Rubberbanding

Starrr - Harshlands wrote: »

+1

sad part is that it is not even a very difficult fix if the devs really wanted to.

I do not know about that being "not very difficult" but I imagine it is doable.

The sad part is that we get asked for details on where this happens. But they could easily log every time this happens. This logging could happen on the client (and be summarized periodically or when patches get downloaded, or whenever else), or on the server, or both. And they could get incredibly precise information this way, about what video card is being used, and where exactly the rubber banding is happening.

Anyways, I personally imagine that whoever they put in charge of dealing with this either does not really understand the issue or is not very interested in solving the issue, or something... something has been interfering not only with the solution of this problem but also with the investigation of this problem. (And, quite possibly, the issue is that the problem is so bad that they are overwhelmed with how bad it is.)

Granted not all rubber banding has been fixed, but the GMs have acknowledged the situation in this thread here: http://pwi-forum.perfectworld.com/showthread.php?t=846152 and I'm sure they have sent in reports to the Devs at PW-CN. You can tell by the previous few patches they have fixed some rubber-banding. Both these patches below have shown the Devs have taken the initiative to correct the problem. Hopefully tonight's patch should further address the rubber banding, but only time will tell.

Patch 409: http://pwi-forum.perfectworld.com/showthread.php?t=856202

Patch 415: http://pwi-forum.perfectworld.com/showthread.php?t=863552

TheDan - Sanctuary wrote: »

Granted not all rubber banding has been fixed, but the GMs have acknowledged the situation in this thread here: http://pwi-forum.perfectworld.com/showthread.php?t=846152 and I'm sure they have sent in reports to the Devs at PW-CN. You can tell by the previous few patches they have fixed some rubber-banding. Both these patches below have shown the Devs have taken the initiative to correct the problem. Hopefully tonight's patch should further address the rubber banding, but only time will tell.

Patch 409: http://pwi-forum.perfectworld.com/showthread.php?t=856202

Patch 415: http://pwi-forum.perfectworld.com/showthread.php?t=863552

If you ask me they are doing a horrible job because it has been consistently getting worse as time goes by.

The rubberbanding issue won't be fixed or at least the Devs are unwilling to fix the cause of it. I suspect it's due to the geolocation updates PWI had. When the game first started, you were able to drop from high altitude and hit the ground with no rubberbanding. After these types of updates started showing up is when your rubberbanding starts. However, Devs will not remove those patches. The main reason they had that patch is to prevent hacking and cheats in the game. Before the major geolocation patch (they had minor ones before) rubberbanding wasn't as bad. After that major patch, it all went to hell. At the same time all sorts of running, flying, etc.. cheats stopped working.

PS: I don't use these cheats so I am not sure if I have the terms right. I heard this from someone who tells me these things. But he was QQing to me how the patch messed up everything for him around the same time the massive rubberbanding got really bad.

TheDan - Sanctuary wrote: »

Granted not all rubber banding has been fixed, but the GMs have acknowledged the situation in this thread here: http://pwi-forum.perfectworld.com/showthread.php?t=846152 and I'm sure they have sent in reports to the Devs at PW-CN. You can tell by the previous few patches they have fixed some rubber-banding. Both these patches below have shown the Devs have taken the initiative to correct the problem. Hopefully tonight's patch should further address the rubber banding, but only time will tell.

Patch 409: http://pwi-forum.perfectworld.com/showthread.php?t=856202

Patch 415: http://pwi-forum.perfectworld.com/showthread.php?t=863552

Sorry to say but the rubber banding has actually gotten worse since the last few patches. Now just a normal jump anywhere will cause you to be stuck in the loop until you fly out of it. For the lower lvls with out a flier it can be a more irritating issue. With land mounts it is the same, jump an you band on even flat surfaces now. This issue has been around for over 2 years. It is time to get it corrected.

can u tell by all the other threads they start like tt changes and tw changes that they dont respond and they either dont care or are powerless to help. This game is fading fast and they are likely trying to get you to play forsaken world or another pwi product, ie no new game content, and fail patches and no issue fixing, just forums for us to qq while they laugh/count monies.

Baalzor - Heavens Tear wrote: »

i am no expert in this...why is this happening ? really looking forward in some answers on this.

i'm not an expert either, so everything i'm about to write is guesswork. moderately educated guesswork, but still.

the game client tracks how you want to move, and moves you around the 3d game world, showing you the results (you in a new position) as quickly as it can. every so often it tells the server, "i moved this player over to there" so that the rest of the game can know how you're moving. it's important that the game client can show you the results of your moves as quickly as possible, so the game action will be smooth to you, on your screen. it's also important that the game client should update the server often and well, so your squadmates can see you move naturally, smoothly, and without lag.

the game server gets updates from clients every so often, but it can't necessarily trust those. fiddling with the network traffic between the client and the server is one way of botting and cheating in multiuser games; wall hacking and in-instance flying would be trivial if the servers blindly trusted what the clients said about your movements. so the servers sanity-check your movements, and if something seems fishy, the server will... do something to set things right.

apparently, one commonly used "something" is to tell the client, "nuh-uh. last i knew for sure, that player was still over there. put 'em back and start over!" which, in the case of a wall ****, puts you back on the floor. and if the client and server for some reason disagree on how you ought to be able to move... rubberbands you.

the obvious solution is to use the exact same movement code in client and server, so they won't disagree (as much). i do not know why this isn't done (enough). there'd still be some disagreements, Murphy's law you know, but they shouldn't have to get as bad as what we're seeing.

the less obvious solution is to have the server calculate possible trajectories: "well, going from that last known position to where they claim to be now, through the walking/running/jumping they claim to have used --- closest they could have legitimately come would be there", and have the client warp you there instead. this would create a different, probably still noticeable and annoying but perhaps not as disruptive, phenomenon; you'd be getting zigzagged to spots near where you thought you were going, but not all the way back (usually). i do not know why this doesn't seem to be done. it'd put more load on the servers, true, but paying for CPU cycles damn well should be in the budget of any game that hopes to make money!

