Þetta er mjög áhugaverð grein um hitboxin í CS: Source og af hverju þau eru EKKI gölluð.
Það eru nokkur video í þessu, ég uppaði þeim á innlent dl svo fólk færi ekki að nöldra ;)
Netcode/Hitboxes
Well its not really a full blown investigation its just I always wanted to clear something up about all these videos I saw where the hitboxes were leading in front of the player and rediculous headshots were the result. So, thus this little thing was born…

Well I loaded up my own server, tweaked my rates a bit got sv_showimpacts 1 on and sv_showhitbox 2. This way I could see the bot too.

I got it to show the hitboxes ahead of the bot just like in many of the videos we see about Valve's netcode/hitbox issues. I got to do what I always wanted a video to show, I shot the hitboxs with many shots, there's no doubt that I didn't shoot where the hitboxs said they were… only to find, suprise surpise, no hits were registered. Where however proceeding to shoot the actual model did register a hit.

I then reset my config and its rates back to what they used to be and I found the hitboxes while walking just about a half body ahead and running just a little more, nowhere near as drastic as has been sugessted (sp!?) before, even then proceeding to shoot the shown hitboxes wouldn't always register a hit where directly shooting the model would.

There is a way to get the hitboxes to line up with the model and stay there, this I have known about and used before in fact with my old computer, cl_interpolation I believe the command was or interpolation 1. It's a quite well known cvar anyway; however this made my game very jerky very unplayable, to me at least. But at least it would keep them exactly in line if I wanted them to be.

All in all its lead me to this conclusion, CS:S Hitboxes are fine (we all know what the hitboxs are by now probably) Valves netcode however may appear slightly odd in the point it puts hitboxes ahead of the player but how many of us go and play CS:S on our client server, when in an Internet server it makes sense because of latency.

I really wanted to get sv_showhitbox to work on an online server, luckily of course i have my own server for my clan, however the cvar will not work and it does mention something about high bandwidth and listen servers only, either way it wouldnt work. I did put on sv_showimpact on and people could see when their client register a hit. This did prove to be very interesting as I myself had two seperate ocassions of the hour or so i was playing where my client registered a hit but the server didnt. Both these ocassions (sp?!) were when i was using the scout, the reason for this is its a very good way to tell if you are hitting people or your hitting the space infront or behind and getting the hit, I was sniping through the double door on dust2 from middle to t spawn with my scout and i shot a guy and he shot me with the awp which killed me. However my client had registered a hit on that player as it clearly showed me on my screen, the server did not register this hit for whatever reason, this also happened when i was at the back of b and got sniped.

These really were split second shot things and as surprised as i was i found it odd to be what seemed like an unrare ocurrance.

So A lot of the “hitbox lag” we experience is down to the server like a lot of people will tell you on this forum anyway, the actual code itself seems pretty good and its clear to me that those videos do things to make it all appear so much worse, they may of even done the same thing i had to, to get the hitboxes so far ahead.

On a couple of side notes I found a nice way to trick net_graph 3 to giving you an insane fps, into 7 figures sorta thing even negatives cl_interp -0.1.

I got the video hosted: http://download.stuff.is/index.php?dir=hl2/skizy/&file=csshitboxinvestigation.wmv

FPS Effect on Hitboxes/netcode
I tried 3 differant settings, 60 fps which ran like the videos above did because fraps would only capture 40 fps at a time and wouldnt go to 60 so i had to scrap that, so i went for 30 and 15 fps. My test on 30 fps had rather unexpected results, it didnt “feel” like it lagged at all, I did get the feeling it was slightly more than laggy than my normal fps (obviously) but it didnt really come into it, there wasn't really any effect on hitbox lag or netcode it just wasnt as smoot as it normally was.

There was a scout kill at 30 fps that when watching it again looked as if i had shot behind the model, this really was only noticable when i watched the movie back at the time it felt totally ‘legit’. Even thinking of it now it may look like the shot was behind the hitbox but i think i shot when the scope was over the model, i think theres a slight lag (slight being the word) between clicking, shot registering, scope unscoping, sound playing, blood splatting. This lag may be less than a second but when you got a model jumping across that short space through the door on dust 2 then thats all the time needed.

My most shocking result was 15 fps, it was jerky my player would warp back but still my shots registered and it felt fine just more laggy.

I'm not sure exactly why this was i have a theory (remember im not the most technically minded and all that) that my computer was running at the 80 fps but was only showing 15 fps right? well on a computer that actually ran at 15 fps there would be more performance issues and im not gonna go buy some old computer to try that out.

http://download.stuff.is/index.php?dir=hl2/skizy/&file=fpseffectsbloodchill.wmv

Gun Recoil
I investigated into the revoil patterns of the guns as well after I remember reading today someone said the recoil patterns for the guns are not always the same but have a few patterns or something, I went to test this out as i sprayed four clips into four differant spaces on a wall in dust 2 ct spawn, i have an optical mouse i turned it upside down so it didnt move one mm (i can guarantee that) and held the fire down, the general recoil pattern of the m4 both silenced and unsilenced were the same to the other reptitions with them both going up and the majority just ahead of the original shot, however it was not the same everytime and there are indeed slight variations. This is also true for the mp5. Me being the 1.6 lover i am went to test this out on 1.6 and low and behold i got the same results roughly the same but varying.

http://download.stuff.is/index.php?dir=hl2/skizy/&file=recoilclippingbloodchill.wmv

Player Clipping
This was something mentioned a looong time ago about how when running alongside or behind another player you would be jerky and warp back like lag sort of effect and how this should be fixed and was yet another example of how many flaws source has to 1.6. Not sure if you noticed this but it has actually got a lot better, im not sure if it was patched but i couldnt clip to the same degree as before and hardly jerked about, anyway i went onto source and 1.6 and was able to clip on both.

Your thoughts are welcome and I hope the video link works, its 3 minutes long the quality isnt great, I didn't mean the music to be on fraps just recorded all sound and i had music on at the time unwittingly, its simple ugly but its to the point and serves its purpose so dont judge the video like oh thats rubbish a baby could of done a better job!

Now my fingers hurt so Ill stop typing, hope someone reads or looks at the video.