83 replies to this topic

[Renthrak]

Melcyna, on 01 November 2012 - 11:01 PM, said:

in a server based decission, both the victim and the shooter sees only the respective amount of latency to the server between what they see and what is actually happening.

Plus with it server based, you are immune to other players latency effect for the most part, all that matters is how much YOUR latency is to the server.

This is why I don't want to see things shifted client-side. It's not that the client can't handle the load. The problem is that there are 16 clients, but only 1 server. It doesn't seem useful to have 16 different computers arguing about what really happens, when they can just tell the server what each player does, and let the server decide how that comes together.

Melcyna, on 01 November 2012 - 11:01 PM, said:

So using server based decission for them is understandable, what's NOT ACCEPTABLE is how grossly out of sync their interpolation is which is supposed to reduce the latency effect on the desync between what is actually happening and what the players see on his client.

This is the real issue. Currently, the mechanisms to deal with the lag between client and server models of the game world are not functioning properly. Fortunately, the devs are listening to us on this. Here is an excerpt from the 'Known Issues' thread:

Quote

Hit Detection/Network Movement Code

We are rolling back some networking code fixes. This will address many of the hit box issues currently in the live environment. The rollback does not fix the problem entirely, and we are working hard on improving the overall experience vs performance. This is a long term engineering task and a top priority internally.

What I find most encouraging is the last sentence, stating that this is a TOP PRIORITY. This means, for us players, that we can rest assured that the issue will be fixed as quickly as PGI are able, which is all we could ask.

[GaussDragon]

Renthrak, on 02 November 2012 - 06:57 PM, said:

This is why I don't want to see things shifted client-side. It's not that the client can't handle the load. The problem is that there are 16 clients, but only 1 server. It doesn't seem useful to have 16 different computers arguing about what really happens, when they can just tell the server what each player does, and let the server decide how that comes together.

I'm not sure, but I think what you're thinking of is peer-to-peer. In a client-side game, 16 clients ARE telling the server what's happening instead of the other way around.

Renthrak, on 02 November 2012 - 06:57 PM, said:

What I find most encouraging is the last sentence, stating that this is a TOP PRIORITY. This means, for us players, that we can rest assured that the issue will be fixed as quickly as PGI are able, which is all we could ask.

Even with them doing a rollback, you're assuming it was good to begin with. I'm happy it's a top priority, and I wouldn't be surprised if F2P was a big reason they went server-side, since it's hard to ban accounts (read: credit cards) in a F2P game. I guess it comes down to a good old fashioned cost/benefit analysis, are the resources needed to keep up with the hackers worth making the game's performance noticeably better? Everyone is different. Personally, I can deal with hackers from time to time. They're usually pretty blatant. Having to change my lead times based on my fluctuating ping is a constant frustration. I don't really care what other people do, most of my frustrations are on my end. However, everyone prioritizes different things. As seen with this thread, my priorities are evident.

Edited by GaussDragon, 02 November 2012 - 09:20 PM.

[Melcyna]

they went server based technically because there isn't any other way really...

we're talking F2P games here... where players can come from anywhere between as close as US states (sub 100ms latency) to as far away as a continent several thousand km away with latency as high as 300-400ms.

deciding things client side for some things are possible even with high latency players like hitscan weapons, but for others they get incredibly messy as latency rises.

for example, if we have a projectile weapon and someone with a 500ms latency fired it at you, he may see in his client that you are within clear line of sight and thus should be hit, but 500ms + your latency to the server of say 50ms later... you obviously are not where you are as according to his client.

So how do we decide the projectile path? His client says you are right within the path of the shot and should be hit, your client shows something completely different (since his client has a massive latency)

Should the shot bend then? and hit you? Should the shot miss but still damage you? Should the shot miss and did no damage? ie: which client decission then do we take? If we take the shooter client decission always for weapon damage then you will end up being shot through impassable object and terrain according to your client

Unless if you only play with ppl with similarly low latency to the server... (ie: you play with absolutely no one but the ppl from your own continent) then the problem still exist but in much smaller scale, and easier to ignore.

Only problem with that? This is F2P game, there's no F2P game in existence that can survive by restricting the playerbase to a single continent (where latency is dependable).

Edited by Melcyna, 02 November 2012 - 11:44 PM.

