climb, since I think I finally have a good setup for that.

Yagamoth

2015-08-24 12:51:16 am

INTJ

Would it be worth charging up to level 3 sword (assuming you have level 3 sword at that point)?

Crow!

Edit history:

Crow!: 2015-08-24 06:54:45 pm

Crow!: 2015-08-24 04:38:56 pm

Crow!: 2015-08-24 04:38:44 pm

Crow!: 2015-08-24 04:24:39 pm

2015-08-24 04:19:20 pm

What's that gemma?

I tried it out a bit, and setting the AI to charge to level 3 for my current Dread Slime setup seems to increase the variance of fight times, but I probably will be unable to tell if it improves or hurts the average without spending a lot more time running trials than I'm willing to.

Below are my notes for comparing using Luna + Dragon Buster vs using Shade vs the Dread Slime.
TL;DR: my current estimate is you'll save 1.5 minutes on average by using the new strategy.

Forging from Watts in Kakkara takes 20s if you were already visiting the Item Shop there anyway.

Leveling Shade to 4 took Stinger 3:45 (ideal is probably closer to 3m)

Stinger's Dread Slime with level 4 Shade took 2:16 (measured from entering room to explosion start)

So, if we can complete the Dread Slime fight in about 5 minutes with the Level 8 Sword + Luna strat, then we're breaking even. Any faster, and we're gaining time.

Practice times (AI set to level 2)
3:10
3:25
3:35
3:55
4:15

So, this strategy should save 1-2 minutes. About 30 MP gets spent by the Girl, which I think is reasonable (the Sprite would have spent more than that, and would have had to spend some time draining before going on with the fortress.)

lvl 3 tests:
3:00
3:05
4:25

For what it's worth, Lich has the same defense as Dread Slime, but I think Freeze is probably going to be the faster kill most of the time. There might be some value in timing level 3 attacks in between sets of Freeze casts, particularly if you're doing 1p2c or outright co-op.

EDIT: Here's a video demonstration of my setup for the fight.
http://www.twitch.tv/iicrowii/v/12642463

I tried it out a bit, and setting the AI to charge to level 3 for my current Dread Slime setup seems to increase the variance of fight times, but I probably will be unable to tell if it improves or hurts the average without spending a lot more time running trials than I'm willing to.

Below are my notes for comparing using Luna + Dragon Buster vs using Shade vs the Dread Slime.
TL;DR: my current estimate is you'll save 1.5 minutes on average by using the new strategy.

[spoiler]
Forging from Watts in Kakkara takes 20s if you were already visiting the Item Shop there anyway.

Leveling Shade to 4 took Stinger 3:45  (ideal is probably closer to 3m)

Stinger's Dread Slime with level 4 Shade took 2:16 (measured from entering room to explosion start)

So, if we can complete the Dread Slime fight in about 5 minutes with the Level 8 Sword + Luna strat, then we're breaking even.  Any faster, and we're gaining time.

Practice times (AI set to level 2)
3:10
3:25
3:35
3:55
4:15

So, this strategy should save 1-2 minutes.  About 30 MP gets spent by the Girl, which I think is reasonable (the Sprite would have spent more than that, and would have had to spend some time draining before going on with the fortress.)

lvl 3 tests:
3:00
3:05
4:25
[/spoiler]

For what it's worth, Lich has the same defense as Dread Slime, but I think Freeze is probably going to be the faster kill most of the time.  There might be some value in timing level 3 attacks in between sets of Freeze casts, particularly if you're doing 1p2c or outright co-op.

EDIT:  Here's a video demonstration of my setup for the fight.
http://www.twitch.tv/iicrowii/v/12642463

Yagamoth

Edit history:

Yagamoth: 2015-08-25 03:14:02 am

2015-08-25 01:13:17 am

INTJ

Excellent research, thank you very much crow Smiley

Edit: Have you considered using only normal slashes or level 1 attacks? I know that's probably slowed down by having to recast lunar energy more often and the healing animation delaying, but I saw the 2nd slash and that looked like decent enough damage for it to be potentially faster

Crow!

