Let me start by saying my issues started seemingly overnight in 2015 in CS:GO. It was the only game I played consistently and had over 3000 hours on record at the time (I now have 8k
). The game "feel" was hard wired into my brain, my fingers, my game sense, my timings and my muscle memory. Then all of a sudden, terrible hit reg, impossible recoil, weird spread, inconsistent mouse input, insane reactions and peekers advantage from enemies, getting running headshot by everyone at all ranges while I would have to tap a static target in the centre of the head up to 10 times until I got a headshot...by which time I got turned on or third partied by a 1 tap. For the previous 3 years I had been a 1 tap headshot machine, it was my play style, then overnight it was impossible to do.These issues persist to this very day. I sometimes get a rare occasion where it feels like a switch has been flicked and everything falls into place again, but it's rare and never lasts long. It's usually for half a match, maybe a session, maybe a couple of days and sometimes It can last up to a week, maybe once per year which for some strange reason I have noticed has been in November for the past 3 years. Anyway!
Back when this issue started I used to get spikes of up to 3% choke IN CS:GO whenever I got into an engagement, someone peeked me, etc. I'm 100% certain this was the cause of my issues. It was a common problem that not everyone...in fact not many people at all relative to the player base, had at the time. I noticed also the same kind of loss in PUBG and other games at the same time but it eventually went away in other games and stayed in CS:GO for another 2+ years. In 2017 Valve released an update. In the patch notes was "Fixed a bug causing small amounts of choke for some players, particularly on DSL connections". At the time I was on VDSL2, Fibre to the cabinet, last 300 meters or so copper.
At the time I remember that being a strange patch note. How can you fix something that occurs on a specific connection technology with a game update? But I thought, fine, ill give it a shot. The game now reported 0% packet loss at all times. I thought GREAT, is it finally fixed? Nope!
I still got Ferrari peeked at every turn, I still had terrible hit reg, I still had sluggish inputs, I still felt like I was way behind the server, and I still had delayed kills. Etc, etc. The only difference now was that instead of people stuttering and jittering or just outright teleporting around corners, they were peeking extremely fast with insane reaction times (my reaction times are 130ms on a good day, 150ms on a bad day or when very drunk for reference. I can get number one weekly on detection ultimate in aim lab.) and impossible accuracy while moving. The experience was exactly the same, the symptoms just presented themselves differently and net graph now said 0% choke but the experience was just as terrible, if not worse, more delayed.
This is when I started to think "ok, they didn't fix the choke issue, they somehow mitigated it/compensated for it". But to do that, they would need to be doing some insane packet buffering delaying everything by a large amount. Now that would explain the gameplay experience, but it wouldn't explain choke not showing in net graph unless they were out right ignoring it/misreporting it or net graph conveniently wasn't working anymore. Then it made me think a little about SDR (Stream Datagram, Valves own protocol and network for their game traffic).
What I think is actually going on is that many people around the world are in locations that are suffering from packet loss when gaming and Valve have implemented some form of retransmission within their SDR protocol. I'll get to my reasonings for this conclusion in a moment. But if this is indeed a wide spread problem it would make sense that all modern games/relays/servers implement the same kind of thing and why people that have these issues seem to have them in all games. With modern official game servers no longer being direct connection and going through relays, traffic can be handled however they see fit once it's on their network and you would be none the wiser since you can't follow the route once you've entered their network. You can only run limited tests to the relay but are in the dark when to comes to what your traffic is doing after that point and have to rely on internal game telemetry reported from the relay.
My reasonings:
The obvious -
1: Well first of all that strange patch note for a particular connection type.
2: The choke existing for 2 years before suddenly disappearing.
3: The same issues persisting with slightly different symptoms.
The not so obvious -
4: I've been monitoring Wireshark captures on and off for years and there seems to be something very inconsistent with packet sizes. In CS2 on a full deathmatch server I can receive as many as 4 packets with high frequency in a single tick. Huge packets, I'm talking 3 full packets 1300+ bytes (1200 max routable defined in source engine + headers) trailed by a smaller packet. They all have small deltas that add up to around 16ms expected for a 64 tick server. This will be how the packets look when having a particularly bad experience. When the experience is better there will less packets per tick. When I look at the SDR stats in steam (game library and press ctrl +\) in a full dm server when the experience is very bad I'll be seeing as many as 140-150 packets sent from the server per second. On a day when it's a bit better experience I'll see around 128 packets per second. The better the experience the lower the number of packets per second I receive from the server. Why on earth does a game as basic as CS need to send more than 3 full packets per tick? That's a lot of data for the type of game it is. Were talking over 4000 bytes per tick, It doesn't make sense. If retransmission is happening then older data would need to be bundled along with the data on future ticks. This could explain the variation in packet sizes/split|number of packets and a correlation between that and game experience..
5: Retransmission would technically mean packets are being delivered rather than dropped.....even though they did actually drop but were then resent adding extra variation in latency. That would explain why telemetry reports 0% packet loss even if you are dropping packets.
6: The variation in timings for the server to respond to events suggests high jitter, which retransmission would introduce.....however the SDR telemetry reports my connection to be 100% perfect so it would need to be dismissing jitter that is inherent to packet loss by design.
7: Using clumsy in CS:GO I would have to force massive amounts of packet loss before net graph would even report it and before it would even have any visible effect on the apparent smoothness of the game. Enemies still didn't warp and teleport, my character model wouldn't teleport either until I hit crazy amounts of loss, like 70%, then everything would start teleporting like crazy. Up until that point the only real visible effect on gameplay (outside of massive peekers advantage and terrible hit reg) was the huge amount of variation in timings from server responses to hits, which happens to match the general gameplay experience, just to a much higher degree. This was also one of the main issues when the original choke issue started.
8: In CS:GO I would set my rate to 20000 (you cant set it that low in CS2 anymore) which would cause me to have around 50% packet loss......but there was no teleporting. How can you not teleport in an online shooter with 50% packet loss unless there is some kind of HUGE amount of compensation going on? Eg, retransmission.
I'd like to ask people here to make Wireshark captures over as long of a period of times as you can/want to/care to. It doesn't matter what game you play but I think valorant and cs2 official deathmatch would be the best testing grounds here. Save your captures and take note of whether it's a particularly bad session or a good session, keep a diary of your experiences, location, dates, and times alongside each capture file and see if it corresponds to larger/more split packets or smaller/less split packets. Also if there are people here that 100% do not have issues with their online play, it would be handy to see captures from you too so people with both good and bad experiences can compare. I haven't provided any data in this post yet, I'll be started to collect fresh again from tomorrow and keeping proper track of it. I hope some of you will do too.
Cheers!
TLDR: I think the majority of problems are cause by global scale issues with networking affecting certain areas and the issue is so widespread that game developers have had to "design" their way around it.
