From the 10 matches I played the delay was definitely less noticeable than it would have been in even some green bar connections in SFIV. Probably had similar to like a 2 or 3 frame delay like in KI, but that’s still pretty good in a game that now has input buffer for combos. Unfortunately it’s hard for me to really test that when most of the gameplay I’ve done offline was a small amount of games at Evo and the training mode during the beta.
I of course ran into about 3 or 4 people where the rollbacks were pretty bad. Like worse than what I usually get in KI, but when it was smooth it felt good.
As I said before, the problem is that there’s no indication of the actual connection quality given. There’s no way to judge whether a connection should be good or whether you should expect rollbacks. IMO, that’s something that’s needed in the next phases of testing so we can give accurate responses.
jitter is king. there’s a wide range of latency which is playable and can be adjusted to.
but if the latency isn’t consistent, it’s impossible to adjust and the game is pretty much destroyed.
i’m not really sure what they can do about packet loss since networking is like some black magic shit.
there is probably a way to lessen the chances of severe packet loss though or ways to mitigate (or even eliminate) the effect of less extreme loss to create a more stable/consistent experience.
An average ping time is more useful for anyone trying to get a game in though, since you can easily calculate delay settings from ping times.
On a related note, I’d really wish that they’d display the actual delay settings before the match. Heck, they can take a page from Skullgirls and even let the players adjust it before said match. Of course, seeing as people are idiots and will probably set it to 0 everytime causing rollbacks up the wazoo, they should probably just have 2 settings, “fast” which is just the average time a packet takes to go one way in ( delay = (ping /16.67)/2 ) and “stable” which is just delay based on the ping time ( delay = ping/16.67 ).
i imagine the game will calculate that itself though.
maybe capcom will let us choose a preference like smooth vs precise but I don’t see them letting you set your rollback frames like ggpo. maybe it will default to auto and let you select in lobby though, that would be ideal.
my point is delay doesn’t mean shit if you have anything more than a little jitter.
if netcode is better than release sfxt then that’s great. i played sfxt doubles with lenin and it was plenty fun, that’s NJ to CA and then playing against people from all over too.
The problem i see so faar Capcom ignores that “Region” means “Country” i would love to choos “SOME” Countrys in my search but i only can say “all in my country” or ALL; that is maybe okay for USA which is a big Country… but take Europe for excample… my connection in SF4 is excelent for all countrys in Europe… but of cours its bad for the most not europeen countrs (except Canada and some in USA) only searching in germany is a bit simple mindet… there are to less players… so it would be great if the searching would consider Europe as ONE thing instead of every Country…
Right now (i.e. in the beta) those values will be fixed so they have a yardstick for performance analysis.
It’ll be interesting to see if they eventually offer a connection stability indicator in future tests. Likely this was omitted so they could collect poor connection data alongside the good. Were people able to discern poor connections from good in the beta, I expect many would have opted out of playing that person and Capcom wouldn’t have been able to collect a nice variety of performance statistics.
You need that variety of data in these tests to strike a balance. Personally, I would have dropped the locale indicators (flags) to amp up the placebo effect even more, but from what I can tell most people just ran with whomever they were matched up against regardless as during the beta window actually getting to play for many was a bonus in itself.
I expect user defined connection tweaks like those you’ve mentioned will make it into the final game. One size fits all doesn’t strike me as very wise for rollback though I gather that how they handled SFxT regardless, no?
While this is 100% speculation on my part, I wonder if the netcode in general in the beta can never be quite as good as the final product due to the fact that they are logging and gathering a ton of information from this test for analysis. When the final product is out, chances are they won’t be doing anything like that. I wonder if there is any overhead for all of this data they are gathering from people playing the beta.
Of course. It makes no sense to have a UDP stream for real-time match data outside of internal debugging scenarios. The impact to overall connection quality would skew connection and tuning criteria.
Stats will be collected once two players are paired up (initial connection conditions, locale, platform, etc) with further relevant data collected and buffered locally in-game and submitted once a victor is decided.
You actually got your hands on the Beta? I couldn’t even get past the start-up screen when it was active, before the closed it. When are they re-releasing it, anyone know?
Yeah the netcode is pretty bad. I hope this is part of the “stress test” for tonight, because I’m assuming that means they are letting more people on than normal so it’s expected. But netcode is pretty unnaceptable IMO at this point. Good connections are far and few between now, compared to how it was before.
Not even CLOSE to being on Killer Instincts level. I’ve definitely had dropped inputs as well, which I had in SFxT but never experience in KI.