Edit history:

Crow!: 2015-08-25 08:52:01 am

Crow!: 2015-08-25 08:50:43 am

Crow!: 2015-08-25 08:00:23 am

2015-08-25 07:52:16 am

What's that gemma?

If you're using a human to control the Boy, I think that level 1 attacks are the way to go. However, for a 1 controller run, the AI does an extra normal slash as it starts charging, and adds a long delay between when the AI reaches his target charge level and when it actually attacks, and both those quirks incentivize using a slower but stronger attack.

Also, it would probably be better to move one or both of the non-timed buffs to just before the fight starts rather than doing all 3 buffs during the fight like I did in the video.

Edit: if my math is right, bouncing a Lunar Boost onto the Dread Slime will guarantee that your attacks will hit, rather than having a 24% chance to miss.

It'll also add a chance for the game to use the "low defense roll" formula, though the "low defense roll" will actually have exactly as much defense as normal (amusingly, in the 1% chance that a 99 evasion enemy gets a "low defense roll", its defense will be increased by 25%. The formulas apparently just assumed that having more than 80 evasion is complete nonsense, despite the fact that all their enemies which evade have exactly that).

If you're using a human to control the Boy, I think that level 1 attacks are the way to go.  However, for a 1 controller run, the AI does an extra normal slash as it starts charging, and adds a long delay between when the AI reaches his target charge level and when it actually attacks, and both those quirks incentivize using a slower but stronger attack.

Also, it would probably be better to move one or both of the non-timed buffs to just before the fight starts rather than doing all 3 buffs during the fight like I did in the video.

Edit: if my math is right, bouncing a Lunar Boost onto the Dread Slime will guarantee that your attacks will hit, rather than having a 24% chance to miss.

It'll also add a chance for the game to use the "low defense roll" formula, though the "low defense roll" will actually have exactly as much defense as normal (amusingly, in the 1% chance that a 99 evasion enemy gets a "low defense roll", its defense will be increased by 25%.  The formulas apparently just assumed that having more than 80 evasion is complete nonsense, despite the fact that all their enemies which evade have exactly that).

Crow!

Edit history:

Crow!: 2015-08-28 02:12:39 pm

Crow!: 2015-08-28 02:12:35 pm

Crow!: 2015-08-28 02:10:31 pm

Crow!: 2015-08-28 11:17:31 am

2015-08-28 11:15:51 am

What's that gemma?

I spent some time today poking around in the event system for the magic% glitch in the Japanese version of the game. With the help of Google Translate, I was able to figure out where I was in the event script with a few events I randomly found, and with some trial and error I found the event to jump to the The End screen. Thankfully, the event numbers seem to be the same between our versions.

Here's the instructions:

As usual:
Highlight a barrel so its name appears, then use it on the boy.
Talk to the Cannon Travel guy a short time after the status effect textbox disappears (the timing is frame perfect).

Upon talking to the cannon travel man and getting the blank dialogue and gold boxes:
Left, B
Left, B
Left, B
Left, B
Left x26, B

Notice that the second left press of the batch of 26 will be even laggier than the first (and will sound like you pressed it twice, even though you didn't). The rest of the lefts are lagless.

Deuceler

2015-08-28 11:19:06 am

SPEEDruns not SAFEruns

Nice stuff Crow! I've been wanting to mess with Magic% (because why not) and this gives me an excuse to fire up my JP cart. I'll need to go back and read from last year to figure out barrel timing etc.

Overfiendvip

2015-08-28 12:05:45 pm

never a frown

Thank you for taking the time to do all of this stuff, Crow. The one time I tried on snes9x with HHS's script, it didn't seem to lead to "The End" even though it appeared to try to (or something; don't really know), and for various reasons, figured I would try again later if I had to, so you saved me from what would have probably been a tremendous headache for me.

It appears that this series of inputs saves about a second over how I executed the English series in my SDA run. This is with using what I would refer to as normal speed thumb mashing on the last series of "Left" inputs. It can certainly save more time if taken more riskily, since you can basically mash as fast as you want through the last series of "Left", but the faster you mash, the higher the chance of miscounting. I expect Japanese to save about a dozen seconds over English just off version differences now (~2 from this, ~10 from everything else such as text and cutscene events), but will have a better idea whenever I return to it, since I primarily analyzed events and not all the little details inbetween.

