User panel stuff on forum
  324 posts on 11 pages  First page12345678910Last page
General Discussion
2010-04-30, 16:06
Administrator
334 posts

Registered:
Jan 2006
Medar wrote:
ParadokS wrote:
If 60ms vs 13ms player get into shaft fight, they both use 13ms shaft right? that sounds fair!... Except, what happens now when the lag is moved from high ping players shaft, to low ping players movement?

Using the word movement here is very wrong. It's transfered from how the shooter sees the shot to how the other players see the shot.

ParadokS wrote:
But for the low ping player he sees the enemy's beam just flying around and not hitting him

With 60 ms player vs. 13 ms that's really not an accurate description of what's happening. The shaft been is in a different position, but not with such a huge margin.

ParadokS wrote:
For 60ms player he can just aim directly at opponent and he knows he will hit, and when he sees enemys shaft etc it's the real deal.

You are conveniently forgetting couple of key facts here. The 60ms player has... well 60ms ping. So when he finally sees your "real deal" shaft it's not the real deal for him any more than his shaft is real deal to you. Secondly, when he starts to react to that shaft beam that he sees, he still has that 60 ms ping and his reactions will be delayed.

To put it simply: In 13ms vs. 60ms shaft on shaft fight the 13ms player holds the advantage when it comes to dodging the enemy shaft.

And movement is based on what you see, so changing what you see, is the same as changing movement or how physics work in my world.
ready!
2010-04-30, 16:09
Member
284 posts

Registered:
Oct 2006
Heh, nice to read these threads now. I didn't know this was enabled and had been wondering about my aim during 2010, but since I've had a lot of issues with my computers, I've simply put it up to those problems. I've certainly been changing my aiming, movement and tactics when I've played with a higher ping, not that I really feel jaded about it even when I'm finding out this late. But I do sympathize with players who've had more important games.

Mostly I had noticed a difference in sg aim, and at times it felt like I hit LG better in 4on4 games, but since it didn't feel consistent, I thought it was just me getting lucky. Because I spec a lot, I had also noticed some kills that seemed a bit weird, but I didn't really think anything more of it. In povdmm4 games and 2on2 dm4 my shaft has seemed a bit weird, again hitting at times when I didn't think I should've and then missing at times when I've predicted movement. I use fakeshaft 0.8.

Ironing out the kinks is an important issue of focus now and I fully support cutting down the benefits of antilag to a degree so as to not change the game too much for those with a higher ping. On thing that really does bother me is this dodging + shaft beam + blood issue. It's simply impossible to make it right for both players (one lagged, one 13ms). Even if I have the advantage at 13ms, I'm still being hit in a sense after the fact and this delay increases with my opponents lag right? I try to use a lot of jump escaping, especially on povdmm4 (when I'm in a bad spot and getting shafted, I jump and let the shaft give me speed and just bunny to a safer position or use the speed to circle strafe jump attack on the opponent). Now I'm getting hit a lot more even after I've done my escape :/

Like so many others, I really like the idea, but I think the most important thing now is to find the balance. And this balance should be very clearly defined in terms of what it means in game, so that people know what to expect.
2010-04-30, 22:58
Administrator
1025 posts

Registered:
Apr 2006
Question is: If this works as good as intended to do, why is people still complaining? :E

Do a fucking poll about it and ask if players wan't this or not, how hard can it be.

Creds to Medar who is just trying to fix broken stuff.
2010-04-30, 23:53
Member
386 posts

Registered:
Apr 2006
fog wrote:
Question is: If this works as good as intended to do, why is people still complaining? :E

There has never been a single thing in quakeworld's long history on which you can get all of its players to agree. Even after all this time there are still people who object to fullbright skins, gl clients, pings being too low for everyone (I'm fucking serious. I heard someone complain about this three days ago) and mm3. Can you conceive of a single feature which would not make people object, even if they had to create imaginary problems in order to do so?

fog wrote:
Do a fucking poll about it and ask if players wan't this or not, how hard can it be.

