Topic by Doc_Skellington

Posted: 9/11/2014 6:23 PM

Quote

Doc_Skellington

He/Him

Active player (252)

Joined: 8/12/2013
Posts: 185
Location: Belgium

So, here are the damage of all the Physical Keyblades (do I have to try with the non-physical Keyblades ?)

             |    Damage Dealt       |  Frames of Delay      |Frames of Delay Aerial | 
Attack       |Strike :Thrust :Finish |Strike :Thrust :Finish |Strike :Thrust :Finish |
-------------+-------:-------:-------+-------:-------:-------+-------:-------:-------+ 
Kingdom Key  |   10  :  10   :  15   | 22/17 : 22/17 : 70/61 | 22/17 : 22/17 : 68/59 |
Three Wishes |   20  :   10  :  30   | 19/17 : 19/17 : 66/60 |       :       :       |
Pumpkinhead  |   20  :  20   :  15   |       :       :       |       :       :       |
Wishing Star |   15  :  15   :  15   |       :       :       |       :       :       |
Lady Luck    |   20  :   20  :  10   |       :       :       |       :       :       |
Olympia      |   20  :  10   :  30   |       :       :       |       :       :       |
Fairy Harp   |   20  :  20   :  20   |       :       :       |       :       :       |
Crabclaw     |   15  :  15   :  35   |       :       :       |       :       :       |
Divine Rose  |   35  :  10   :  20   |       :       :       |       :       :       |
Oathkeeper   |   25  :  40   :  35   |       :       :       |       :       :       |
Oblivion     |   ?   :   ?   :   ?   |       :       :       |       :       :       |
SpellBinder  |   ?   :   ?   :   ?   |       :       :       |       :       :       |
Metal Chocobo|   ?   :   ?   :   ?   |       :       :       |       :       :       |
Lionheart    |   ?   :   ?   :   ?   |       :       :       |       :       :       |
Ultima       |   40  :   40  :  40   |       :       :       |       :       :       |

I'm not going to try with Diamond Dust and One-Winged Angel, except if you want this information too. Well, in fact Cloud is one of the easier boss (or mid-boss) to do the tests : he is on the ground and Sora can walk without any trouble. Most of the time, it's +/- 50% (weakness and resistance as well). Ok, so the cards you get depend on the floor, but FatRatKnight is better than me in finding the formulas. Hmm, that explains a few things, FatRatKnight.

New PCSX2-rr

Posted: 9/12/2014 12:19 AM

Quote

Ghostwheel

Joined: 3/22/2013
Posts: 60

The only other Keyblade I care about is Oblivion, because I've been using that on Marluxia 1+2 as a backup for Oathkeeper quite a bit. Is he even weak to it? I never really understood that. This tells me you don't need fairy harp at all. That's huge. The only thing you get out of it over Lady luck is swing speed and card break recovery, you don't actually get more damage. So you can get your crabclaws and divine roses here, and get 3-5 Oathkeepers from castle oblivion (I think between Marluxia 1+2 opening 1 pack will save time). Shocking and awesome revalations: 3 wishes, pumpkinhead, and olympia all have the same strike damage as Lady luck, the only real difference is the thrust! I'm going to have to work through some things. I'm still not sure on this: You said Blizzard and blizzard raid do 50(75 with enemy card) damage per hit, but that you tested this on heartless. Does it do the same damage to Cloud? Or does it do half damage? He's a boss without special weaknesses or resistances so I want to know how to compare magic and attacks. Anyway, thanks again guys. I'm curious to see what comes out of the card drop investigation because thats really weird. How are you geting these values by the way, Tounet? Are you reading cloud's HP and checking how much damage he takes? Or is there some other way to determine damage?

