iiinova — where your digital self keeps evolving

Pangaea @toptwentynotes.bsky.social · 27/09/2026

i believe on linux you can also go without the 4gb patch and force large address awareness through a wine setting directly

110

Pangaea @toptwentynotes.bsky.social · 18/09/2026

i'd just recommend that if you get either of these issues, you check and make sure they're not the problem, only a few people have reported these though, so hopefully it shouldn't be common

000

Pangaea @toptwentynotes.bsky.social · 18/09/2026

i've received bug reports of the Angel overlay not appearing, or the game crashing upon startup with Angel installed. these issues were due to MSI Afterburner conflicting with the overlay, and Windows 11 Smart App Control. unfortunately it seems like i can't fix the issues directly

110

Pangaea @toptwentynotes.bsky.social · 17/09/2026

be sure to update to this one as soon as possible, the old version is incompatible with version 2 for netplay! hopefully we won't need to do any huge migrations like this for the next couple versions, but it's worth telling everyone you play with to switch over for now

0105

Pangaea @toptwentynotes.bsky.social · 17/09/2026

new version of Angel is out, i fixed lots of the bugs that were reported in the first version. accessing online play should be quicker and more reliable for everyone, especially people with long-distance connections. there's a few new features too! github.com/super-contin...

github.com

Release Version 2 · super-continent/Angel

New Features: Training mode hotkeys for Save/Load state An automatic check for updates upon startup, with a notification that links to the download page An optional display in the top right corner...

14743

Pangaea @toptwentynotes.bsky.social · 12/09/2026

i'm not aware of any major crashes in the base game, but an LAA patch might fix those if it's a crash that was happening on character loading screens. hopefully it works!

001

Pangaea @toptwentynotes.bsky.social · 12/09/2026

if you want to support my work, i would greatly appreciate any donations, i still have features i want to add to the mod in updates ko-fi.com/pangaea__

ko-fi.com

Support Pangaea

037

Pangaea @toptwentynotes.bsky.social · 12/09/2026

finally releasing a project i've spent a lot of time working on. a mod for Arcana Heart 3 XTEND which adds several improvements, including rollback! special thanks to @armonte.meltyblood.org for doing much of the reversing work download and install guide here: github.com/super-contin...

github.com

GitHub - super-continent/Angel: An improvement mod for Arcana Heart 3 LOVEMAX SIXSTARS!!!!!! XTEND

An improvement mod for Arcana Heart 3 LOVEMAX SIXSTARS!!!!!! XTEND - super-continent/Angel

18431295

Pangaea @toptwentynotes.bsky.social · 02/08/2026

doesn't currently have a public release, i'm working to improve the experience a bit and get a proper ah3 modding project set up

081

Pangaea @toptwentynotes.bsky.social · 02/08/2026

created a hitbox viewer mod for Arcana Heart

25118

Pangaea @toptwentynotes.bsky.social · 12/06/2026

i'm honestly pretty shocked it got this good too considering what i was working with a few months ago, i expect that i can make it even better with a bit more training too

020

Pangaea @toptwentynotes.bsky.social · 12/06/2026

the model is playing as both characters here

110

Pangaea @toptwentynotes.bsky.social · 12/06/2026

as a comparison, this is what the model was doing a couple months ago bsky.app/profile/topt...

010

Pangaea @toptwentynotes.bsky.social · 12/06/2026

different type of training, it's used to make models that predict stuff better at actually accomplishing a specific task instead of just predicting stuff bsky.app/profile/topt...

010

Pangaea @toptwentynotes.bsky.social · 12/06/2026

for those who dont know what "RL" means: its reinforcement learning. the models i was showing before were trained to mimic player behavior as accurately as possible, this one has a second phase of training which tells it to estimate which decisions were "good" and focus on those actions

1102

Pangaea @toptwentynotes.bsky.social · 12/06/2026

Xrd model now uses RL, this isn't a finished training run, and no online RL has been applied yet as i'm still working on the mod setup for that, but i'm very happy with the progress! here's the model playing against itself, i think it's a massive step up from the last footage i posted

55015

Pangaea @toptwentynotes.bsky.social · 06/06/2026

this is correct! thanks, didn't think to check how RISC decayed, will have to update the instruction name to make it more obvious

010

Pangaea @toptwentynotes.bsky.social · 06/06/2026

things i've tried so far: - getting hit/hitting an opponent with varying RISC values - blocking/making an opponent block with varying RISC values none of them have resulted in any change the the meter gained by either player, i'm quite confused, because it varies by character, i assume they use it

100

Pangaea @toptwentynotes.bsky.social · 06/06/2026

working on figuring out the meter gain values for each character, i've created a table of each characters values, most of them are fairly obvious, just meter added and scaled by tension pulse somehow, but i can't figure out what "Guard Balance Tension" is supposed to be. any ideas?

