109 replies to this topic

[TexAce]

Anyone care to explain this rewind thing more in detail? Don't hold back, I understand l337 and nerdcore, I just dont know what this one actually is.

I hope with the ping headroom we can enable IK again.

Edited by Big Black Wolf, 26 June 2015 - 03:27 AM.

[TWIAFU]

FupDup, on 25 June 2015 - 08:47 PM, said:

Maybe the complaints about "Broken hitboxes, OMFG!" will finally stop.

Bad pilots will find a new excuse to explain away bad piloting/gunnery skills.

[zagibu]

Big Black Wolf, on 26 June 2015 - 03:11 AM, said:

Anyone care to explain this rewind thing more in detail? Don't hold back, I understand l337 and nerdcore, I just dont know what this one actually is.

Let's say you have 200ms ping. If you fire at a target, you basically fire at a position that is 200ms old. Then it takes another 200ms for your fire command to reach the server. Without anything to help address this delay, you would miss a lot, because you would fire at a position your target was at 400ms ago. With host state rewind, the server basically rewind's everything 400ms into the past, so that it hopefully has roughly the same situation that you saw on your client.

Of course, that's a gross oversimplification, there is also client-side prediction and fire commands are probably timestamped, but it should give you an idea on how HSR works.

The new thing now is that it doesn't only rewind the positions, but also the animations (leg/arm poses, torso rotation, etc.), so hopefully, less shots that seem to hit your arm but damage your CT.

Edited by zagibu, 26 June 2015 - 07:06 AM.

[Zordicron]

If this fix is what PGI says it will be, then it is finally here.

Most of my weapon balance posts have involved HSR issues. Spreadsheet warriors like to ignore the HSR stuff. Since open beta, I have been saying "when the weapons do ALL the dmg they are supposed to, ALL the time, Paul will have to start over 100% with balance."

If this "fix" does what it looks like it could...... well, I hope Paul is taking a couple days off before he gets started on the complete rework ahead of him.

[Karamarka]

As an Aussie i'l believe it when i see it : P

[Lugh]

Vxheous Kerensky, on 25 June 2015 - 11:26 PM, said:

I am still taking rear armor damage from ballistic shots to the front from time to time, pretty sure due to HSR predicting that the shot was mid body or something. AC-5 and Gauss seem to be the highest culprits for this

I've been getting kills when I was aiming at what looked to be the dudes side because he'd moved correctly to roll my lasers, and it all went to his beet red back and BOOM.

Karamarka, on 26 June 2015 - 07:31 AM, said:

As an Aussie i'l believe it when i see it : P

But even if you see it, won't it be 200ms late?

Edited by Lugh, 26 June 2015 - 07:39 AM.

[oldradagast]

I'm trying to figure out where this magical place is where lights are god-tier and played only by terrible pilots who don't understand skill?

Every evening I play, I look at the queues, and it is rare for Lights to hit 15%, and I don't think I've ever seen them at 20% except maybe just after the release of a popular new Light mech. If they were really god-tier, unstoppable monsters protected by magical HSR shields, everyone would be playing them, just like everyone played the Stormcrow and Timberwolf before they took a nerf to the knee.

I don't know... it looks to me more like people either just hate Light mechs and want them out of the game so we can focus on tunnel-vision murderball with no capping, or they wanted Lights to "suck so bad" that they are on the only ones left playing them.

As for HSR, yes, it is spotty and bad. Lights are the easiest targets to kill when hit, so HSR problems show up more on them - that, and they move fast. But I'd be willing to be that plenty of damage is lost against larger mechs, but nobody notices it as much because they have more hitpoints and thus people don't really track how they "should have died in that shot."

[Lugh]

oldradagast, on 26 June 2015 - 07:39 AM, said:

I'm trying to figure out where this magical place is where lights are god-tier and played only by terrible pilots who don't understand skill?

Every evening I play, I look at the queues, and it is rare for Lights to hit 15%, and I don't think I've ever seen them at 20% except maybe just after the release of a popular new Light mech. If they were really god-tier, unstoppable monsters protected by magical HSR shields, everyone would be playing them, just like everyone played the Stormcrow and Timberwolf before they took a nerf to the knee.

I don't know... it looks to me more like people either just hate Light mechs and want them out of the game so we can focus on tunnel-vision murderball with no capping, or they wanted Lights to "suck so bad" that they are on the only ones left playing them.