(the latter method doesn't seem to be done at all. i've rubberbanded while falling through empty air, happens a lot when autopathing to city of arrivals. server interpolation of a fall through empty air should never cause me to rubberband back up to the start of the fall, so this method seems not to be used.)

LenieClarke - Heavens Tear wrote: »

i'm not an expert either, so everything i'm about to write is guesswork. moderately educated guesswork, but still.

The company I used to work for wrote networking software for military sims. The clients are trusted there, but the concept of client/server sync is the same. (Military sims actually don't use a server - the clients all communicate with each other so it's client/client sync. The "world state" is just what's broadcast on the network.)

fiddling with the network traffic between the client and the server is one way of botting and cheating in multiuser games; wall hacking and in-instance flying would be trivial if the servers blindly trusted what the clients said about your movements. so the servers sanity-check your movements, and if something seems fishy, the server will... do something to set things right.

The cheap and easy way (costs less CPU time) is for the server to take the last known True client position. The client sends the new position. The server calculates the time and distance (including obstacles) between those two positions. If the distance is something the character could have covered legally in that amount of time, the move is allowed. If the character couldn't have covered it, then the move is not allowed and the client gets an instruction to go back to the last known True position. When your client gets that, it rubberbands back to a previous location.

As best as I can tell, there's only one code-related source of rubberbanding. When you drop from a height, at some point the speed at which your client calculates you falling seems to differ from the speed the server thinks you can fall. So the server rubberbands you because it thinks the client is doing something illegal when in fact it's just making you fall as the developers coded it to. I suspect they tweaked their falling velocity algorithm at some point, but only updated it on the client, forgetting to do so on the server. There also seems to be a problem when you combine falling with moving when you hit the ground - the server doesn't seem to think you're allowed to do that.

The other major source is not directly code-related. If there's network lag, the client's packets saying what it's doing may not arrive in a timely fashion. In a trusted environment like a military sim, each client just timestamps their position update packet and the timestamp is trusted by everyone. In a game, that could be exploited by someone modifying their client to put false timestamps, making the server think you've been running for a lot longer than you really have and thus allowing a greater position change.

So the server has to guess the time between position updates partly based on client timestamps, and partly based on when it received a packet. If you're suffering a lot of lag, your client may send a packing saying "Ok I've moved to point A" then 1.0 seconds later send another packet saying "Ok I've moved to point B". In a non-laggy environment, the server receives these two packets about 1.0 seconds apart, calculates that the distance change is allowable in that time, and allows it. But in a laggy environment, the first packet could get delayed. The two packets then arrive, say, 0.5 seconds apart, the server decides you've moved more quickly than you're allowed to, and rubber bands you.

the obvious solution is to use the exact same movement code in client and server, so they won't disagree (as much). i do not know why this isn't done (enough).

It's not done because the client does collision detection every frame as you move. It would take too much CPU time for the server to do that every tick (from the interval system, I suspect a tick in this game is 0.05 sec) for every player in the game. The server doesn't even bother doing collision detection for mobs and pets on certain items - they can run through doors, trees, and other stuff you have to run around.

You'll notice there's also some slop between where moving mobs actually stop and where they show up on your screen. Sometimes you'll kill a mob in one spot, and the loot will show up in a slightly different spot.

the less obvious solution is to have the server calculate possible trajectories: "well, going from that last known position to where they claim to be now, through the walking/running/jumping they claim to have used --- closest they could have legitimately come would be there", and have the client warp you there instead. this would create a different, probably still noticeable and annoying but perhaps not as disruptive, phenomenon; you'd be getting zigzagged to spots near where you thought you were going, but not all the way back (usually).

That's actually what military sims do. It's called dead reckoning. The client does true position calculations for itself, and dead reckoning calculations in parallel. All other clients do dead reckoning calculations for that client. When the deviation between the true position and dead reckoned position exceeded a certain threshold, the client sent out a true position update which all other clients then used to update their dead reckoned position of that client.

We had all sorts of sophisticated dead reckoning algorithms using position, first, second, and sometimes even third derivatives to try to maximize the time between true position updates. Some of them got pretty involved in the physics behind the vehicle (planes for example were dead reckoned based on a more accurate phugoid motion rather than a straight line).

Soul_knife - Dreamweaver wrote: »

Sorry to say but the rubber banding has actually gotten worse since the last few patches. Now just a normal jump anywhere will cause you to be stuck in the loop until you fly out of it. For the lower lvls with out a flier it can be a more irritating issue. With land mounts it is the same, jump an you band on even flat surfaces now. This issue has been around for over 2 years. It is time to get it corrected.

But dude, are you sure that it is that way?, cuz me, I don't have that much problems with rubberbanding, I only rubberband in certain situations, but not everywhere and everytime. Then there's that girl who sayd that she didn't even noticed rubberbanding!!!. If you're not exagerating, so this isn't a regular problem, I mean, the rubberbanding affects every person in a different way and intensity, which makes it even harder to know what is causing this rubberbanding! >.<

EDIT: This makes me think that the intensity of the problem may have something to do with every player's internet connection/PC's features/real life geographic location (since once I heard that the more away you live from the server's location, the worse your connection with the server is, no matter how fast yourinternet connection is)

MystiMonk - Sanctuary wrote: »

I don't have much of problem with this if you click on the map you are fine.I play another game where they just switched severs more of downgraded one and the rubberbanding in that is bad way more so than this game.I don't even notice it in this game.

This girl doesn't seem to have that much rubberbanding.