For what it’s worth I tried reproducing your issue by creating retention on the upstream line (client to server), basically: occasionally buffering a few outbound packets and then delivering them all at once.
It does create double animations (mispredictions), slight shot delays, does not spike the netgraph as it takes place on the upstream band, does not appear as loss either as it’s just delay in delivery.
Though it does not appear exactly as in your video so it might be off.
Maybe you have a faulty node in your region that throttles packets? Something along those lines. Does it happen on other servers? regions? in other games?
For reproducing: clumsy on “udp” with outbound throttling and a low delay (~30ms) and reasonable frequency (~5% chance) so the server is not given a chance to adapt.
Its hard to determine if its happening in other games since the netcode I believe is quite different in Overwatch compared to some other FPS so I can’t confidently say it doesn’t happen. The only fact I know for sure is both ISPs exhibit the same symptoms.
Currently we’re in agreement with Cgg that this is probably some kind of Sim misprediction. Since the server is considered the ‘source of truth’ of what actually happens in game, what is likely happening is that VeiledFlame’s client is mis-predicting when his attack animation can go off. The client then receives a packet from the server which then corrects the animation, resulting in the odd behavior.
I’ve changed the title of this thread to better reflect the area that this is occurring in. If anybody else in Alberta is seeing this behavior, please post the following information:
A looking glass test to the overwatch server. Please make sure to check all three of the boxes on the bottom left.
Copy the code I have below and paste your tests into the designated lines:
~~~~
WinMTR goes here
~~~~
~~~~
Looking Glass goes here
~~~~
In the mean time, we’re gut checking ourselves with the engineers on this to make sure we’re on the right track. Thanks for all the help so far VeiledFlame. We’ll provide more insight as we can.
What I suggest if it helps, I can make you connect to a game for which I know the netcode. I’ll be able to tell if there’s something odd there.
Because it’s a different system it could provide new insights and help us understand. I’ll share the data (in private). It takes 1h max if you’re down.
Sorry to have barged in btw, that ended up catching my curiosity.
Appreciate the all the help so far Drakuloth, I’ll keep an eye out for any updates. If you guys happen upon anything relating to either Shaw or Telus let me know as well so I can make them aware.
That would be great if you wouldn’t mind. The biggest thing is trying to find out if there’s network instability with other games, so this would be incredibly helpful.
Sure leave me your battletag and I’ll add you when I’m off work. I’ll use the more stable of the two ISPs for the test, and yeah my VPN is all good to go on that PC.
Sounds great, and thanks again for helping me out. It’s been…interesting trying to figure this out to say the very least.
We tested through Shaw against an AWS instance based in Oregon. He had his client run at 125Hz both ways (standard for this game) on a ~40ms ping. Everything went perfectly fine in game. I couldn’t find anything odd or unusual.
Just to be a little technical, said game doesn’t extrapolate clients on the server as OW does (it just executes commands as they come, so players with inconsistent connections visibly hitch in the game). I was expecting something obvious (or none of it) - the latter happened.
Just like the netgraph in OW the connection (as seen from the server) is textbook consistent. I have packets from his clients evenly spaced at 8ms intervals with very little variation in the dumps.
As I was explaining, it looks like the issue has to do with the server extrapolating a (missed) command and then executing the actual action with delay when it’s finally arrived (causing a rollback of the animation on the client/a slightly delayed shot in the game). Because there’s no visible lag spikes or loss on the graph I thought it could be a node throttling packets on the way up to the server - a node holding onto packets for some ms, then delivering them all at once.
That’s where it gets tricky. I went into more details and I actually have exactly that in the dumps, but not to the extent I would have expected. I have a number of small lag spikes in an otherwise perfectly perfect dump, basically packets visibly held for ~10ms each time, and later delivered in a burst (at 125hz that’s 2 packets…).
So I’m really unsure. Either there’s something going on that’s not directly related to the local ISPs as a whole, or OW in his case (considering the otherwise heavenly connection) is doing too aggressive a buffering, is then caught off guard by random unexpected throttling. Commands in OW run at 16ms intervals, 10ms is not even one command. But maybe in perfect conditions OW lets the server catch up with the client stream?
I have one especially clean record of 80s of gameplay. There are 3 cases of ~10ms throttling in this dump.
@drakuloth
Its been just over a couple weeks so I figured I’d check in. Has QA pinpointed the cause of the issue at least? All I know from my support ticket is that it required ‘extreme conditions’ to mimic my issue.
We haven’t heard anything back from QA yet other than that they’ve seen the issue before but only in very extreme cases (similar to how CGG intentionally mangled their network.) The engineer we’re working with is basically seeing if he can reproduce the event and then trying to figure out in what situations a player would see it. I’d keep sending players having the same problem as you to this thread so we can get more reports. As of now you’re the only official report of the behavior, which means the priority of investigating the issue is lower.
That’s unfortunate. Yeah CGG essentially mimicked the result by hindering the connection, so I’m just trying to figure out if its on my end or blizzards end.
Posting from my other account since I can’t repost on my main, but is there any update on this? I’ve been waiting quite a while now to hear back from a simple report on the issue. All I need to know is where the issue is originating. Or even just how QA reproduced the problem would be great.
Without accurate reproduction steps or more reports there’s not a whole lot we can do here. I’ve bumped our thread to see if we got lucky, but the last time we talked with the engineers about this it was basically them happening upon the problem completely at random while doing some things during debugging that a player like yourself wouldn’t ever be able to do because it requires backend tools.
Edit: We heard back, and without more information there’s nothing else we can do on our end. We don’t have information on how to consistently reproduce the problem using a normal setup, and the kinds of things one can do to reproduce the problem are completely unnatural for a normal PC/network setup, so the information is pretty useless for testing. We totally understand that this is a frustrating error, but unless you can direct more people to this thread who have the problem and who can get us more information, we’re at an impasse here.
For now your best plan is probably going to be finding the best VPN you can and using that to play to at least minimize the problem.
Appreciate the response. From what I’m gathering here its some way shape or form of a network issue that seems to happen at random. And more specifically it seems to be at least on a provincial level to Alberta, I went to my brother’s house in a small town 3 hours north of here and brought my setup, same issue there too with another ISP (wasn’t fun hauling that setup out there).
That’s really all I wanted to know so I could investigate it myself. As much as I’d like to use a VPN it does almost nothing to fix the issue from my findings. Unfortunate but the game isn’t playable for me in the current state its in. Not sure if its just isolated to Overwatch but not much more I can do when my network tests are coming out clean.
I wanted to follow up on this forum here and throw my hat in, as a player who get frequent rubber banding, inaccurate hit detection and other major issues and who has been advocating for a solution since very early in the games life cycle.
I know for a fact that I’ve reported these issues well over a single digit amount myself alone and have followed literally hundreds of Canadian players who all share this inconsistent experience.
Please, I’d love some way to understand why this is such an issue for you and your team, but from everything I’ve gathered, it’s simply your ignorance that perpetuates this biased and poor service you claim to be fair.