Posted: 9/13/2014 6:00 PM

Quote

Doc_Skellington

He/Him

Active player (252)

Joined: 8/12/2013
Posts: 185
Location: Belgium

The "problem" of Oblivion is that it's a non-physical Keyblade (so I'll have to find a Heartless) and most of the bosses have resistance. I'll do the tests as soon as possible. Ok, I'll create a Moogle Room in Atlantica, then. Actually, Cloud has a resistance of more than 50%... It deals about 20 damage. I'm sure the values must be stored somewhere in the ROM but I haven't found them yet. So, I use the RAM Watch to see the Cloud's HPs.

New PCSX2-rr

Posted: 9/14/2014 6:09 PM

Quote

Ghostwheel

Joined: 3/22/2013
Posts: 60

What is the damage Oblivion does to either form of marluxia? That's the only boss I can conceivably use it on. Does the resisstance Cloud has for blizzard raid stay true for other bosses? If so, is it 20 damage instead of 50? Or a 50% cut that got rounded somehow. Maybe its 1/3? The whole point of this exercise is to compare damage of certain attacks against bosses, not heartless.

Posted: 9/15/2014 1:29 AM

Quote

FatRatKnight

Editor, Skilled player (1173)

Joined: 9/27/2008
Posts: 1085

Alright, I have been figuring out the data structure of the internals to figure out how to have my bot successfully detect which card shows up. Which was proving difficult, as it uses entirely separate objects for "PrizeCard" (sub-object) and "Heartless card" (main-object). Now that I have the way to detect all combat objects, I should be able to produce a bot that edits the RNG and collect a bunch of data. While the bot may not tell exactly how the game determines the prize, it will know how to identify what it is we're getting and put that information somewhere. By no means is it perfect, but my usual methods of messing with the RNG isn't getting me anywhere, as I don't have any good clues as to which RNG roll out of dozens is being used to determine the card. As a side-effect of all this, I can probably get a script to identify other sorts of objects that the game may be using in combat. I don't know how useful that would be, but I can probably have a realtime script give some information for combat. As for what information I would display... That is another matter entirely.

Posted: 9/15/2014 3:08 AM

Quote

Ghostwheel

Joined: 3/22/2013
Posts: 60

That's great, FatRatKnight! Probably the most useful information for a TAS would be the "Heartless" card info. Predicting the value of a card for heartless type boses like Ursula, Darkside, and Marluxia 2 wouldm an that you could be safer not sleighting your attack combos. For me, I'm still interested in blue card frequency by floor, but I totally understand if this is getting to ridiculous to decipher

Posted: 9/15/2014 6:46 AM

Quote

FatRatKnight

Editor, Skilled player (1173)

Joined: 9/27/2008
Posts: 1085

Proof of concept script for locating combat objects in memory. Watch the TAS on VBA, and run the script. I also slipped in a piece of code that extracts some other information, but it's commented out. Any text editor should work in modifying the script to remove the comment markers. Of course, it's just a concept. I can now get object data easily. Reading it is another thing entirely, but working out the information contained within shouldn't be too hard, now that I know where precisely to look. They even have internal names conveniently left in for me to read and get some kind of clue what they relate to! In any case, the prize cards actually use different objects entirely between map cards and enemy cards. "PrizeCardInit" is used for map card prizes, and "Heartless card" is used for enemy card prizes. I did spot "PrizeCard" somewhere, but this is linked to from "PrizeCardInit". In any case, it's a mess of pointers in there. But I figured out the crazy system in there, I hope.

Posted: 9/15/2014 8:47 PM

Quote

FatRatKnight

Editor, Skilled player (1173)

Joined: 9/27/2008
Posts: 1085

Putting that concept to use. Needs refinement. I select out a moment in the TAS: The battle where it first gets the Red Nocturne. After letting it run for a while, it collected this data:

   0    1    4    6   14  To. Card IDs
-----------------------------
 171  214  193  215  413 1206  0
  41   57   56   65  123  342  1
  80   67   98   78  137  460  2
 254  270  277  312  463 1576  3
 231  236  231  238  372 1308  4
 166  197  191  176  372 1102  5
 276  269  290  251  443 1529  6
 151  168  143  152  309  923  7
  60   40   44   38  108  290  8
 107   87   85  104  187  570  9

1537 1605 1608 1629 2927 9306  Totals

459: 694 - Enemy card

No errors

I stopped it at 10000 trials in that battle. Apparently, there's a 7% chance of an enemy card in Teeming Darkness? I haven't taken too close a look at what Map Card each ID maps to, but I recall ID 13 is a green card, I believe Meeting Ground. I have a start in my search. When I have time, I'll continue this.

