Armanlex

@nuejaebs This stat is no good, it shows matches played with all characters and not just the character that has the highest rank. I would suggest looking at their 100 recent matches with that character. If that is below 50, then I would be surprised.

This is the win rate of a player who has now gotten GoD 2 GoD is currently the top 0.5% percentile of players globally What serious competitive environment allows you to reach a rank like that by winning less than half of your matches?

@koenjideck @ugotpauld @RichardWinz Tekken is like 99.99999999999% deterministic. They had some rare bugs in t7, mostly at the wall, but I think those are probably gone in t8, haven't seen anyone mention them in years, so now it may be at 100%.

@ugotpauld @RichardWinz I think it depends if your animation system is deterministic enough to allow that and for a lot of these code bases it's not always fully deterministic. Tekken is not perfectly deterministic and has suffered from desync bugs for decades, for example

The issues for rollback in 3D fighting games, and a solution.

@akiraktu @miharasan Ooooh, that's a very interesting angle I didn't consider. As in the rollback artifact will usually only be visual and not affect the gamestate, so the rollback artifact won't be as jarring? Hmmmm...

In Tekken 8, a character’s position, posture, rotation, as well as attack and hurt hit collision data are updated continuously on a per-frame basis. Each frame’s state depends on the result of the previous frame, which means that intermediate frames cannot be skipped, approximated, or coarsely rounded without breaking simulation consistency. First, it is important to clarify that the displayed latency value of 80–100ms represents the round-trip network time (RTT) between the two players. This value reflects user-to-user network conditions such as connection quality, routing paths, and physical distance, and is not a metric that evaluates the quality or performance of the netcode itself. In online matches, both players’ systems must reproduce identical simulation results based on per-frame input data. As long as there is a delay before the opponent’s input arrives (that is, network round-trip time), it is impossible to perform accurate collision detection or state calculations using inputs that have not yet been received. Rollback netcode mitigates this limitation by predicting inputs and re-simulating frames once the actual inputs arrive. However, this technique fundamentally assumes that inputs arrive late. Rollback cannot make inputs arrive earlier, nor can it reduce physical distance or alter network routing. At this point, it is useful to explain the difference compared to Street Fighter 6. Street Fighter 6 is designed as a 2D fighting game, in which character positions and hit collision updates occur in discrete steps over multiple frames (roughly every 4–5 frames). Because of this structure, when rollback occurs, the resulting differences are relatively coarse, making timing discrepancies less noticeable both visually and in terms of player input. Tekken 8, by contrast, updates hit collision data every single frame, with collision volumes continuously moving and deforming in 3D space. As a result, even a one-frame discrepancy can directly determine whether an attack hits or misses, or whether a move is successfully evaded or not. Due to this structural requirement, the effects of the same network round-trip time tend to be more perceptible in Tekken 8 than in Street Fighter 6. This difference is not a matter of netcode quality or superiority, but rather a result of different precision requirements dictated by game design. To completely eliminate the effects of network latency in Tekken 8, one would need to choose one of the following options: •Reduce collision and state updates to multi-frame intervals •Significantly increase input delay •Lower hit collision precision All of these options would fundamentally compromise Tekken’s gameplay. They would break the conditions required for frame-precise movement, whiff punishment, and guaranteed punishment, effectively turning the game into a different experience. In other words, the current behavior is not the result of something being “left unfixed,” but rather the result of optimization pushed to the structural limits of the design. Any remaining latency is not something that can be resolved through netcode tuning, but instead reflects the inherent constraints imposed by user-to-user network conditions and physical distance.

@victoriquedesu @miharasan Its to reduce network delay to a minimum, since fgs are by their nature extremely precise games. And since they are 1v1, they don't need a server to be an authority, so using a direct p2p connection makes a lot of sense. Plus not having a server reduces operating costs a lot.

@Armanlex_ @miharasan Thank you for the explanation. But why do fighting games require such complicated and inefficient system?

@victoriquedesu @miharasan And those authoritative models also have massive downsides. Ever heard of rubberbanding, or peakers advantage? Those don't exist in fighting games, but fighting games have their own problems. Networking is a pain in the ass and there's no perfect solution.

@victoriquedesu @miharasan Because those games run on an authoratitive server model, which can validate and enforce gamestate. In fighting games both computers need to be in sync while simulating the game independently, and both players network instability necessarily translates to the other player.

@victoriquedesu @miharasan Tekken 8's wonky hitboxes are not related to the netcode, those happen while playing offline too.

@miharasan Any modern 3d online game (from fortnite to dota2) requires high precision, yet all of them has way better online experience. Why is that? Also, if T8 is so sophistycated why do we have so much clips of weird wiffs and characters hitting someone befind their back?

@miharasan And if collision updates occur in batches, but within a batch they happen sequantially for every frame so that the simulation is identical to offline, that could still cause major rollback artifacts. I'm really trying to understand how that could work.

@miharasan Could you please explain more what you mean by coarseness, and how could it be possible for sf6 to not update collisions ever frame? I can't comprehend how that could work. If updates don't happen every frame, then that could lead to a different gamestate compared to offline.

@alikhalil809 I understand why things look weird behind the scenes. What I don't understand is what fails in the code and allows them to suddenly become visible. I saw another bug where after coming back to practice mode the p1 character will start animating random moves from other characters.

@DesignerLeake Ah then it can't be what I was thinking. Those 13th 14th gen cpus need to have their mobos bios up to date to prevent random crashes.

@Armanlex_ 11th gen. i5

Reinstalled the game... Default settings...

@Qveeny9 He is -10. Your fastest ws punish is probably something like ws4 which is typically i11. Most characters, if not all, punish the second hit with a dick jab.

so somebody please explain to me wtf I wasn't able to punish him? like I'm really about to throw my controller out this window #T8_report

@NechoPlays @Harada_TEKKEN Nah, that probably isn't worth it. Look up if there are issues with your cpu/mobo; but if there aren't I wouldn't suggest updating anything. The chance of something going wrong is small, but if it goes wrong you could end up needing to buy a new mobo.

@Armanlex_ @Harada_TEKKEN I have a Ryzen processor. I'll update BIOS and check if that helps. Thanks!

#T8_report @Harada_TEKKEN hey I've been having constant crashes post matches and also in the graphics settings menu. Something is wrong with the game, for me is unplayable. Also, some stages have been having sever FPS drops. Please fix.

@NechoPlays @Harada_TEKKEN But know that updating your mobo bios is risky, so read up on the process thoroughly.

@NechoPlays @Harada_TEKKEN Do you have a 13th or 14th gen intel cpu? If you do, you should update the mobo bios cause those cpu's run with too many volts & crash on ue5 games, and those crashes can appear suddenly cause the cpu is gradually degrading. I updated mine and all my crashes went away.

@bigsaruSSR @_Shiroi_Kaze_ 👀

🤣

@EternalDragon28 Not a bug, you pressed right before the buffer window. The buffer window is 9f. You have f4 (6), n (5), f4. The first f4 was pressed right before the buffer window, the second f4 was right after the window, so the punish came out late. Mash faster or time your punish better.

I was raging man, why didn't he die!!

@KrisuuPL Looks like you did the rage art too late and they hit the ground before rage art connected. Not a bug, it's always been this way.

#T8_report Bryan Fury... Can someone please explain to me what the hell happened here❓ Did you turn the players into Bryan Clowns❓ What is this‼️

@thunderf00t Have you seen the buzz with the "thunderstorm generator" that supposedly can eliminate combustion emissions, increase efficiency massively, create infinite energy too I guess and much more? Randall Carlson is pushing it a lot and he's been on Joe Rogan.