A table of meter gain multipliers by character in Guilty Gear Xrd

210

Pangaea @toptwentynotes.bsky.social · 31/05/2026

i think the only thing i'm lacking now is RL to give the model the goal of winning, not just pure imitation of player habits. also more data, there is really no limit to the amount of replay data i would like to have, it's clearly still overfocusing on a few weird rounds (e.g. the backdash habit)

1110

Pangaea @toptwentynotes.bsky.social · 31/05/2026

switched to training on whole rounds instead of small context windows, i'm pretty shocked by how big the improvement is compared to what i had before. this also enables performance optimizations which make it way easier to run larger models in realtime. here's a Johnny vs Faust round as an example

3377

Pangaea @toptwentynotes.bsky.social · 27/05/2026

if he does things like 5K>5K often enough that could definitely impact the APM, it's counting everything here including when the opponent is in hit/blockstun

110

Pangaea @toptwentynotes.bsky.social · 27/05/2026

unsure if it meaningfully impacts chipp more but this metric doesnt capture a state transition from a state to itself, that would invalidate things like multiple jabs. might fix that and see if anything changes

110

Pangaea @toptwentynotes.bsky.social · 27/05/2026

i'd guess it's due to low sample size, i think if i ever get more replay folders from the less common characters it'll tighten up to a better average

040

Pangaea @toptwentynotes.bsky.social · 27/05/2026

no actually! thats why i'm so surprised by it bsky.app/profile/topt...

040

Pangaea @toptwentynotes.bsky.social · 27/05/2026

zato and jack-o players please send me your replay folders

082

Pangaea @toptwentynotes.bsky.social · 27/05/2026

in this case i actually expected zato to be unusually low, as eddies attacks wouldn't be counted because theyre a separate entity changing state, not zato. there must be some other thing boosting him unusually high, that's what happens when you have few samples of some characters

180

Pangaea @toptwentynotes.bsky.social · 27/05/2026

this isn't a perfect metric, it over-focuses on attacks, probably has some noise because a button press doesnt necessarily cause any given state change, and the average probably gets raised by stuff like mashing FD, i'm surprised at how clean it seems to be for characters with a lot of match samples

1140

Pangaea @toptwentynotes.bsky.social · 27/05/2026

trying to figure out how to measure the average APM of each character in the Xrd dataset. here i'm searching for changes in the player state that coincide with a button press. honestly not sure what's going on with zato

A plot showing the average button press-correlated state changes per minute for each Xrd character, this is supposed to roughly indicate the average APM of each character in the dataset

76918

Reposted by Pangaea

Pangaea @toptwentynotes.bsky.social · 18/04/2026

still only at ~30 replay folders. if anyone has time to send some over i would greatly appreciate it!

001

Reposted by Pangaea

Pangaea @toptwentynotes.bsky.social · 14/04/2026

if anyone has more Xrd replay data folders they're able to send over, i would appreciate it, going to try and continue work on the data science stuff i'm doing, but i want to get the most representative sample of Xrd play that i can

378

Pangaea @toptwentynotes.bsky.social · 25/04/2026

all the labels here are the state names used internally by Xrd, you might recognize some of them, but plenty are also named different from the labels in-game

030

Pangaea @toptwentynotes.bsky.social · 25/04/2026

the input prediction model takes the state that characters are in as input. these are placed in a vector space, starting out in random positions, but are slowly positioned so that states that are "related" end up closer together. here's a visualization of the random placement vs the sorted result

All states contained in the Xrd dataset, placed randomly throughout a vector space and projected to 2D with UMAP

All states contained in the Xrd dataset, placed throughout a vector space so that related states are closer together and projected to 2D with UMAP

1185

Pangaea @toptwentynotes.bsky.social · 18/04/2026

still only at ~30 replay folders. if anyone has time to send some over i would greatly appreciate it!

