Program Moveset bin format structure & a bit more

AlexXsWx · Apr 10, 2021

EDIT 18Jul21: for DOA6 there's this post: https://www.freestepdodge.com/threads/doa6-moveset-bin-format-partial-how-to-mod-them.8918/

I've been digging through moveset bin format for quite some time now, and even though there are still unknowns I'm now at a good point to share my findings and tools with everyone, along with some gifs I made along the way

Here's an example of what moveset bins contain:

Legacy information for hit reaction for BT NH / BT crouch NH:
https://gfycat.com/blondspeedyanemone

Frame advantage on block / nh (and even ch / hch as long as it doesn't stun or GB), stored as an actual advantage number (so the game does the math to make it match):
https://gfycat.com/sophisticatedwaterloggedafricanparadiseflycatcher
https://gfycat.com/severeplusharcherfish

Overall information which move ids to apply in which situations (stun / juggle / ground hit etc), here's some modding of that:
https://gfycat.com/bareacclaimedcat

Some unused moves and sometimes even inputs for them:
https://gfycat.com/acclaimedsandyleafwing
https://gfycat.com/drearycompassionatearabianhorse

Full AI configuration - pre-made combos and probability weights
e.g. here I constrained COM lvl 8 to just one combo (KK46K, PP4KKK) after I recorded the replay:
https://gfycat.com/pleasantfrequentasianporcupine
and here I changed probability weights to always start offense with an OH (which it doesn't have), and always defend with crushing moves:
https://gfycat.com/acclaimedmelodicindianglassfish

General information about character like juggle momentum multiplier:
(this one nicely shows how juggle system works)
https://gfycat.com/impressivefairafricanwildcat
https://gfycat.com/jealousanykakarikis

... or juggle starting height multiplier (sort of "character height"):
https://gfycat.com/illfatedspeedycardinal

Code:

// momentum mult. | height mult. | // char name
1.0200 0.9880 // kasumi
1.1025 0.9880 // alpha
1.0000 1.0000 // brad wong
1.0200 1.0000 // mila
1.0000 0.9880 // helena
1.0200 1.0000 // kokoro
0.9150 1.0890 // bayman
1.0000 1.0000 // christie
1.0000 1.0020 // rig
1.0000 1.0020 // jann lee
1.0000 1.0020 // zack
1.0200 0.9880 // ayane
1.0000 1.0020 // hayabusa
1.0000 1.0000 // nyotengu
1.0200 1.0000 // eliot
1.0000 1.0020 // tina
1.0000 1.0000 // lisa
1.0000 1.0020 // akira
1.0200 1.0000 // sarah
0.9150 1.0890 // bass
1.0200 0.9620 // leifang
1.0200 1.0000 // hitomi
1.0200 0.9880 // phase 4
1.0000 1.0020 // hayate
0.9150 1.0890 // raidou
1.0200 0.9880 // marie rose
1.0200 1.0000 // momiji
1.0000 0.9620 // gen fu
1.0200 0.9620 // pai
1.0000 1.0020 // jacky
1.0000 1.0000 // ein
0.9150 1.0890 // rachel
0.9150 1.0890 // leon
1.0200 0.9880 // honoka
1.0200 1.0000 // mai
1.0000 1.0020 // naotora

A chunk for collisions, I'm talking hitboxes / hurtboxes / pushboxes:
https://gfycat.com/slowcookedbarebirdbat

Event timelines, among many other stuff, contain information about followup intervals:
https://gfycat.com/athleticexcitableamazonparrot

And I briefly touched format of actual animations (not part of moveset bin, but it's still used for it to tell how long an animation is, or to define root motion):
https://gfycat.com/grandminiaturecatfish

Anyways, here you can find folder with movesets, few motions, and the code I used to parse & document the files:
EDIT (24apr2021): please make sure to have verbose logs on (see this message for more details)
https://www.mediafire.com/folder/43n9rlglsxufo/doa5lr_moveset

Be advised that it is mainly for other programmes who can read code, and it's far from user friendly or complete, so don't expect to be able to use it out of the box.

Keep in mind I mainly tested on Rig's bin / animation, and while I was briefly checking other characters from time to time, a lot of validation is currently failing (and there are some overlaps in AI combos section, apparently I got some terminators wrong).
I've also seen that structure is seriously off for Marie, Honoka and one other char (can't remember right now), so they need to be inspected more closely.

KasumiLover · Apr 10, 2021

AlexXsWx said:
I've been digging through moveset bin format for quite some time now, and even though there are still unknowns I'm now at a good point to share my findings and tools with everyone, along with some gifs I made along the way

Here's an example of what moveset bins contain:

Legacy information for hit reaction for BT NH / BT crouch NH:
https://gfycat.com/blondspeedyanemone

Frame advantage on block / nh (and even ch / hch as long as it doesn't stun or GB), stored as an actual advantage number (so the game does the math to make it match):
https://gfycat.com/sophisticatedwaterloggedafricanparadiseflycatcher
https://gfycat.com/severeplusharcherfish

Overall information which move ids to apply in which situations (stun / juggle / ground hit etc), here's some modding of that:
https://gfycat.com/bareacclaimedcat