It's too soon for a decisive poll, surely, since the current implementation of antilag is little over two days old and people are still too busy yelling about antilag in general for any reasonable vote to take place, partly due to the sneaky way it was tested. Wait until the general population has been educated and have experienced it for themselves before they reject it due to imaginary side-effects.

fog wrote:
Creds to Medar who is just trying to fix broken stuff.

Word, Medar. Represent.
2010-05-01, 00:12
Moderator
1329 posts

Registered:
Apr 2006
Here are the results obtained by using an aimbot:

- Aimbot always aims directly to the center of the opponent
- A special test weapon was used, it shoots 32 pellets (one pellet per quake-unit) and the weapon does not have any spread (this means the shot is as wide from point blank as it is over any distance). The wideness of the shot is a bit less than player hitbox.
- The test server had "normal" movement speed of 320 enabled, and 700 for "faster movement"
- There were two ranges tested, 200 ("shortest distance" and 550 ("longer distance"

In the test, one player stands still firing the weapon using an aimbot. The other player circles around the other player using automated movement. This way no human factor will be present, however some randomness due to whatever circumstances is still there.


The results


From the methods you can see, that aiming directly at the center of the player model, the unpatched antilag was not even close to 99% accurate on higher pings.


During testing, we tested LG too with antilag 0 (disabled). If you aim directly at the center of the model with LG (aimbot was used to used to ignore the human factor), you will get 99% accuracy on pings 12ms, 25ms and 38ms. 51ms has only 60% accuracy, 64ms hits randomly, if ever. This test was taken using the "shortest distance" circling method and serverside maxspeed of 320.
Servers: Troopers
2010-05-01, 07:59
Member
8 posts

Registered:
Sep 2007
Renzo wrote:
Here are the results obtained by using an aimbot:

- Aimbot always aims directly to the center of the opponent
- A special test weapon was used, it shoots 32 pellets (one pellet per quake-unit) and the weapon does not have any spread (this means the shot is as wide from point blank as it is over any distance). The wideness of the shot is a bit less than player hitbox.
- The test server had "normal" movement speed of 320 enabled, and 700 for "faster movement"
- There were two ranges tested, 200 ("shortest distance" and 550 ("longer distance"

In the test, one player stands still firing the weapon using an aimbot. The other player circles around the other player using automated movement. This way no human factor will be present, however some randomness due to whatever circumstances is still there.


The results


From the methods you can see, that aiming directly at the center of the player model, the unpatched antilag was not even close to 99% accurate on higher pings.


During testing, we tested LG too with antilag 0 (disabled). If you aim directly at the center of the model with LG (aimbot was used to used to ignore the human factor), you will get 99% accuracy on pings 12ms, 25ms and 38ms. 51ms has only 60% accuracy, 64ms hits randomly, if ever. This test was taken using the "shortest distance" circling method and serverside maxspeed of 320.

Thats a test with aimbot on automated player movement. It doesnt tell us jack shit... In the real (quake)world, in a 1on1 situation you have 2 players fighting. So what are theese guys doing?
Well they do the following:
1. Move around to get strategical advantages of different positions based on where opponent currently is and how opponent is aiming.
2. Predict opponents movements
3. Shoot at predicted locations

With this antilag "function" we have a very very bad situation when we have one 13ms player and one 51ms player. The 13ms player suffers from point 1, because he doesnt always notice being hit, hence his movement becomes worse, which in turn raises the efficiency of the other player. So a person who in 13vs13ms situation that would have say 37% LG efficiency will now, with antilag present, have 40-43% efficiency due to the real 13ms player being forced to change his movement patterns to less optimal movement patterns.

So when the 13ms player suffers from what i previously mentioned, it will start a chain reaction where because he gets frustrated about not being able to move away properly, his focus will drop and he wont predict the opponents movements (point 2) as well as he would normally. Leading to him not aiming as well (point 3), dropping his own aim efficiency.

The way i see it, antilag with its current implementation is not a win-win situation.
2010-05-01, 08:52
Member
8 posts

Registered:
Sep 2007
On top of that everyone knows that its already a pain in the ass to hit players lagging around on the map...
2010-05-01, 09:52
Moderator
1329 posts

Registered:
Apr 2006
javve wrote:
It doesnt tell us jack shit...