001

Pangaea @toptwentynotes.bsky.social · 17/04/2026

new BBScript version, this adds support for the latest Guilty Gear Strive update. it also stabilizes some changes kkots made, which fix formatting for scripts where ArcSys devs didn't close scopes correctly, as well as adding much more Xrd info github.com/super-contin...

github.com

Release v1.5.8: Support latest GGST version, stabilize other changes · super-continent/bbscript

This update adds support for the latest version of GGST, as well as stabilizing changes from 1.5.7

0105

Pangaea @toptwentynotes.bsky.social · 14/04/2026

378

Pangaea @toptwentynotes.bsky.social · 08/04/2026

something went very wrong this time

261

Pangaea @toptwentynotes.bsky.social · 05/04/2026

this doesnt seem to apply to directions though, it can charge for flashkick as leo perfectly fine

010

Pangaea @toptwentynotes.bsky.social · 05/04/2026

the model seems to have trouble holding buttons down, i think it might be from the temperature sampling but i don't know if i can fix that

120

Pangaea @toptwentynotes.bsky.social · 05/04/2026

new model, this time it's trained to predict further ahead in the future, hoping that if i tweak this right it will learn to plan ahead better and maybe execute real combos, there's already some improvement to sols short strings but it doesn't seem to be enough yet

274

Pangaea @toptwentynotes.bsky.social · 05/04/2026

early version of a new model, seems to be a bit better at short combo strings but still cant do much more than that, also still has some pretty silly decision making in neutral. its starting to feel less like a bot and more like a very weird human player which is cool though

174

Pangaea @toptwentynotes.bsky.social · 04/04/2026

alright, now that the button issue is fixed, here's a newer model, both players are bots. it's currently quite random but i'm trying to improve that

4269

Pangaea @toptwentynotes.bsky.social · 04/04/2026

this is a silly problem to have missed but now that i've fixed it i can start improving the model again

020

Pangaea @toptwentynotes.bsky.social · 04/04/2026

after trying to debug the 5D spam assuming it was an issue with the model itself, i've realized it's because i forgot to change the Xrd controller binds back to the default after playing a bit, meaning the program automating the inputs was pressing the incorrect buttons the whole time

2274

Pangaea @toptwentynotes.bsky.social · 04/04/2026

been working on player input prediction using the replay dataset. both players here are bots. still a rough result, but i've made a lot of progress since starting

3284

Pangaea @toptwentynotes.bsky.social · 01/04/2026

see my recent posts for dataset-related stuff, i recently got a few more replay folders that i'll be adding soon bsky.app/profile/topt...

010

Pangaea @toptwentynotes.bsky.social · 01/04/2026

at the moment i've got a lot of other projects that i'm working on, i think it would be possible to implement something that gathers data about games that are being played live, but it would need infrastructure i currently dont have. i am currently working on gathering a replay dataset though

110

Reposted by Pangaea

Pangaea @toptwentynotes.bsky.social · 28/03/2026

if you would like to help build this dataset, i would greatly appreciate anyone sending me more replay folders, especially the ones for any character with <100 appearances! just compress the "REPLAY" folder inside the "SAVE" folder in your Xrd install and send it over

The total number of times a character appears in the current Xrd dataset:
Faust: 793
Ky: 642
Raven: 569
Sol: 448
Slayer: 388
Dizzy: 351
Sin: 294
Baiken: 257
Haehyun: 230
Elphelt: 209
Millia: 194
Bedman: 113
Leo: 109
Ino: 106
Chipp: 96
Johnny: 82
May: 68
Answer: 54
Venom: 45
Potemkin: 44
JackO: 35
Jam: 33
Axl: 27
Ramlethal: 19

21110

Pangaea @toptwentynotes.bsky.social · 29/03/2026

i've been trying to work out how i might give a better intuition for the distance values, cS activation range is a good idea. i might go with that and then a marker for something like "half screen away". need to take some measurements

010

Pangaea @toptwentynotes.bsky.social · 28/03/2026

21110