As for HSR, yes, it is spotty and bad. Lights are the easiest targets to kill when hit, so HSR problems show up more on them - that, and they move fast. But I'd be willing to be that plenty of damage is lost against larger mechs, but nobody notices it as much because they have more hitpoints and thus people don't really track how they "should have died in that shot."

Everyone is always annoyed when the 'little guy' gets the better of them. It's true in real life and true in game too.

[oldradagast]

Lugh, on 26 June 2015 - 08:03 AM, said:

Everyone is always annoyed when the 'little guy' gets the better of them. It's true in real life and true in game too.

Basically. I've run into no shortage of people who claim "Lights are the kings of the battlefield!" Which, I guess, is why the Light queue rarely hits 15% - people must hate winning and have chosen to play harder mechs, like Heavies - and also why when these same people play a Light, they typically die dealing under 100 damage.

It's a joke. Lights are borderline useless in reality. Better hit-reg is needed, but it will make them even less useful. At least it sounds like with turret removal from Assault, they might be able to cap again, but still... I expect the Light queue to be buried at about 9% after all this is done.

[Neema Teymory]

Navid A1, on 25 June 2015 - 08:56 PM, said:

You know... i'm both excited and concerned about this.

Excited: because better hitreg.

Concerned: If something so painfully primitive and obvious such as this only gets noticed well more than a year after the creation of HSR... then it only brings up more suspicions on other aspects of the game... things that do not work properly not because there is a reason, but because it has not been figured out yet.






I would LOVE to see the face of wannabe l337s in their firestarters throwing l2p comments around denying every video or personal experiences (some of them in particular in fact)

To address this concern, voiced also by several others in this thread:

Animation rewinding was always something we were aware of, and intentionally chose not to support in the initial implementation of the HSR algorithm. The reason for this is, that at the time, the HSR algorithm's complexity made animation rewinding prohibitively expensive. In certain degenerative cases, server performance would be impacted to the point that the game would be unplayable. This negative effect outweighed the potential positive effects that animation rewinding would introduce. We felt that delivering the HSR algorithm as you know it now (without animation rewinding) was still a great improvement to the game, even though it wasn't as complete as we would have liked.

In the coming round of changes, several key optimizations were made to the way the HSR algorithm works that now make animation rewinding viable. Which is why these changes are coming now.

With that said, as you have seen, Russ mentioned that several other bugs were fixed along the way. These also contributed to problems with the HSR algorithm. So, to be clear, animation rewinding isn't the only reason for "hit reg" problems. We are hopeful that the changes, as a whole (including the bug fixes) should be a noticeable improvement for users of any ping.

zagibu, on 26 June 2015 - 07:05 AM, said:

Let's say you have 200ms ping. If you fire at a target, you basically fire at a position that is 200ms old. Then it takes another 200ms for your fire command to reach the server. Without anything to help address this delay, you would miss a lot, because you would fire at a position your target was at 400ms ago. With host state rewind, the server basically rewind's everything 400ms into the past, so that it hopefully has roughly the same situation that you saw on your client.

This is the right idea. However, ping is a round trip time (i.e. the time it takes a message to get from your client to the server and then back to your client). That means if you have 200ms ping, it actually takes 100ms for a message to go in one direction. Therefore, what you see is actually 100ms old. Then, it takes another 100 ms for your fire command to reach the server. So by the time the server is processing your fire request, it is 200 ms ahead of what the client decided to shoot at. The server compensates for exactly this time difference.

[Sarlic]

Thanks Neema, have a beer on me.

[Zordicron]

Neema,

could you give us a word on if there will be improvment to jitter affected players, you know, fluctuating ping?

Ihave had both bad, and fluctuating at the same time for a long while, and I can attest the jitter was far more annoying then the actual latency. Will the new "fix" have an effect on packet loss etc caused by excessive jitter?

thanks

[Monkey Lover]

Neema Teymory, thank you and everyone at Pgi for this fix. Its my feeling this will improve the game greatly. This should increase the player base as much or more.than any feature.

Edited by Monkey Lover, 26 June 2015 - 10:46 AM.

[Lugh]

Eldagore, on 26 June 2015 - 10:26 AM, said:

Neema,

could you give us a word on if there will be improvment to jitter affected players, you know, fluctuating ping?