Posted: 9/15/2014 8:49 PM

Quote

solarplex

Active player (471)

Joined: 2/1/2014
Posts: 928

FatRatKnight wrote:

Putting that concept to use. Needs refinement. I select out a moment in the TAS: The battle where it first gets the Red Nocturne. After letting it run for a while, it collected this data:
   0    1    4    6   14  To. Card IDs
-----------------------------
 171  214  193  215  413 1206  0
  41   57   56   65  123  342  1
  80   67   98   78  137  460  2
 254  270  277  312  463 1576  3
 231  236  231  238  372 1308  4
 166  197  191  176  372 1102  5
 276  269  290  251  443 1529  6
 151  168  143  152  309  923  7
  60   40   44   38  108  290  8
 107   87   85  104  187  570  9

1537 1605 1608 1629 2927 9306  Totals

459: 694 - Enemy card

No errors
I stopped it at 10000 trials in that battle. Apparently, there's a 7% chance of an enemy card in Teeming Darkness? I haven't taken too close a look at what Map Card each ID maps to, but I recall ID 13 is a green card, I believe Meeting Ground. I have a start in my search. When I have time, I'll continue this.

This is interesting to watch you continue to work and rework these scripts, FatRatKnight.. I would watch you stream this anyday lol..

Posted: 9/16/2014 12:17 AM

Quote

Ghostwheel

Joined: 3/22/2013
Posts: 60

FatRatKnight wrote:

Putting that concept to use. Needs refinement. I select out a moment in the TAS: The battle where it first gets the Red Nocturne. After letting it run for a while, it collected this data:

   0    1    4    6   14  To. Card IDs
-----------------------------
 171  214  193  215  413 1206  0
  41   57   56   65  123  342  1
  80   67   98   78  137  460  2
 254  270  277  312  463 1576  3
 231  236  231  238  372 1308  4
 166  197  191  176  372 1102  5
 276  269  290  251  443 1529  6
 151  168  143  152  309  923  7
  60   40   44   38  108  290  8
 107   87   85  104  187  570  9

1537 1605 1608 1629 2927 9306  Totals

459: 694 - Enemy card

No errors

This is fantastic! Well it looks like the distribution in TT is split 2/3 1/3 for red cards and green cards. That's pretty easy to adjust for with 4 red card types. I always thought that enemy cards in Teeming darkness were 10% but it makes sense for it to be slightly lower. No reason to guess the closest nice, round number if it's really a little different. I am really interested to see what this looks like for other floors. Specifically all the first world set floors, Neverland, and Twilight Town. Then we start dipping into more card types and blue cards. I think the TAS has point in there where it fights an encounter at each of the first world set worlds

Posted: 9/16/2014 6:36 PM

Quote

Doc_Skellington

He/Him

Active player (252)

Joined: 8/12/2013
Posts: 185
Location: Belgium

Oblivion : 17/17/5 (the tests were made on either form of Marluxia). Blizzard/Fire deals 20 on each boss... But, 50/3 doesn't equal 20. So, maybe, it's a value that the developers chose ? Your findings are absolutely fantastic, FatRatKnight !

