CS:GO -- Does 2 milliseconds matter?

[lexlazootin]

Lets say you move 10cm in 3ms for a flick shot. that's 120km a hour if I'm understanding your point correctly.

yea, i don't think so.

[spacediver]

Lets say you move 10cm in 3ms for a flick shot. that's 120km a hour if I'm understanding your point correctly.

yea, i don't think so.

Well a lot of flicks are much less than 10 cm, but I think we're all agreed that 3 ms is too short for a flick shot.

But the precision question is interesting.

somewhat relevant:

http://www.ncbi.nlm.nih.gov/pubmed/7789435 (recreational throwers)

http://www.ncbi.nlm.nih.gov/pubmed/21775713 (skilled throwers)

In these studies, the kinematics and forces involved meant that the final muscular contractions in the fingers were funneled, or channeled by the neuromuscular feedback from the gross motor movements (I believe the measurements were taken relative to certain kinematic events, rather than to someone saying "Go!"

There are likely feedback circuits that can be developed in mouse control - for example, the constellation of muscles involved in generating a flick motion can form a circuit with the muscles involved in pressing a button. I wouldn't be surprised if expert gamers could get precision under 10 ms. One could even do some simple math and figure out the precision needed to consistently hit a 90 degree flick rail in quake live, if you could somehow time the movement to get a sense of the available time window for the shot.

[RealNC]

I would need to draw images here to explain.

OK, consider this. You're getting at least 128FPS. You do a flick shot. The spots that the shot can register are "quantized" in 7.8ms steps due to 128 tickrate (distance in time units here works the same as in racing.) Imagine a screenshot of the game, with red dots that show the possible hit registration points at the current speed you flick the mouse. Those dots would have a distance of 7.8ms between them. Now, if you fall below 128FPS, these dots will change distance. But that's client-side. Server-side, the dots stay the same. So when you shoot, the dot you hit on your end will be mapped to the nearest neighbor dot on the server. (If you get more than 128 FPS the dots also change distance, but the mapping to the server works correctly because the distances get smaller, not bigger.)

A small change of 4ms will result in the shot getting registered on the next dot. This is the cause of hit registration issues when you have low frame rates during high-speed flick shots. You cannot of course tell what the issue is, since you can't tell 4ms time differences apart. What happens instead is that you miss more shots in low frame rate scenarios compared to when you get high frame rates, but you can't actually tell why that is.

[lexlazootin]

I understand what you are saying but under the grand scheme of things 2 milliseconds doesn't matter to me. or if you drop a packet the EXACT moment you shoot because it's such a small time frame it's pretty much random at that point if you get the shot.

[Chief Blur Buster]

That said, I am hearing multiple reports that ending of aim (crosshairs first overlapping enemy) and shot (the kill) is often occuring during the same tick for some people. Assuming a 128Hz tick rate, that would be a window of ~8ms, or a center of +/- 4ms.

(A curious question: How do you determine it's occuring during the same tick? Or is this something Valve said? Or?)

[Chief Blur Buster]

A small change of 4ms will result in the shot getting registered on the next dot. This is the cause of hit registration issues when you have low frame rates during high-speed flick shots. You cannot of course tell what the issue is, since you can't tell 4ms time differences apart. What happens instead is that you miss more shots in low frame rate scenarios compared to when you get high frame rates, but you can't actually tell why that is.

Another easy way to understand this -- At 50fps panning the screen at 2000 pixels/second (one screenwidth per second), you have a 40 pixel step in motion between frames (2000 divided by 50 equals 40). But at 200fps, panning the screen at 2000 pixels/second, there's only a a 10 pixel step in motion between frames (2000 divided by 200 equals 10). If you're shooting moving targets, a higher framerate also means the enemy is less likely likely to "skip" to the other side of the target in its granular, quantitized, framerate-limited motion. And even if it didn't skip visually, the server might have (aliasing effects between tickrate and framerate, I think...).

In this case, framerate dramatically aiming in CS:GO, even if the framerate is far beyond the monitor's Hz, and even when framerate is far beyond the server tickrate.

I understand what you are saying but under the grand scheme of things 2 milliseconds doesn't matter to me. or if you drop a packet the EXACT moment you shoot because it's such a small time frame it's pretty much random at that point if you get the shot.

In competitive gaming, with a highly attuned individual (Olympics league reaction times) competing for tournament prizes/money, it doesn't seem random. Every millisecond matters, when you're trying to cross the Olympics finish line -- and this is what is happening in the esports arena...

