Redditors acting like they know more than Valve about game systems that Valve themselves created is a tale as old as time. Mods please start removing this low effort content. Im tired of being beaten over the head by self proclaimed experts and talking heads.
Fair-Definition3178
Good video, commented on youtube but please send this to valve on their email too! Great explanation overall and you are right, moving and shooting are not consistent right now… it was worse, way worse on the beginning of cs2, i think they will correctly change this if we keep telling because this game is trash right now… just played CSGO yesterday and its another level of fluidity both moving and shooting.
Strg-Alt-Entf
Great video. Couldn’t agree more with the statement that movement should NOT be subticked! It’s aweful and the main cause of problems in cs2 right now.
I also fully agree, that animation should be held back and sent at e.g. 128 tick rate. This would actually not defeat the purpose of subtick! It still makes input feel like 128 tick, although you are plying on a 64 tick server, which valve likes more because it’s cheaper. That’s the whole purpose of subtick anyway: getting rid of expensive 128 tick servers, right? Correct me if I’m wrong here, but I don’t think this was ever supposed to give players an amazing experience. All we ever asked for was 128 ticks…
nuclearwomb1
Good job man, cleared some things up for me
kinginprussia
I’m an imbecile, so I’ve heard subtick described in multiple ways that kind of left me ‘eh’, but the way it’s described here is the first time I’ve been left dumbfounded that valve thought this was a good idea at all.
Was it not obvious that allowing the server to redefine events based on client timestamps is a really bad idea?
CheeseWineBread
The explanation is accurate. And the solution for movements seems on paper good. (the 128 ticks servers won’t happen). If they keep the instant hitreg position + firing animation (next frame), I won’t be bothered to have movements on ticks (or half a tick to reproduce 128 ticks GO movements).
Hyperus102
I have quite a few problems with this.
1. The server doesn’t “interpolate the difference” for movement. It will quite literally split the tick into multiple timesteps. If you start moving between ticks you essentially get a first part that will have you stand still and a second part that you accelerate on. On top of that conflating the shooting with movement mechanics is problematic at best. The two could aswell be called entirely seperate systems. 2. Yes, peekers advantage is worse. [I noted this](https://x.com/H7pernerd/status/1725491850088255861) when per frame prediction started being a thing. The problem is that you can’t attribute what people feel is peekers advantage in CS2. Because: 1. If you didn’t shoot before you died or you shot and the shot actually went through its not related to this and we did not see a lot of clips of people shooting first and then dying and 2. Peekers advantage wasn’t considered worse when latency was much higher for many people due to [higher recv margins](https://www.reddit.com/r/GlobalOffensive/comments/1eedvxn/comment/lffdkfm/?utm_source=share&utm_medium=web3x&utm_name=web3xcss&utm_term=1&utm_content=share_button). What this affects more so is unpeekers disadvantage, where you see yourself a tick further behind the wall and the peeker a tick further out from behind the wall. 3. Crouching or counterstrafing isn’t more effective(as in, throwing opponents off) due to subtick. If anything, you are slower to others. In CSGO a tick recording you crouch will already apply the entire tick worth of movement, hence now there is one tick interval, in which you might have a “half crouch” that gets interpolated on a defenders screen, which means that initial movement is slower. Same goes for all movement. The server doesn’t retroactively apply input earlier than without subtick. You didn’t send a packet to the server and the server would say “well, the next timestep will have you advance in position because you pressed here”, it will instantly apply it, which works because it gets visually interpolated for everyone and reduces latency. You also make 15ms look way longer than it actually is. Even if the initial acceleration was faster, and again, its not, I doubt this would greatly affect the balance of peeker vs peeked. 4. What you describe at 05:25 would be extrapolation, not interpolation, that is, If I am understanding you correctly. But the server doesn’t do that. It doesn’t assume your position, exactly because of this. Instead your position on the server is *always* delayed compared to client. That goes for all inputs, so you can not derive peekers advantage from that. 5. The slight variance in kill feedback actually being a huge deal needs to be backed up by data. The time of multiple shots afterwards is much much larger(more than an order of magnitude) than any inconsistency this could cause in the first place. I have heard many complaints of kill delay, not one of them was timing consistency, else we would get complaints at low latencies too and we don’t really see that. We are talking about an already delayed response. Its not like its sometimes instant and then randomly takes 15ms. More realistically, the change is from something like 20 to 35, not great, but nowhere near enough to explain pros shooting many more bullets. Remember that a normal human reaction time(even for most pros, maybe a little lower) is 200ms or so. This would basically guarantee that even with instant feedback, if you are reacting instead of predicting, you will fire atleast 2 more bullets with an M4A4. Add to that that reaction time will greatly degrade with more visual input, especially with shaking(add the recoil stutter that you mentioned, thats a real issue imo) and 3-4 is perfectly normal.
Nothing much to say about the conclusion because it is based on flawed assumptions. The only thing I will say is that, atleast ground movement, on subtick is much, much more consistent than pure tickbased strafing. You are free to test this yourself or even simulate it(bit more complicated than just splitting ticks unfortunately, also need forced steps up to which you have no friction one tick interval after movement start for consistency, we know Valve does this), we know enough about how subtick gets processed to do such things.
nyotao
it feels better though bullets go where u aim
D47k0
Cs 2 Devs are you watching; someone exposed your crappy piece of engineering. Fix the damn game you lousy coders.
DuumiS
i really wonder if valve will ever make the movement feel like it was in csgo or they will just let the time pass and wait for everyone to get used to it and forget about it. i tried the csgo legacy version a few days ago and i was surprised by how good the movement actually was, i literally forgot about it. also spraying with the ak was much more smooth in csgo and the awp feels better, flicks feel way different and i also noticed that in cs2 u cant quickscope like u could in csgo
11 Comments
valve pls fix, ty
Redditors acting like they know more than Valve about game systems that Valve themselves created is a tale as old as time. Mods please start removing this low effort content. Im tired of being beaten over the head by self proclaimed experts and talking heads.