That test tells you everything about the antilag precision with hitscan weapons.

It was intentionally done so that no human error will be a factor. The test shows that if you want to hit >99% with that special weapon built for specifically testing this, you always need to aim in the center of the model. If you aim on the edge of the model, you will hit only 50% of the pellets (since the shot is wide).

The base accuracy for >99% seems to be 13ms and 25ms without antilag at all, so what we want to see is antilag scoring that high hit percentages with any given ping. As you can see, the antilag without prediction fix does not actually hit properly when you aim directly in the center of the player.


As for the situation you describe, I agree partially.


If there is a situation of 13ms vs 51ms. Let's say this 13ms player can hit 40% with lg on that 51ms player. There is no antilag enabled, the 51ms player suffers from lag-effects when he gets hit, and can even warp a little.

If both players have same 51ms and antilag is enabled, the situation remains the same. The player being hit by 40% lg will suffer the same way as it did before, since the hitrate is as good.

If the players have 51ms and 13ms ping, and the 51ms player is the one hitting 40% with lg, the situation changes. The 13ms player should not suffer from the effects that come with 51ms when he is being hit, it should feel pretty much the same as someone with 13ms was shafting another 13ms player.


That being said, you are right about the visual representation. It looks like (on your screen) that the opponent shafting you at high ping does not hit, since the bolt lags, while in reality you could be taking damage. If you add some sort of damage indicator (like flashes to screen) you would notice the hits better, but this way you'd be giving out some clarity of the screen away (increased distractions from flashing).


But the thing is, the most important part is to make the antilag safe in a way, that it won't allow hits if your hitscans can't be kept on top of the visible model, the thing we tested for last night. This way you can't aim "behind" the moving object and still hit, even if you should not.


And the last thing, there should never be a situation like 12ms vs 51ms, or even 38ms vs 51ms, even if you had antilag enabled. This is because 51ms is the first ping in qw that makes your movement suffer a lot from hits. You can also no longer turn properly after teles (aero middle-tele turn is a good example).
Servers: Troopers
2010-05-01, 10:03
Member
130 posts

Registered:
Jan 2006
javve wrote:
Thats a test with aimbot on automated player movement. It doesnt tell us jack shit... In the real (quake)world, in a 1on1 situation you have 2 players fighting.

Obviously this test wasn't designed to test those effects, but to confirm that the technical implementation works as it's supposed to.

javve wrote:
With this antilag "function" we have a very very bad situation when we have one 13ms player and one 51ms player. The 13ms player suffers from point 1, because he doesnt always notice being hit, hence his movement becomes worse, which in turn raises the efficiency of the other player.

I do not believe this is the case at all against a 51 ms player. The distances in question are simply way too small to throw the 13 ms player off. I do encourage you to test for yourself on a server with the new patch.

javve wrote:
So a person who in 13vs13ms situation that would have say 37% LG efficiency will now, with antilag present, have 40-43% efficiency due to the real 13ms player being forced to change his movement patterns to less optimal movement patterns.

This is just not true. First of all like I said you exaggerate the changes a lot. Second of all you forget that the 51 ms player won't have 13 ms conditions, he will just be able to aim to same way compared to what he sees on his screen as he would on 13 ms. He doesn't get any more "right view" about the enemy's shaft than the enemy gets of his shaft. He's 51 ms ping is still there and all his decisions and dodges are lagged.

This is the same thing I explained a few posts earlier while replying to Paradoks: In any LG vs. LG fight the lower pinging player can still always dodge better.

There have been no signs of people aiming better with higher pings than lower pings with the new patch.
There have been no signs of people aiming worse against higher pinging players than lower pinging players with the new patch.
I see no theoretical reasons why we would see either of these effects.
2010-05-01, 10:27
Member
8 posts

Registered:
Sep 2007
I agree with what you say on the technical parts. And as ive said earlier, this is perfect in theory. But imo you fail to see the human factor...

Humans will alter movement patterns without even thinking about it, its something we do without stopping to even think about it. We dont re-evaluate how to move every time we get into LG range with the same opponent. This will happen first encounter with the high pinged person, and then we will contine playing with changed movement patterns troughout that game. Because our brains tells us to without any conscious thought behind it from our own parts.