There are many competitive players that can accurately time a shoot of stuff moving past their crosshairs (without pausing crosshairs on target). Some of them can pan really fast and still frag pretty quickly, if you watched some Twitch.tv videos of them. Shooting moving targets (whether target moves past, or you 'turret' the crosshairs past them, or whatever) is in territory where milliseconds of consistency matter -- similiar to an expert marksman shooting skeet (clay pigeons) very early in its flight where it's moving fast. Being just merely 2-3ms off with fast-moving clay pigeons will cause you to miss your shot of the clay pigon in real life.

Recreational players at home (even myself included), sometimes can't always fathom why the heck a single millisecond matters -- but it makes a lot of sense in the Olympics leagues when you're competing for big money in the esports tournaments with other people who's within mere milliseconds of your abilities.

Aside: It'd be fun to do some statistical scientific analyses of pro CS:GO players. Speed of object moving past crosshairs. Average hit rate for each specific speed. Other interesting cases of precision/accuracy/latency feats. Statistical graphing & visualization of hit failures and successes based on framerate, tickrate, etc. Scatterdot graphs. 3D barcharts. Bell curves. See patterns/trends. I'd love to see data visualizations of this, but it would quite require a few hooks into the game engine and/or server, or some automated high-speed-camera analyses (including image recognition), or some creative method of capturing data accurately in a wholesale manner for statistical analysis. Don't know if anyone would ever fund an exhaustive study such as this, and release findings to the public -- but if it ever happened -- it should be interesting

[flood]

I would need to draw images here to explain.

OK, consider this. You're getting at least 128FPS. You do a flick shot. The spots that the shot can register are "quantized" in 7.8ms steps due to 128 tickrate (distance in time units here works the same as in racing.) Imagine a screenshot of the game, with red dots that show the possible hit registration points at the current speed you flick the mouse. Those dots would have a distance of 7.8ms between them. Now, if you fall below 128FPS, these dots will change distance. But that's client-side. Server-side, the dots stay the same. So when you shoot, the dot you hit on your end will be mapped to the nearest neighbor dot on the server. (If you get more than 128 FPS the dots also change distance, but the mapping to the server works correctly because the distances get smaller, not bigger.)

A small change of 4ms will result in the shot getting registered on the next dot. This is the cause of hit registration issues when you have low frame rates during high-speed flick shots. You cannot of course tell what the issue is, since you can't tell 4ms time differences apart. What happens instead is that you miss more shots in low frame rate scenarios compared to when you get high frame rates, but you can't actually tell why that is.

of course

Consider that an AWP flick-shot is in the ballpark of 2-3ms.

it was kind of vague what was meant here but i took it to mean that you can complete the flick in a timespan of 2-3ms.

in the end csgo is a tactical game and not a game where being faster by 5ms will give you a huge advantage over someone else... the game itself is way too random and inconsistent (weapon inaccuracy, netcode, framerate microstuttering, messed up jumping hitboxes) for individual hand eye coordination to be a deciding factor.

[flood]

Recreational players at home (even myself included), sometimes can't always fathom why the heck a single millisecond matters -- but it makes a lot of sense in the Olympics leagues when you're competing for big money in the esports tournaments with other people who's within mere milliseconds of your abilities.

except there things are actually accurate to within milliseconds.
in csgo the server tickrate (~128), client framerate (~200), monitor refresh rate (144), mouse polling (500/1000) are completely independent and out of sync

[lexlazootin]

Oh boy, i failed English but i'll give explaining this a try ;P

If the refreshing of the server is 128(7.8ms) times a second and the player moves 215units with a AK out that's 1.68units of movement every refresh (A player is 64 units wide). It would take 38ticks for you too walk your entire body distance across, so if you were waiting for someone too cross your corsshairs around a corner with a AWP you have a FULL 38ticks or 196.4ms to HIT the player and most people leave ~about one bodys lengh away from the way too give extra leeway of time into the shot, doubling the time given.

That's A LOT of time, and unless you are swiping your move at 128km a hour trying flickshots (please don't) you are going to hit or miss the player REGARDLESS of a few milliseconds

peoples reaction time tests on a small web-based "reaction test game" is always much better then ingame, when people get "Insane" quick shots around a corner awping, it's about 190ms.

[flood]

i can get 150-160ms shots for t side picking mid in dust2 but ya

https://www.youtube.com/watch?v=qFGpLdGEL9k

this is the only type of shot where being off by a tick could make him miss completely