[Tywald]

It doesn't have to be a perfect solution right now and i would be okay if i could just hit the darn Light mechs somewhere and do some damage when i point at them whit my lasers. I could be dancing around whit one and my lasers burn its cockpit but the hits are being registred on its arms and side torsos, At least i know that if i continue to shoot at the cockpit i'll sooner or later destroy the mech or force it away.

[Viper M40]

Yes, there is one solution. Open more servers in diferents places around the wolrd, and get people connecte to the nearest server. The problem will be there like in all online shooters, but having a server in your country makes the problem very small.

Greetings

[Melcyna]

True, and players in continents other than US will ask for it too...

Each server cluster however will cost money... so it comes down to how much resource the dev or the publisher have, and where they intend to spend that resources, and if there is enough player base in those region to benefit from the nearer server and if the profit from it will offset the extra cost.

Note as well of course that the cost varies, for example a server cluster of the same spec capacity cost more in Australia than US... so maintaining one in Australia will yield smaller profit per player served than say US etc..

It is also usually cheaper to deploy more server in the same location and continent than it is to deploy another one elsewhere, since expanding the capacity of existing server cluster tend to be much simpler (in terms of hardware and procedure). This is why most games tend to concentrate their server in as few locations as possible.

Tywald, on 03 November 2012 - 12:38 AM, said:

It doesn't have to be a perfect solution right now and i would be okay if i could just hit the darn Light mechs somewhere and do some damage when i point at them whit my lasers. I could be dancing around whit one and my lasers burn its cockpit but the hits are being registred on its arms and side torsos, At least i know that if i continue to shoot at the cockpit i'll sooner or later destroy the mech or force it away.

This is what interpolation is supposed to help us with...

Most games generally have a decent interpolation so even with significant latency (but not too massive as there's a limit from which interpolation cannot handle beyond it) what you see should be a pretty close approximation of what is actually happening.

HOWEVER, PGI botched it big time with 25th oct patch and made it worse... it wasn't good to start with pre 25th oct patch, but after the patch it was WORSE.

They then reversed it on 31st so now we're just waiting on what the next iteration's netcode and interpolation is going to be like... IF they are any decent, it should be at least sufficient to allow us to hit the lights reliably.

[Lumanil]

DaZur, on 30 October 2012 - 05:07 AM, said:

Food for thought... and sadly this can only qualified by the devs via a a damage audit or some seriously coordinated in-game testing methodology.
'
Right now everyone is pointing to damage not registering on the target mech. Is it possible it's the other way around? I.e... The damage is actually registering on the target mech but the paper doll indicator is not syncing with the actual damage inflicted.

I say this because I cored an assault mech last night with a AC20 hit that as far as the paper doll was concerned, on a completely fresh / undamaged CT.

I tought about this too.
One day i've seen a mech, which lost its right arm only to see on the paper doll that it was still "red" and only it's shoulder was destroyed.

[Fugu]

It has gotten way worse since last patch.
Currently I often feel like my Catapult weighs 200 tons. When I move I can actually ignore almost all balistic incoming fire as it almost never hits me. Even a Jenner can be ignored for a pretty long time because the lasers only hit me in random locations.

On the other hand I have to lead the lasers on a Jenner like nobodys business. Aiming 20m into the direction he is heading and still just barely scratching his arm is not unusual.

Especially new players seem to get frustrated about that. I shouldn't be able to use a Catapult C1 as a fast assault, seeming indestructable just because new players aren't used to the extreme discrepancy between where they see me and where I actually am.

[Melcyna]

Indeed, so let's hope PGI can sort out the network code and interpolation ASAP, before even more prospective players are lost...

This open beta frankly i think did more harm than good overall given the state of the game as it is, and how it must've looked to ppl who checked it out. Their first impression can't be good the way it looked on 25th Oct condition... not that it's much better prior, or now (since they reverted it to pre 25th Oct a few days ago)... but i can imagine what the new players would think seeing their shots not registering, not pretty would be an understatement.

Edited by Melcyna, 03 November 2012 - 04:54 AM.

[Cest7]

They need to make it so that mechs only glow when the server regs damage on them

[Targetloc]

I really hope the network code fixes address the fire delay on ballistics. I played a few rounds today where my AC shots were firing off 45 degrees behind where my reticle was pointing.