Worse movement patterns leads to you getting buttraped in qw. Thats the part i dont like about antilag. And if im not entirely mistaken, that is what paradoks refers to as "transfering lag" from high ping player to low ping player, this to even out the playing field. We all know that this function does not transfer lag. But with your brains, you should get my point here.
2010-05-01, 10:34
Member
130 posts

Registered:
Jan 2006
javve wrote:
Humans will alter movement patterns without even thinking about it, its something we do without stopping to even think about it. We dont re-evaluate how to move every time we get into LG range with the same opponent. This will happen first encounter with the high pinged person, and then we will contine playing with changed movement patterns troughout that game. Because our brains tells us to without any conscious thought behind it from our own parts.

That's quite an extraordinary theory.

I do of course agree that at some point your ability to dodge the high pinging player's shaft gets noticeably worse. I do not however think that the point is at 51 ms at all.
2010-05-01, 10:39
Member
8 posts

Registered:
Sep 2007
Medar wrote:
javve wrote:
Humans will alter movement patterns without even thinking about it, its something we do without stopping to even think about it. We dont re-evaluate how to move every time we get into LG range with the same opponent. This will happen first encounter with the high pinged person, and then we will contine playing with changed movement patterns troughout that game. Because our brains tells us to without any conscious thought behind it from our own parts.

That's quite an extraordinary theory.

I do of course agree that at some point your ability to dodge the high pinging player's shaft gets noticeably worse. I do not however think that the point is at 51 ms at all.

Its more about the techincal implementation mate. The clients should get the same coordinates of where the players are. Not individual coordinates based on the players pings. Because at time xx:xx:xx if im doing something, and person Y can attack me on the position i was at during time xx:xx:01 , it will give him advantage over me because at that exact moment in time, i might think that im behind a wall. That i just managed to get behind it and dodge him. Or i decide to dodge left instead of right? Maybe now u see why some people see this as a transfer of lag?
2010-05-01, 11:23
Member
386 posts

Registered:
Apr 2006
javve wrote:
Its more about the techincal implementation mate. The clients should get the same coordinates of where the players are. Not individual coordinates based on the players pings.

This feature in no way alters the positional data sent to the clients. Everyone still sees the same thing.

Also, isn't this kind of theorising about dodging patterns pointless when you can easily test it in practice? This is something I have done and have noticed no differences in my accuracy between 19ms without antilag and 51ms with the current antilag implementation (except perhaps that it's harder to pin people coming from povdmm4 high spawn on the higher ping, but that's not a reliable observation due to there being too many variables involved).

If you can come up with some consistent results proving that antilag on 51ms gives you an advantage due to the enemy's altered perception, then perhaps people will reconsider, but there is no current evidence to suggest that this is the case.
2010-05-01, 11:26
Member
130 posts

Registered:
Jan 2006
javve wrote:
Its more about the techincal implementation mate. The clients should get the same coordinates of where the players are. Not individual coordinates based on the players pings. Because at time xx:xx:xx if im doing something, and person Y can attack me on the position i was at during time xx:xx:01 , it will give him advantage over me because at that exact moment in time, i might think that im behind a wall. That i just managed to get behind it and dodge him. Or i decide to dodge left instead of right? Maybe now u see why some people see this as a transfer of lag?

Of course it's a "transfer of lag". If you read my explanations about it, that's exactly how I paint it. The position where one player needs to shoot changes at the expense of the position where other players see the shot land.

In my blog entry and in these threads I've given extensive explanations about why situations you describe are not problematic in practice. I've also explained why the higher pinging player doesn't get an advantage over the lower pinging one.

QW without antilag and QW with antilag are two technical solutions to the problem of having to deal with latencies on the internet. I'm convinced that the problems that exist with antilag implementation are far smaller than those that exist without it.
2010-05-01, 11:28
Member
8 posts

Registered:
Sep 2007
Stev wrote:
javve wrote:
Its more about the techincal implementation mate. The clients should get the same coordinates of where the players are. Not individual coordinates based on the players pings.