Here's an example of the above, this one saving 2-3 seconds over my SDA run's inputs, and could still potentially be slightly faster (3rd B press and final B press were delayed, and mashing can always be faster assuming it won't affect counting) - http://puu.sh/jRMGq.mp4

Thank you for taking the time to do all of this stuff, Crow. The one time I tried on snes9x with HHS's script, it didn't seem to lead to "The End" even though it appeared to try to (or something; don't really know), and for various reasons, figured I would try again later if I had to, so you saved me from what would have probably been a tremendous headache for me.

It appears that this series of inputs saves about a second over how I executed the English series in my SDA run. This is with using what I would refer to as normal speed thumb mashing on the last series of "Left" inputs. It can certainly save more time if taken more riskily, since you can basically mash as fast as you want through the last series of "Left", but the faster you mash, the higher the chance of miscounting. I expect Japanese to save about a dozen seconds over English just off version differences now (~2 from this, ~10 from everything else such as text and cutscene events), but will have a better idea whenever I return to it, since I primarily analyzed events and not all the little details inbetween.

Here's an example of the above, this one saving 2-3 seconds over my SDA run's inputs, and could still potentially be slightly faster (3rd B press and final B press were delayed, and mashing can always be faster assuming it won't affect counting) - http://puu.sh/jRMGq.mp4

Crow!

Edit history:

Crow!: 2015-09-02 01:16:07 pm

2015-09-02 01:15:37 pm

What's that gemma?

Something to investigate for the multiple controllers run:

Along the same lines as the co-op run's version of the Northtown guard skip, you might be able to avoid talking to the guy tending to the stove in the Scorpion Army's Salamando-heated town. Player 3 would talk to a townsperson, then the multitap would be turned off, and player 1 would move in position to talk to the stove. So long as Player 1 does not move, I don't think the Scorpion Army guy's text will start. This would save some time watching him walk off screen.

This would probably make the cutscene play out weridly, but since the only important thing about the visit is getting Salamando magic, that should be fine.

Unfortunately, testing this is rather difficult, so I haven't done it yet.

StingerPA

Edit history:

StingerPA: 2015-09-02 04:57:03 pm

2015-09-02 04:44:44 pm

Quote from Crow!:

Something to investigate for the multiple controllers run:

Along the same lines as the co-op run's version of the Northtown guard skip, you might be able to avoid talking to the guy tending to the stove in the Scorpion Army's Salamando-heated town. Player 3 would talk to a townsperson, then the multitap would be turned off, and player 1 would move in position to talk to the stove. So long as Player 1 does not move, I don't think the Scorpion Army guy's text will start. This would save some time watching him walk off screen.

This would probably make the cutscene play out weridly, but since the only important thing about the visit is getting Salamando magic, that should be fine.

Unfortunately, testing this is rather difficult, so I haven't done it yet.

Nice idea. I just tested it, and it works. You don't even need a multitap to do this since there's an NPC right there when you walk in, just have the 2nd controller talk to somebody, move up to the platform while the text box is up, and then talk to the stove. It is still possible to trigger the scorpion army guy from up there, though, so be careful about walking into him.

You were very correct about the cutscene being wonky. It turns out the scorpion army really likes the town after it gets cold again, so they never leave.

[quote="Crow!"]Something to investigate for the multiple controllers run:

Along the same lines as the co-op run's version of the Northtown guard skip, you might be able to avoid talking to the guy tending to the stove in the Scorpion Army's Salamando-heated town.  Player 3 would talk to a townsperson, then the multitap would be turned off, and player 1 would move in position to talk to the stove.  So long as Player 1 does not move, I don't think the Scorpion Army guy's text will start.  This would save some time watching him walk off screen.

This would probably make the cutscene play out weridly, but since the only important thing about the visit is getting Salamando magic, that should be fine.