New PCSX2-rr

Posted: 9/17/2014 11:55 PM

Quote

FatRatKnight

Editor, Skilled player (1173)

Joined: 9/27/2008
Posts: 1085

Hah, my improved data collection script! Now with less running off the screen edge, better identification, and a brand new feature of using the Output Console to dump collected data for copy/paste. As for the card generation algorithm, I'm now finally getting a rough idea of what's going on, although it's all educated guesses. First, if you lack Map Cards, the game will skip generating an enemy card and go straight to producing a zero-value Map Card. Now, if you have any Map Cards, a single 100-sided roll is made to determine whether it should generate an enemy card. If so, make it. If not, start rolling for map cards. If it picks a Map Card ID that it doesn't like, then it will roll dice again until it comes up with an allowed Map Card ID. This "roll until success" method generally skews the chances of other cards showing up, and will also explain why it apparently rolls a random number of times based on my tweaks to the RNG itself. Besides, while running the bot, I've seen streaks of the exact same card showing up repeatedly, where previously I've not been seeing any of that card for many, many hundreds of RNG rolls. If rolls #0 ~ #6 did not generate an allowed card, then the game will ultimately end up picking RNG roll #7 no matter which of #0 ~ #6 you started it at. Which would explain these streaks In any case, if this theory is correct, then there's simply one list that identifies the card, and as you progress in the game, different stuff in that list gets allowed. Aside from skews caused by "roll until success", the relative probabilities of finding this one type of card versus finding this other type should remain roughly equal. So find the floor with the most blue cards and the least of other cards, and you're good! ... These are all guesses, though.

Posted: 9/19/2014 4:46 AM

Quote

Ghostwheel

Joined: 3/22/2013
Posts: 60

Well all of that makes sense I guess. Even tough it's a guess, I'll put stock in it. One algorithm to get cards for all the floors makes a lot of sense, as does rolling multiple times and rolling enemy cards independently (because that probability can change, maybe even based on the enemy type that drops the card). I also have experience with weird chains of card drops (getting an 8 meeting ground 3-5 times in a row on multiple occasions in Traverse) but I don't know if that's the same kind of thing at all. From what you're telling me, it sounds like as long as any two card types are on a floor, the relative frequency of those card types is the same. I guess it's time for a map card investigation then. Thanks a bunch, this really helps me out. If I can even get a little more consistent with the one most infuriating part of the run, that's a gigantic step. All this work has inspired a route where I can get even closer to the TAS by getting cards for berserk strategies more consistently. It'll be a lot riskier and the total damage I'll have with my deck is a lot closer to the minimum needed (which is why the values themselves were super helpful) I'm really excited to see you finish your work, Tounet, and I'm always around to support!

Posted: 9/24/2014 5:28 PM

Quote

Doc_Skellington

He/Him

Active player (252)

Joined: 8/12/2013
Posts: 185
Location: Belgium

Sorry for the delay : I had to do a big work for biology. Indeed, that makes sense (and explains a few things). I'm very glad you've found a new route ! So, I don't need to create a Moogle Room in Atlantica ? If I don't have to, let's go to Neverland ! And you can count on me, too ;).

New PCSX2-rr

Posted: 9/24/2014 10:22 PM

Quote

FatRatKnight

Editor, Skilled player (1173)