Good video, commented on youtube but please send this to valve on their email too! Great explanation overall and you are right, moving and shooting are not consistent right now… it was worse, way worse on the beginning of cs2, i think they will correctly change this if we keep telling because this game is trash right now… just played CSGO yesterday and its another level of fluidity both moving and shooting.
Great video. Couldn’t agree more with the statement that movement should NOT be subticked! It’s aweful and the main cause of problems in cs2 right now.
I also fully agree, that animation should be held back and sent at e.g. 128 tick rate. This would actually not defeat the purpose of subtick! It still makes input feel like 128 tick, although you are plying on a 64 tick server, which valve likes more because it’s cheaper. That’s the whole purpose of subtick anyway: getting rid of expensive 128 tick servers, right? Correct me if I’m wrong here, but I don’t think this was ever supposed to give players an amazing experience. All we ever asked for was 128 ticks…
Good job man, cleared some things up for me
I’m an imbecile, so I’ve heard subtick described in multiple ways that kind of left me ‘eh’, but the way it’s described here is the first time I’ve been left dumbfounded that valve thought this was a good idea at all.
Was it not obvious that allowing the server to redefine events based on client timestamps is a really bad idea?
The explanation is accurate. And the solution for movements seems on paper good. (the 128 ticks servers won’t happen). If they keep the instant hitreg position + firing animation (next frame), I won’t be bothered to have movements on ticks (or half a tick to reproduce 128 ticks GO movements).
I have quite a few problems with this.
1. The server doesn’t “interpolate the difference” for movement. It will quite literally split the tick into multiple timesteps. If you start moving between ticks you essentially get a first part that will have you stand still and a second part that you accelerate on. On top of that conflating the shooting with movement mechanics is problematic at best. The two could aswell be called entirely seperate systems.
2. Yes, peekers advantage is worse. [I noted this](https://x.com/H7pernerd/status/1725491850088255861) when per frame prediction started being a thing. The problem is that you can’t attribute what people feel is peekers advantage in CS2. Because: 1. If you didn’t shoot before you died or you shot and the shot actually went through its not related to this and we did not see a lot of clips of people shooting first and then dying and 2. Peekers advantage wasn’t considered worse when latency was much higher for many people due to [higher recv margins](https://www.reddit.com/r/GlobalOffensive/comments/1eedvxn/comment/lffdkfm/?utm_source=share&utm_medium=web3x&utm_name=web3xcss&utm_term=1&utm_content=share_button). What this affects more so is unpeekers disadvantage, where you see yourself a tick further behind the wall and the peeker a tick further out from behind the wall.
3. Crouching or counterstrafing isn’t more effective(as in, throwing opponents off) due to subtick. If anything, you are slower to others. In CSGO a tick recording you crouch will already apply the entire tick worth of movement, hence now there is one tick interval, in which you might have a “half crouch” that gets interpolated on a defenders screen, which means that initial movement is slower. Same goes for all movement. The server doesn’t retroactively apply input earlier than without subtick. You didn’t send a packet to the server and the server would say “well, the next timestep will have you advance in position because you pressed here”, it will instantly apply it, which works because it gets visually interpolated for everyone and reduces latency. You also make 15ms look way longer than it actually is. Even if the initial acceleration was faster, and again, its not, I doubt this would greatly affect the balance of peeker vs peeked.
4. What you describe at 05:25 would be extrapolation, not interpolation, that is, If I am understanding you correctly. But the server doesn’t do that. It doesn’t assume your position, exactly because of this. Instead your position on the server is *always* delayed compared to client. That goes for all inputs, so you can not derive peekers advantage from that.
5. The slight variance in kill feedback actually being a huge deal needs to be backed up by data. The time of multiple shots afterwards is much much larger(more than an order of magnitude) than any inconsistency this could cause in the first place. I have heard many complaints of kill delay, not one of them was timing consistency, else we would get complaints at low latencies too and we don’t really see that. We are talking about an already delayed response. Its not like its sometimes instant and then randomly takes 15ms. More realistically, the change is from something like 20 to 35, not great, but nowhere near enough to explain pros shooting many more bullets. Remember that a normal human reaction time(even for most pros, maybe a little lower) is 200ms or so. This would basically guarantee that even with instant feedback, if you are reacting instead of predicting, you will fire atleast 2 more bullets with an M4A4. Add to that that reaction time will greatly degrade with more visual input, especially with shaking(add the recoil stutter that you mentioned, thats a real issue imo) and 3-4 is perfectly normal.
Nothing much to say about the conclusion because it is based on flawed assumptions. The only thing I will say is that, atleast ground movement, on subtick is much, much more consistent than pure tickbased strafing.
You are free to test this yourself or even simulate it(bit more complicated than just splitting ticks unfortunately, also need forced steps up to which you have no friction one tick interval after movement start for consistency, we know Valve does this), we know enough about how subtick gets processed to do such things.
it feels better though bullets go where u aim
Cs 2 Devs are you watching; someone exposed your crappy piece of engineering. Fix the damn game you lousy coders.
i really wonder if valve will ever make the movement feel like it was in csgo or they will just let the time pass and wait for everyone to get used to it and forget about it. i tried the csgo legacy version a few days ago and i was surprised by how good the movement actually was, i literally forgot about it. also spraying with the ak was much more smooth in csgo and the awp feels better, flicks feel way different and i also noticed that in cs2 u cant quickscope like u could in csgo