Some unused moves and sometimes even inputs for them:
https://gfycat.com/acclaimedsandyleafwing
https://gfycat.com/drearycompassionatearabianhorse

Full AI configuration - pre-made combos and probability weights
e.g. here I constrained COM lvl 8 to just one combo (KK46K, PP4KKK) after I recorded the replay:
https://gfycat.com/pleasantfrequentasianporcupine
and here I changed probability weights to always start offense with an OH (which it doesn't have), and always defend with crushing moves:
https://gfycat.com/acclaimedmelodicindianglassfish

General information about character like juggle momentum multiplier:
(this one nicely shows how juggle system works)
https://gfycat.com/impressivefairafricanwildcat
https://gfycat.com/jealousanykakarikis

... or juggle starting height multiplier (sort of "character height"):
https://gfycat.com/illfatedspeedycardinal

Code:

// momentum mult. | height mult. | // char name 1.0200 0.9880 // kasumi 1.1025 0.9880 // alpha 1.0000 1.0000 // brad wong 1.0200 1.0000 // mila 1.0000 0.9880 // helena 1.0200 1.0000 // kokoro 0.9150 1.0890 // bayman 1.0000 1.0000 // christie 1.0000 1.0020 // rig 1.0000 1.0020 // jann lee 1.0000 1.0020 // zack 1.0200 0.9880 // ayane 1.0000 1.0020 // hayabusa 1.0000 1.0000 // nyotengu 1.0200 1.0000 // eliot 1.0000 1.0020 // tina 1.0000 1.0000 // lisa 1.0000 1.0020 // akira 1.0200 1.0000 // sarah 0.9150 1.0890 // bass 1.0200 0.9620 // leifang 1.0200 1.0000 // hitomi 1.0200 0.9880 // phase 4 1.0000 1.0020 // hayate 0.9150 1.0890 // raidou 1.0200 0.9880 // marie rose 1.0200 1.0000 // momiji 1.0000 0.9620 // gen fu 1.0200 0.9620 // pai 1.0000 1.0020 // jacky 1.0000 1.0000 // ein 0.9150 1.0890 // rachel 0.9150 1.0890 // leon 1.0200 0.9880 // honoka 1.0200 1.0000 // mai 1.0000 1.0020 // naotora

A chunk for collisions, I'm talking hitboxes / hurtboxes / pushboxes:
https://gfycat.com/slowcookedbarebirdbat

Event timelines, among many other stuff, contain information about followup intervals:
https://gfycat.com/athleticexcitableamazonparrot

And I briefly touched format of actual animations (not part of moveset bin, but it's still used for it to tell how long an animation is, or to define root motion):
https://gfycat.com/grandminiaturecatfish

Anyways, here you can find folder with movesets, few motions, and the code I used to parse & document the files:
https://www.mediafire.com/folder/43n9rlglsxufo/doa5lr_moveset

Be advised that it is mainly for other programmes who can read code, and it's far from user friendly or complete, so don't expect to be able to use it out of the box.

Keep in mind I mainly tested on Rig's bin / animation, and while I was briefly checking other characters from time to time, a lot of validation is currently failing (and there are some overlaps in AI combos section, apparently I got some terminators wrong).
I've also seen that structure is seriously off for Marie, Honoka and one other char (can't remember right now), so they need to be inspected more closely.

Do you have any cut moves for Naotora and Kasumi and Phase 4 that didn't make the cut and are unused??

AlexXsWx · Apr 10, 2021

KasumiLover said:
Do you have any cut moves for Naotora and Kasumi and Phase 4 that didn't make the cut and are unused??

I'm sorry but I will not go over every character to check this. You can try to find this information using the files I provided though

DestructionBomb · Apr 10, 2021

I really want to dig into this. Will have to redownload the games at some point.

DestructionBomb · Apr 10, 2021

Posted it on one of the FB groups. Curious if there'll be interesting talks about it.

XxNoriakixX · Apr 10, 2021

If you can somehow find data to add lobbies to the pc version, then I'll patreon you for a year lol

WAZAAAAA · Apr 10, 2021

hovering characters on the character select screen while monitoring stuff that gets loaded using ProcMon, lool good shit

rjqnraos19 · Apr 11, 2021

So can we also improve the CPU AI? I'm so exciting now.

AlexXsWx · Apr 11, 2021

rjqnraos19 said:
So can we also improve the CPU AI? I'm so exciting now.

Yes, to an extent. I didn't dive too deep into AI stuff

DestructionBomb · Apr 11, 2021

AlexXsWx said:
Yes, to an extent. I didn't dive too deep into AI stuff

That's pretty dope actually, didn't expect that one. If there's a way, i'd imagine CPU boss battles to be an actual thing than actual cheese lol. Particularly Alpha-152, Raidou, Phase-4, or someone might be able to once again replicate Akira into an incredibly hard boss but in DOA's style. iirc, every single character I mentioned have very poor AI overall.

AlexXsWx · Apr 24, 2021

Just realized something - I print some important logs in "debug" mode, which do not show up by default; this can make following the documentation impossible (either because logs won't show up or because context menu won't be available). So please change displayed log levels to "verbose" to see that logs (that is done in browser developer tools, "Default levels" -> "verbose" for Chrome)

rjqnraos19 · Jul 16, 2021

Any news?