Joined: 9/27/2008
Posts: 1085

I finally used my data collection script to... Well, collect data. Here's what I've got: The ratios of map card values have remained fairly constant, whichever battle I picked from the existing TAS. The chance of values have been generally as follows: 0 - 10% :: 1 - _5% :: 2 - _5% :: 3 - 15% :: 4 - 15% 5 - 15% :: 6 - 15% :: 7 - 10% :: 8 - _5% :: 9 - _5% As well, here's some rough approximations for each area and how often I've been picking up each type of card: 1F Traverse Town 17% - _0:Teeming Darkness 17% - _1:Tranquil Darkness 17% - _4:Sleeping Darkness 17% - _6:Feeble Darkness 32% - 14:Meeting Ground 2F ~ 5F _7% _0:Teeming Darkness _7% _1:Tranquil Darkness _7% _4:Sleeping Darkness _7% _5:Moment's Reprieve _7% _6:Feeble Darkness 21% _8:Calm Bounty _7% 10:Moogle Room ??% 11:Sorcerous Waking ??% 12:Martial Waking ??% 13:Alchemic Waking _1% 14:Meeting Ground ??% 16:Strong Initiative ??% 17:Lasting Daze _7% 18:Stagnant Space _1% 19:Premium Room _1% 20:White Room _1% 21:Black Room The five ??% I listed are either roughly 20% or 1%, depending on what you used for the floor. The following gets 20%: Olympus: Martial Waking Wonderland: Strong Initiative Agrabah: Sorcerous Waking Monstro: Alchemic Waking Holloween Town: Lasting Daze Curiously, Lasting Daze was entirely absent in Monstro. Any enforced battles from the story will use the Traverse Town list. The TAS doesn't fight any normal battles on 7F. So, here's what I got on 8F: 12% _2:Guarded Trove 12% _3:Looming Darkness _1% _5:Moment's Reprieve 12% _7:Almighty Darkness 12% _9:False Bounty 12% 10:Moogle Room _1% 11:Sorcerous Waking _1% 12:Martial Waking _1% 13:Alchemic Waking _1% 14:Meeting Ground _1% 16:Strong Initiative _1% 17:Lasting Daze _1% 19:Premium Room 20% 20:White Room _1% 21:Black Room _9% 25:Key to Rewards For a TASer, this is helpful mainly to know what is possible. For RTA, this should give a pretty good idea what to expect. Keep in mind these percents are just estimates. The bot isn't exactly perfect, here.

Posted: 9/25/2014 4:35 PM

Quote

Doc_Skellington

He/Him

Active player (252)

Joined: 8/12/2013
Posts: 185
Location: Belgium

Thank you so much for your work. Your findings are always amazing ! Well, the data will be very useful to determinate the battles.

New PCSX2-rr

Posted: 9/25/2014 7:39 PM

Quote

Ghostwheel

Joined: 3/22/2013
Posts: 60

Tounet, the fact that you don't need Fairy Harp means that it IS better to make a room in Atlantica so you can use the better Keys to fight Ursula and maybe Riku but I don't know if he will be faster. I know Ursula will be though 37% vs 35% isn't too much of an advantage for the second world set, but that's a big boost in Moogle rooms. Very interesting indeed. This data is very helpful FatRatKnight. I don't think I'll need other floors on the second world set, but Twilight Town is important too. Whenever I or the TAS end up getting there that's the last floor that could potentially be useful to check. Thanks a bunch.

Posted: 10/5/2014 5:27 PM

Quote

Doc_Skellington

He/Him

Active player (252)

Joined: 8/12/2013
Posts: 185
Location: Belgium

Here is the new Atlantica (with a new deck) Link to video Thank you so much, FatRatKnight ! Your shop script is really, really, really useful and it was a lot easier to manipulate the Moogles. So, I took 3 Crabclaws and 3 Divine Roses (I couldn't take more because of the RNG). I decided to use the Lady Lucks for completing the combo (Divine Rose - Lady Luck - Crabclaw). I created 2 combos of LL - LL - TW. Were they good ideas ? For Ursula/Riku Replica, it's more or less the same thing (but with a different mind manipulation).

New PCSX2-rr

Posted: 10/6/2014 2:31 AM

Quote

Ghostwheel

Joined: 3/22/2013
Posts: 60

Great segment, it's perfect in every overworld aspect, at those probabilities any more than 3 of each is pushing it a little. The deck edit was fine, what I could see at least haha. The only thing I don't know is your enemy cards. For boss fights, now is a good time to bring up the Strike Thrust difference. Divine rose has 40 "strike" damage, and this usually refers to the first hit of the combo, except in 2 special cases, which actually come up in the video: -When you are on the ground, too far from an enemy to do an overhand strike, and you end up thrusting first instead. This is the combo you stock at 5:39 and the one after that. -When your enemy is in the air, and by using the first hit of your combo, you move from the ground to the air in an upward strike. This is again thrust first, and is functionally every combo you do on Ursula except for the first. In both, a good thing to look for (other than the damage being dealt) is Sora's vocal cues. What he says when he swings the keyblade directly corresponds to "Strike" and "Thrust". Another good way to remember is that the "Strike" is always a downward motion, whether on the ground or in the air. To remedy this, just scroll cards in the ursula fight (scroll - lady luck - scroll back - divine rose - crabclaw), because it's better to have them in the order you put them in the deck for other bosses. You can probably save a combo or two that way. For the Riku Fight, have you timed Overdrive vs. Attack Haste? Especially with proper combo ordering my hunch is that overdrive is faster. Regardless, even with attack haste, properly ordering your 5:39 combos will save you time as well.