This feature in no way alters the positional data sent to the clients. Everyone still sees the same thing.

If that is true, then ive been misinformed and people can ignore my posts.
2010-05-01, 11:36
Member
386 posts

Registered:
Apr 2006
javve wrote:
If that is true, then ive been misinformed and people can ignore my posts.

Well, perhaps not. What I was saying is that the only change is the physical location of the hit. All positional and graphical data outside of the actual damage and the resulting blood spurt remains identical. I think you do actually have a point, but I think your point only applies to much higher pings, and I think even then that you overestimate its importance.
2010-05-01, 12:10
Member
78 posts

Registered:
Aug 2006
I have a hypothetical question: Is it possible to drop an RL pack in a 4on4 game after having been convinced that you escaped around the corner and therefore start shooting with RL again just to find out that you are getting hit and killed anyway?
2010-05-01, 12:22
Member
130 posts

Registered:
Jan 2006
LethalWiz wrote:
I have a hypothetical question: Is it possible to drop an RL pack in a 4on4 game after having been convinced that you escaped around the corner and therefore start shooting with RL again just to find out that you are getting hit and killed anyway?

No, unless your opponent has some extremely silly ping like 900 ms. And even then that would be really a freak occurrence.

The time difference is just way too small for that to happen.
2010-05-01, 12:30
Member
386 posts

Registered:
Apr 2006
A quick note on human reaction times: Times of 50ms may affect your ability to synchronise visual data on-screen when it comes to aiming because you see crosshairs overlapping models and later make a decision that your current trajectory will create a hit based on this, but those times are far too short to affect actual decision-making.
2010-05-01, 14:15
Administrator
334 posts

Registered:
Jan 2006
well since it's not hard to feel the difference from 39ms to 52ms based on visual input (shaft lagging and increased delay in rl/gl), it should be safe to assume that within the 52ms it actually is NOT too small ping to effect or make you change decsions, if you base visual input.

People are the best "computers" to recognise and identify patterns, don't underestimate that
But as I said, from 13ms-52ms the antilag feels and looks like it works as intended, except for the visual part which could do with a tweak.
My suggestion to that was that if client/server knows about antilag being enabled, it should be possible to do a cl_fakeshaft 2 or something, making rendering more accurate visually.
ready!
2010-05-01, 15:03
Member
485 posts

Registered:
Feb 2006
In a case where you either react in time or not, any amount of time can be enough to change your reaction. It just becomes increasingly less likely to occur, smaller the lag.

So the question is if will this happen often enough to worry about.
2010-05-01, 15:07
Administrator
334 posts

Registered:
Jan 2006
I remember back in the days when your JUMPING was affected by lag.. with todays client you can jump around with ping 100 as if it was LAN.. noone seemd to mind that "fix"
ready!
2010-05-01, 15:09
Member
386 posts

Registered:
Apr 2006
No, no, you don't react to the visual input instantaneously; You recognise it then react to it considerably later. If the image changes between the time you see it and the time you react to it, your brain cannot act fast enough to react to the new stimuli. There are physical limits to the speed at which the machine that is the human brain can work.

For an example on the limitations of human reactions, take the Olympic 100m sprint. Any sprinter that leaves the blocks within 100ms of the starting signal is deemed to have made a false start, and even that is being incredibly conservative considering the dozens of other variables at play, such as the sprinter's distance from the speaker.
2010-05-01, 15:56
Administrator
334 posts

Registered:
Jan 2006
But its not about reacting to every single cell for example, its about reacting to a period of time, wherein you see a pattern of shaft moving or enemy moving.
ready!
2010-05-01, 16:27
Member
386 posts

Registered:
Apr 2006
Exactly, which is why such small periods of time don't apply to decisions made outside of consistent, predictable patterns, as in lethalwiz's hypothetical situation.
2010-05-01, 17:04
Member
485 posts

Registered:
Feb 2006
A time advantage doesn't disappear just because everyone reacts later.

A taller sprinter will have a disadvantage in every start, because it takes longer for the nerve impulse from the brain to reach his leg muscles and his ear will be further away from the loud speaker.