AlexXsWx · Jul 16, 2021

rjqnraos19 said:
Any news?

I'm wrapping up documentation for DOA6, but nothing new for DOA5LR

AlexXsWx · Jul 18, 2021

Here it is for DOA6: https://www.freestepdodge.com/threads/doa6-moveset-bin-format-partial-how-to-mod-them.8918/

rjqnraos19 · Aug 31, 2021

I'm sorry for bothering you. I was trying to edit CPU AI but I can't do anything because I don't know about programming or coding at all.
I opened the .bin files on chrome and I just gave up.
Can you fix the CPU AI like this?

Hayabusa: Disable or reducing 3•4kp - CPU spamming this over and over. It's boring and makes too easy to defeat CPU hayabusa.
Christie, Marie: Fix guard - their Guard is broken. They can't block anything. They do guard but not work at all.
Sarah, Jacky: Throw the offensive hold - They keep attack when player do OH and getting Hi counter.
Jacky: Adding some combo after Critical burst - After Critical Burst Jacky just do nothing.
Akira: Punishing throws - Akira do nothing when player do throws. just guarding forever.
Rachel: Fix low hold - She can hold low attack 100%.

I really want to fix these things...
Sorry for my bad english.

AlexXsWx · Aug 31, 2021

rjqnraos19 said:
I'm sorry for bothering you. I was trying to edit CPU AI but I can't do anything because I don't know about programming or coding at all.
I opened the .bin files on chrome and I just gave up.
Can you fix the CPU AI like this?

Hayabusa: Disable or reducing 3•4kp - CPU spamming this over and over. It's boring and makes too easy to defeat CPU hayabusa.
Christie, Marie: Fix guard - their Guard is broken. They can't block anything. They do guard but not work at all.
Sarah, Jacky: Throw the offensive hold - They keep attack when player do OH and getting Hi counter.
Jacky: Adding some combo after Critical burst - After Critical Burst Jacky just do nothing.
Akira: Punishing throws - Akira do nothing when player do throws. just guarding forever.
Rachel: Fix low hold - She can hold low attack 100%.

I really want to fix these things...
Sorry for my bad english.

Uhhhh that's a project and a half, and is not really of interest to me, so sorry - not gonna do that.

rjqnraos19 · Aug 31, 2021

AlexXsWx said:
Uhhhh that's a project and a half, and is not really of interest to me, so sorry - not gonna do that.

Okay, Thanks for reply!

rjqnraos19 · Mar 14, 2023

hello again, I finally figured out editing AI moves and combos with hex workshop.
I edited ai Akira's CB combo(4p+k . 66p . 1p . 66p+k) to 1f h+k and 466p+k. now ai akira better than before

but don't know where is the AI's counter like your momiji video.
as I said before, akira never block low and just do nothing when opponent trying to throw him.
also seems akira doesn't hold that much unlike other characters. that's why he block over and over.
can you teach me where is the addresses for ai's counter and editing that?

AlexXsWx · Mar 14, 2023

rjqnraos19 said:
hello again, I finally figured out editing AI moves and combos with hex workshop.
I edited ai Akira's CB combo(4p+k . 66p . 1p . 66p+k) to 1f h+k and 466p+k. now ai akira better than before
but don't know where is the AI's counter like your momiji video.
as I said before, akira never block low and just do nothing when opponent trying to throw him.
also seems akira doesn't hold that much unlike other characters. that's why he block over and over.
can you teach me where is the addresses for ai's counter and editing that?

Well done!
I believe for the gif I changed pAiData->pIdk21[i].aiOffenseTypeWeights to 0/0/0/100 and pAiData->pIdk21[i].aiDefenseTypeWeights to 0/100/0/0.
These weights somehow define which list of "weighted combos" to use (from pAiData->pIdk19[i]->... and pAiData->pIdk20[i]->...).
I don't have an exact understanding of how they are connected, but you can find my notes on it for doa6 (which likely uses the same structure) here - the list of 48 entries after function createBtCombo(...) { ... }

rjqnraos19 · Mar 14, 2023

AlexXsWx said:
Well done!
I believe for the gif I changed pAiData->pIdk21[i].aiOffenseTypeWeights to 0/0/0/100 and pAiData->pIdk21[i].aiDefenseTypeWeights to 0/100/0/0.
These weights somehow define which list of "weighted combos" to use (from pAiData->pIdk19[i]->... and pAiData->pIdk20[i]->...).
I don't have an exact understanding of how they are connected, but you can find my notes on it for doa6 (which likely uses the same structure) here - the list of 48 entries after function createBtCombo(...) { ... }

thank you for the fast reply!
that looks hard. seems I must learn java script.
I just edited the bin file with hex editor

Program Moveset bin format structure & a bit more

Active Member

xX_APO_Prince_Xx

Active Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Active Member

Active Member

Well-Known Member

Active Member

Active Member

Active Member

Active Member

Active Member

Active Member

Active Member

Active Member

Active Member

Active Member