Ihave had both bad, and fluctuating at the same time for a long while, and I can attest the jitter was far more annoying then the actual latency. Will the new "fix" have an effect on packet loss etc caused by excessive jitter?

thanks

Jitter as a whole can be caused by so many different things between the client and the server. Have you done the usual fixes for attempting to fix this problem, being sure to disable your firewall(or program in the ports to open /forward for the game), turned off your anti-virus while playing (yes you can stop surfing porn while you wait in que without dying), has your modem's firmware or your routers firmware been updated to the latest and greatest version?

When experiencing the Jitter have you done a pathping to mwomercs.com(or whatever the server ip is) to find out where in the path your problem might be propping up? This is useful information both for you ( to reach out to your isp to report the device at x.x.x.x is showing large amounts of packet loss can you fix?) and for PGI (maybe a router in the data center is flapping and needs it's interfaces reset) ?

[Neema Teymory]

Eldagore, on 26 June 2015 - 10:26 AM, said:

Neema, could you give us a word on if there will be improvment to jitter affected players, you know, fluctuating ping? Ihave had both bad, and fluctuating at the same time for a long while, and I can attest the jitter was far more annoying then the actual latency. Will the new "fix" have an effect on packet loss etc caused by excessive jitter? thanks

Hey Eldagore,

Unfortunately, the coming changes do not make the algorithm any more resilient to problems introduced by sudden variations in latency. I am actively working on some more changes that should help with this, but they are not going to make it in time for the next patch. And while these changes should improve your experience while experiencing fluctuating latency, they will not be able to completely negate the problem.

If there is some way for you to have a stable, higher ping, this is better than a lower, fluctuating ping. And when I say fluctuations, I mean relatively large fluctuations (> ~50ms), smaller fluctuations _shouldn't_ matter so much. With that said, I think you should still see a noticeable improvement with the latest changes.

[Kh0rn]

Now it will be pay back time to for those who escaped my weapons due too hit reg issues. Skull Crusher is coming for you!

[Sarlic]

Neema Teymory, on 26 June 2015 - 10:59 AM, said:

Hey Eldagore,

Unfortunately, the coming changes do not make the algorithm any more resilient to problems introduced by sudden variations in latency. I am actively working on some more changes that should help with this, but they are not going to make it in time for the next patch. And while these changes should improve your experience while experiencing fluctuating latency, they will not be able to completely negate the problem.

If there is some way for you to have a stable, higher ping, this is better than a lower, fluctuating ping. And when I say fluctuations, I mean relatively large fluctuations (> ~50ms), smaller fluctuations _shouldn't_ matter so much. With that said, I think you should still see a noticeable improvement with the latest changes.

You should post more on the forums.

Go slap any dev on your office to make a decent Developer tracker to use on the forums.

[Dinochrome]

Hailz all

Unfortunately I am totally at a loss as to why "Bone Rewind" is a better thing than "HSR". Is there a succinct explanation that I as a member of the great unwashed Merc rabble could understand?

V/R Dinochrome

[Kristov Kerensky]

Dinochrome, on 26 June 2015 - 11:20 AM, said:

Hailz all

Unfortunately I am totally at a loss as to why "Bone Rewind" is a better thing than "HSR". Is there a succinct explanation that I as a member of the great unwashed Merc rabble could understand?

V/R Dinochrome

It's the same thing really, it just describes a portion of the HSR system, which directly deals with the 'bones' of a Mech, the pieces used in the animation process. By making the HSR read not only the main structure state but also the 'bones', the system looks at where the ENTIRE Mech was at the time of your input, before it was only looking at the CT at the time of your input. This is simplified, but you should get the idea I hope.

And I understand why they didn't go with the full bone rewind, it's expensive, and since they've had to shoehorn the entire HSR system into the engine, getting it all optimized isn't an easy process. I'm glad they've finally been able to do this, along with finding some bugs they didn't realize existed and getting them fixed.

Netcode ain't easy, people don't realize how much really goes on with it, they just assume it's a simple fix, it's not, it's very complicated and takes a lot of work and time.

Cudos to you Neema, and thank you so very much!

[Weeny Machine]

Sarlic, on 25 June 2015 - 08:43 PM, said:

Brace yourselves!

Rambo Light complaining getting killed too easily incoming!

Or...

...or suddenly you think WTF?! when an Oxide's alpha fully registers...