Okay, lets try a QW example.

Player with low health and RL needs to decide if he will next shoot a rocket and risk dropping a pack. He is getting peppered by a boomsticker.

Lets say there is a threshold where he considers his health to be too low to take the risk. So the decision is affected by input from another player.

At time X players decides to shoot RL. Unfortunately, At X+10ms information that he was hit and his health dropped to 19 reaches his screen. With 20ms lower latency, this would be X-10ms, generating a different decision.

At time X+200ms the RL shot actually happens. At X+210ms he starts to realize his mistake. And before X+1000ms he gets shot down, holding RL.


I think we can presume this happens so rarely with low pings that it's insignificant. But it still happens. Which is kind of what LethalWiz was asking.
2010-05-01, 17:22
Member
386 posts

Registered:
Apr 2006
Kalma wrote:
A time advantage doesn't disappear just because everyone reacts later.

A taller sprinter will have a disadvantage in every start, because it takes longer for the nerve impulse from the brain to reach his leg muscles and his ear will be further away from the loud speaker.


Okay, lets try a QW example.

Player with low health and RL needs to decide if he will next shoot a rocket and risk dropping a pack. He is getting peppered by a boomsticker.

Lets say there is a threshold where he considers his health to be too low to take the risk. So the decision is affected by input from another player.

At time X players decides to shoot RL. Unfortunately, At X+10ms information that he was hit and his health dropped to 19 reaches his screen. With 20ms lower latency, this would be X-10ms, generating a different decision.

At time X+200ms the RL shot actually happens. At X+210ms he starts to realize his mistake. And before X+1000ms he gets shot down, holding RL.


I think we can presume this happens so rarely with low pings that it's insignificant. But it still happens. Which is kind of what LethalWiz was asking.

But this only applies to situations which meet similar conditions to the many mentioned in comment #76 here in addition to the condition in which that 10ms is the proverbial "straw that breaks the camel's back" AND you would have to fire inside that incredibly small time-frame (specifically an rl, making it even less likely by a large factor). The odds of all that happening are simply staggering. I can't even begin to put a figure on it.

No reasonable person can say that creates too great a compromise, especially considering the inherent benefits granted to thousands of hitscan shots each person fires during every single round.
2010-05-02, 10:35
Member
485 posts

Registered:
Feb 2006
Stev wrote:
But this only applies to situations which meet similar conditions to the many mentioned in comment #76 here in addition to the condition in which that 10ms is the proverbial "straw that breaks the camel's back" AND you would have to fire inside that incredibly small time-frame (specifically an rl, making it even less likely by a large factor). The odds of all that happening are simply staggering. I can't even begin to put a figure on it.

If we take the more simple example of a player running behind a wall, it's simple to calculate the odds. Shaft fires every 100ms. Going from 33ms to 0ms lag gives an extra hit 33% of the time, 10hp on average.

This kind of thing will happen once in a while. I think it should be acknowledged.

Quote:
No reasonable person can say that creates too great a compromise, especially considering the inherent benefits granted to thousands of hitscan shots each person fires during every single round.

I agree. And the lower pinger still sees the opponent sooner and thus has an advantage at the start of an encounter.

When I tried to explain this during the first antilaggate, blAze found it unacceptable if it happens at all.
2010-05-02, 13:35
Member
462 posts

Registered:
Jan 2006
Well I reserve the right to change my opinion if new logical arguments or evidence is brought up, which I think is the case now. As medar explained, because of player prediction, from a player point of view strange things could have happened even before antilag, so in that sense it's not as fundamental change as I originally thought.
2010-05-05, 08:43
News Writer
283 posts

Registered:
Jan 2007
Shall we go to a vote sometime soon? I liked Para's idea of cl_fakeshaft 2, if I understood it correctly - this would mean the nme's shaft beam would be drawn to be hitting you, if you suffer damage from the shot (yes even if its through a wall etc)?
This would negate the reflex-movement stuff which is a fair logical concern. I'm not sure how feasible it is to implement though, as shaft beam drawing is server side and cl_fakeshaft is client side...
  324 posts on 11 pages  First page12345678910Last page