|
Post by fatclemenza on Nov 20, 2012 18:28:16 GMT -5
Hey all, haven't made a post on this account in ages (maybe ever, but I've had it for a while). I'd like to talk about an observation I made regarding killcams. What better place to do it, right?
Pre-MW3, the connection model was that the host of the lobby had a perfect connection (obviously), and every other player was a certain amount of time behind him. Killcams in this case were very inaccurate, projecting what would be seen as if the host was the killer, which was rarely indicative of what happened when two other players were involved.
The new model eliminates the "host-on-top" model and instead aims to give as many players as possible 4 bars. In order to do this, sacrifices must be made to even out connection differences, aka the new form of lag comp we all hate to death.
In this model, killcams also can't be the same since the host isn't the pinnacle of the lobby anymore. What is now be displayed may truly be what the other player sees. If a killcam shows you not firing as many shots as you really did, or shooting the completely wrong location when in reality you were clearly on target, you're on the bad end of the lag comp. If it shows you actually doing what you were doing, then the gameplay error is on your end, and the killcam should reflect that.
What is the point of all this? If killcams are actually accurate now, then they display the clear problems created by this model of lag comp. You can see how badly your play is affected by it.
Does anybody else agree that things may now work this way? I'd like to hear some thoughts. Please don't hesitate to correct me if my information is inaccurate. There are all just my observations.
|
|
toysrme
True Bro
"Even at normal Health, there's no other choice than the Vector" Den Kirson
Posts: 1,339
|
Post by toysrme on Nov 20, 2012 18:47:28 GMT -5
Pre-MW3, the connection model was that the host of the lobby had a perfect connection (obviously), and every other player was a certain amount of time behind him. incorrect. on many accounts. so many i quit reading. all systems, including the host system, is running their own world. they can be any point faster or slower than any other system. COD works by timestamps. all actions are reported with timestamps. the host system simply buffers everything that happens over a set period of time and looks for the earliest time stamp where someone has died. also, hosts have their own lag. typically this is reported as an additional 30-40ms applied to yourself anytime you are host. this is also true on console. (go run a jtag/RGH or burn a hacked W@W iso that enables the ping counters on console) there is no "host-on-top" model. there never has been. cod4/5 run an unknown logic to determine host. from mw2 on wards your system polls the activision server for the first 50 results which reply in 50ms. it continues refreshing at 10ms intervals until it finds a suitable connection. MW2 instituted an upload bandwidth race where-by any time a new host must be chosen (host migration), all clients run a bandwidth test to the activision server and the highest bandwidth is chosen. BO-1 quickened and simplified this by running the test as soon as you enter the MP client portion of the game and stores the results. with MW2 is also became known what forces a host migration to occure. if: 51% of the clients (minimum of 3) in the game report a ping at or above 400ms for 300 network frames (the default network rate is 20 per second). the the game re-chooses host. authServerName "cod4master.activision.com" badhost_debug "0" badhost_endGameIfISuck "1" badhost_maxDoISuckFrames "300" badhost_maxHappyPingTime "400" badhost_minPercentClientsUnhappyToSuck "0.51" badhost_minTotalClientsForHappyTest "3" band_12players "384000" band_18players "580000" band_2players "64000" band_4players "128000" band_8players "256000" bandwidthtest_announceinterval "200" bandwidthtest_duration "500" bandwidthtest_enable "1" bandwidthtest_fudge "1" bandwidthtest_ingame_enable "1" bandwidthtest_ingame_fudge "1" bandwidthtest_maxpacketsize "1000" bandwidthtest_quota "0.5" bandwidthtest_rounds "10" bandwidthtest_scalar "2" bandwidthtest_timeout "1500"
|
|
toysrme
True Bro
"Even at normal Health, there's no other choice than the Vector" Den Kirson
Posts: 1,339
|
Post by toysrme on Nov 20, 2012 18:50:31 GMT -5
not only that but killcams have never and will never be accurate. COD is a game that runs dozens of frames per second on each local system. "killcams" and "theater" are nothing more than demo file recordings which are done by default at MOST 20 states/second.
you can easily see this by watching an erratic, extremely high sensitivity & quick paced XIM mouse user on console. Demo recordings make them seem "normal" with few extrordinary movements & shots. where real life recordings can show extremely quick paced play. Faster than the majority of PC players...
|
|
|
Post by fatclemenza on Nov 20, 2012 19:30:23 GMT -5
Wow, I never knew any of that. Very nice information though, at least now I have a basic understanding of how things work. Thanks.
|
|
|
Post by corpsecreate on Nov 20, 2012 20:01:19 GMT -5
Lag Compensation has nothing to do with "evening out connections" whatsoever. Read this: developer.valvesoftware.com/wiki/Source_Multiplayer_Networking#Lag_compensationin theater/kill cams, the game shows you what the killer saw except there is one difference. The position of the person that died is NOT the same position the killer saw it happen. The killer always sees the player behind where they truly are (the result of lag, not lag compensation). Idealy to fix this problem, the game would have to keep track of the latency between players A (killer) and B and then in the killcam, show player A's action as they are performed on a player B that has been rewinded back in time by an amount of time equal to the amount of latency. This would result in a far more accurate killcam. Others know more about theater mode than me but there is one thing I can be sure of. Videos replayed in theater are not actual 'video clips' per say. They are a replayed version in real time of that particular game where there is no lag. See my Live Gameplay vs Theater video: I synced up the client sided inputs. You will notice that in theater mode, the enemies are always further forward than they are in the live gameplay.
|
|
GoHarvard
True Bro
It's not that serious.
Posts: 390
|
Post by GoHarvard on Nov 21, 2012 0:09:16 GMT -5
There was a scenario where I had just killed 2 players in a building, called in my UAV and noticed on the mini map that someone was rushing the building to come get me. I laid down and ADS'd the hall way, saw the guy in a reload animation and started firing at his face...but no kill. He ended up killing me without taking a single bullet.
On the kill cam he had already finished reloading 4-5 steps before approaching the corner and saw my first few bullets being fired, he strafed out and killed me. My bullets were all missing to his right, which was the direction he was strafing to. His model was obviously too far away from his real hit box.
Both of us had 4 bars, I noticed an 818 in his clan tag which is an area code in the Los Angeles area close to me.
In MW3 or BO1 I wouldn't experience latency this bad when playing against local players. This is why I think there's more to it than ping times. The guy wasn't even having a good round, he was in the negative (it was TDM). I was thinking to myself if the lag was benefiting him that much, he surely would have been playing better. Makes me think the lag compensation is more dynamic this time compared to before where the 'sense of latency' was the same throughout the match.
|
|
bluespark
True Bro
formerly DS-Hunter
Posts: 159
|
Post by bluespark on Nov 21, 2012 11:02:02 GMT -5
in theater/kill cams, the game shows you what the killer saw except there is one difference. The position of the person that died is NOT the same position the killer saw it happen. The killer always sees the player behind where they truly are (the result of lag, not lag compensation). That would explain why I constantly see the enemy shooting into the air in the kill-cam while I move behind cover and then drop dead. I guess knowing this doesn't make it any less aggravating, but at least I feel educated now .
|
|
|
Post by generalawesomo on Nov 22, 2012 0:34:54 GMT -5
I wonder. If you think that you fired your gun, and the killcam showed that you didn't fire your gun. Can you pick up your gun that you dropped when you died and see if it was fired or not in the first place? To check if it has a full mag in it or not.
|
|