Posted: 10/6/2014 9:56 AM

Quote

FatRatKnight

Editor, Skilled player (1173)

Joined: 9/27/2008
Posts: 1085

I did believe having a script conveniently decode the RNG in shops for you would go a long way into speeding up your search. Good to see confirmation that this belief proved true. I may get pretty quiet for a while, as I can't think of any particular help you'd need from lua scripts in battles. And considering I have no real knowledge of good strategies in this game, the only thing I can really offer is cheering for your progress. I will watch for any progress. I expect great things. I have enjoyed seeing your TAS so far.

Posted: 10/8/2014 3:56 PM

Quote

Doc_Skellington

He/Him

Active player (252)

Joined: 8/12/2013
Posts: 185
Location: Belgium

The enemy cards are in this order : Blue, Red, Blue, Soldier (it will be changed by Overdrive), Hades, Crescendo. Is it fine ? Oh, yes, I forgot this ! Actually, I can't hear Sora's voice when I'm TASing because of the frame advance. But thanks for the trick. So, I'll correct this error. Yes, you're right : Crabclaw and Divine Rose deals a lot of damage. So, with Overdrive, it'll be faster. I'll correct this, too. Don't worry, FatRatKnight, your help has been really precious and so useful for me. Thanks. I'll try not to disappoint you ;)

New PCSX2-rr

Posted: 10/16/2014 6:49 PM

Quote

Doc_Skellington

He/Him

Active player (252)

Joined: 8/12/2013
Posts: 185
Location: Belgium

Hello everyone, Here is the new version of Ursula (and so Riku Replica) : Link to video Changing the deck changes the frame when the door opens, so the RNG of the Ursula is different (but better for me, I must say). I follow your recommendations and used the cards in the good order. Riku Replica is also different (and a little more aggressive).

New PCSX2-rr

Posted: 10/18/2014 8:08 AM

Quote

Ghostwheel

Joined: 3/22/2013
Posts: 60

Those fights look really nice!

Posted: 10/18/2014 4:51 PM

Quote

InfamousKnight

He/Him

Joined: 9/12/2014
Posts: 536
Location: Waterford, MI

I remember when I played this game back when I was 11 years old , I didn't know what enemy cards did and never figured it out. So I beat the game without enemy cards. Back then, I couldn't read very well either. This run is looking quite nice. This game is very suitable for a TAS run. I can't believe people have done single segment runs of this game. This game is solely based on luck with speed running. Routing in this game is pretty fun as well. Will the final boss be easy like in every other RPG speedrun? Find out when that time comes! As for Riku mode, that shouldn't require that much luck. As your card deck is based on the world you're on. So skill is key in that area. People have done riku mode already. I don't think criticals exist in this game? The only real luck involved in riku mode is door cards and Jafar platform manipulation. And I suppose enemy behavior as well.

Posted: 10/18/2014 5:48 PM

Quote

Doc_Skellington

He/Him

Active player (252)

Joined: 8/12/2013
Posts: 185
Location: Belgium

I didn't use the Enemy Cards (not enough), too. Thank you. Well, Ghostwheel is really patient, I think. Even for a simple player, that game can be frustrating (isn't it, you damn Twilight Town ?) I think the last form of Marluxia will be fast, indeed (because of the luck manipulation). Actually, you're right : the enemies' behavior depends on the RNG, too. We can't use Blizzaga, Cloud or other tricks, so it'll be a little longer to farm the cards but not to kill the bosses. But I still don't know if I'll do Riku and I won't think about a route planning until I finish Sora. Thanks for your comment ;) !

New PCSX2-rr