I'm used to lag shooting from Mech2, but at one point I was aiming 3 full mech lengths ahead of a Raven and hist were still registering on his trailing arm.... sigh.

[Lugh]

Lumanil, on 03 November 2012 - 02:56 AM, said:

I tought about this too.
One day i've seen a mech, which lost its right arm only to see on the paper doll that it was still "red" and only it's shoulder was destroyed.

There was a single hit point of damage left in that torso. running into him would have taken it away, a MG would take it away, etc.

We left a Centurion with NO legs upright a couple nights ago the same way. He stopped moving because he didn't want any damage to his legs to kill himself.

[Wireball]

Cest7, on 03 November 2012 - 06:38 AM, said:

They need to make it so that mechs only glow when the server regs damage on them

Direct visual feedback definitely helps. In closed beta when mechs were falling down and then teleporting back and forth as they stood up, I was able to figure out where the mech was and actually do damage by sweeping my lasers around until I saw them impacting something - even though it was only empty space part of the time (I checked the paper doll a few times to make sure it was having an effect).

That being said, if the delay between hitting and glowing was more than about 100ms (just guessing), players wouldn't necessarily be able to correlate their hits with the glowing. Some play testing would be necessary to see how much delay there could be before the association was no longer clear between having fired the lasers say, 500ms ago, and the armor glowing red.

Edited by Wireball, 03 November 2012 - 01:54 PM.

[Kaptain]

Also experiencing strange hit detection. Prior to the latest patch I only had zero damage bugs when mechs were powered down... Now it happens more often and in other situations.

I love this game, but firing delays, hit delays and hit detection have always been my biggest gripe.

[ZivyTerc]

Same here - firing at CT of overheated Centurion, no hits registered. Firing at stationary Raven from 20 meters, no hits registered. Extremely frustrating.

[Connatic]

I was wondering why I've felt invulnerable lately in my Raven. I have played since closed beta.

[Renthrak]

GaussDragon, on 02 November 2012 - 09:20 PM, said:

I'm not sure, but I think what you're thinking of is peer-to-peer. In a client-side game, 16 clients ARE telling the server what's happening instead of the other way around.

What I mean is: Client tells Server, "Player changed throttle to 90%, turned 39 degrees, torso twist at 51 degrees, fire command for weapon 1, 2, and 4, at vector . . ." vs. "Player moving at 135kph, target lock on Player 3, fire missiles, Player 3 hit".

In the first case, it's up to the server to decide how the player's 'Mech responds to the input, and whether the weapons hit anything when fired in that direction. That is preferable to letting the client determine how the 'Mech responds to input, or whether any shots will hit.

[Wireball]

In spectator mode, one can observe players consistently firing somewhat behind mechs running circles around them. It's possible to see what's "officially" going on. I guess that makes sense, since the server is sending the spectator what it thinks the relative positions of the mechs are.

In some first person shooters, if you have high latency, there's a noticeable delay between pulling the trigger and one's weapon actually firing. I can't remember specifically, but I think Quake 2 or Quake 3 behaved like that? It can be annoying, but at least you can see what's actually going on. I wonder if it might make more sense to have Mechwarrior behave that way - it's kind of misleading to get instant feedback when in fact your weapons are taking a tenth of a second before firing at where the enemy was.

[LarryDaBird]

Here's the problem, serverside hit detection. Yes it prevents aimbots but there is a reason that all of the successful shooters mostly use clientside. Because I would rather be hit once in a while while I am entering cover than miss a target that I obviously just drilled with a 50 point alpha....

[Wireball]

LarryDaBird, on 05 November 2012 - 10:21 PM, said:

Here's the problem, serverside hit detection. Yes it prevents aimbots but there is a reason that all of the successful shooters mostly use clientside. Because I would rather be hit once in a while while I am entering cover than miss a target that I obviously just drilled with a 50 point alpha....

I've often wondered why aimbot prevention has to be done in realtime. Why not allow it, but cross-check and compare to see when it's happening? It should be possible to detect clients that are sending data that's significantly deviant from other clients and flag those players.

E.g. Client sent us data saying they were doing this. Then they hit this target over here. Calculating...at the rate of turn of this mech, there's no way they could have brought their weapons to bear. Aimbot detection tripped - no credit for match. Or flag player for review.