Unfortunately, testing this is rather difficult, so I haven't done it yet.[/quote]

Nice idea.  I just tested it, and it works.  You don't even need a multitap to do this since there's an NPC right there when you walk in, just have the 2nd controller talk to somebody, move up to the platform while the text box is up, and then talk to the stove.  It is still possible to trigger the scorpion army guy from up there, though, so be careful about walking into him.

You were very correct about the cutscene being wonky. It turns out the scorpion army really likes the town after it gets cold again, so they never leave.

Crow!

2015-09-02 07:00:45 pm

What's that gemma?

Another thought: can you remote-talk the stove? Salamando is slot 1 and the stove is slot 2, so since Salamando doesn't exist yet, it's probably the room's remote-talk target.

Yagamoth

2015-09-03 05:40:32 am

INTJ

Tried remote talking with Autofire on various attempts - doesn't seem to trigger unfortunately. But good find ^^

Crow!

Edit history:

Crow!: 2015-09-11 12:09:16 pm

Crow!: 2015-09-11 12:08:41 pm

2015-09-11 11:53:01 am

What's that gemma?

Yaga had some trouble with the Sprite sometimes dealing 1 less damage than he should for no apparent reason in the Haunted Forest, which was preventing him from knocking Chobin Hoods to the ground on a low damage roll. I spent some time hunting down answers, and I had a lot of false leads, but I eventually found the issue: the action grid's AI status bonuses are sometimes not removed.

Depending on where a character's AI grid is located, when the character is AI controlled, they get some small stat bonuses and penalties. However, taking control of an AI via the Select button does not immediately remove the action grid's stat changes. The default, maximally passive AI position takes a -1 strength penalty, so unless you do something that makes the game update your stats, you will for most of the game have 1 less attack power than you should.

Alternatively, if you set the AI to a grid position with a bonus to your attack, then you will retain that bonus to attack. The left column of the AI grid has +1 strength. (As an aside, HHS's claims regarding what stat changes are where on the grid are completely wrong.) This bonus - or at least, understanding how to avoid the penalty - is probably useful in "Glitchless," and might sometimes be handy elsewhere.

Some actions which update a character's stats include:
Swapping off of a character via Select (To test: does abandoning a character via the multitap do it?)
Looking at the character's weapon menu
Leveling up

(amusingly, looking at the Status screen does NOT update the character's stats)

Yaga had some trouble with the Sprite sometimes dealing 1 less damage than he should for no apparent reason in the Haunted Forest, which was preventing him from knocking Chobin Hoods to the ground on a low damage roll. I spent some time hunting down answers, and I had a lot of false leads, but I eventually found the issue: the action grid's AI status bonuses are sometimes not removed.

Depending on where a character's AI grid is located, when the character is AI controlled, they get some small stat bonuses and penalties.  However, [b]taking control of an AI via the Select button does not immediately remove the action grid's stat changes[/b].  The default, maximally passive AI position takes a -1 strength penalty, so unless you do something that makes the game update your stats, you will for most of the game have 1 less attack power than you should.

Alternatively, if you set the AI to a grid position with a bonus to your attack, then you will retain that bonus to attack.  The left column of the AI grid has +1 strength. (As an aside, HHS's claims regarding what stat changes are where on the grid are completely wrong.)  This bonus - or at least, understanding how to avoid the penalty - is probably useful in "Glitchless," and might sometimes be handy elsewhere.

Some actions which update a character's stats include:
Swapping off of a character via Select  (To test: does abandoning a character via the multitap do it?)
Looking at the character's weapon menu
Leveling up

(amusingly, looking at the Status screen does NOT update the character's stats)

Yagamoth

2015-09-11 06:01:10 pm

INTJ

Does that mean we could potentially gain extra stats by using the grid in certain ways? oô

Crow!

2015-09-11 10:57:55 pm

What's that gemma?

Yes. I'm not sure when it would be worth the time expenditure, but here's a simple example:

Once you've picked up your second character, set both characters' action grids to one of the following places:
oooo
Xooo
Xooo
oooo

Then, swap to your ally. Now both characters have +2 attack instead of -1 attack, for a difference of 3 damage per attack (sometimes less for low damage rolls, and more for charged attacks).