Simulation fidelty in online multiplayer

[stupor-mundi]

I'm starting this thread in Great Battles, to draw some attention to the topic, which I have described within Tank Crew under

https://forum.il2sturmovik.com/topic/82052-slow-motion-syndrome-in-multiplayer-on-advance-and-secure-server/

in the hope that some devs will see it here.

 

I haven't flown in many months, since with my current hardware and the FPS I get and the effort it would be to setup head tracking again, I think it wouldn't be a good idea.

However, I strongly suspect, that the effect I'm describing in tanks, will also be present when flying, but maybe much harder to detect.

There have apparently been some threads about this effect on the Russion forum, where apparently, what I call slow-motion-effect, is translated by Google translate into English as "Jelly" effect.

 

The ingredients of the slo-mo effect are:

• weaker client hardware, I suspect single thread performance to be the most important, since my laptop has 16 threads and that doesn't seem to help much.
• client network connectivity, not currently a cause for me, but I've experienced it in the past, when I was in a location where only wifi was available
• server processing power, probably
• possibly some server configuration, I'm pure guessing
• map/mission design
• number of players on the map, in the are of the player. When tanking, that is usually the total number of tank players.

 

With most of these as a given, when I go tanking on a particular, typically Advance and Secure map/mission, I could begin at a point where there are 6 players, estimate that I have 95% percent of my tank's correct performance (reference for that is Offline), go on playing while more players join, and let's say at 9 players the slo-mo becomes so extreme that I would estimate the tank's performace to be around 25%, and I have to quit.

 

Or, with most of these as a given, including the server, and including the number of players, I could play a particular map with only minimal slo-mo effect, at 8 players, while a different map at the same number of players is completely unplayable.

 

At higher number of players however, roughly above 15, all Advance and Secure maps are unplayable for me due to slo-mo effect.

But that is with my client hardware as a given. Clearly there are other players who can still use those maps at 20 players without problems.

 

Now you might think, my client hardware is just below what's needed for IL2. But, on other servers, I do not see this effect, or only at again much higher tank player numbers, for example on Finnish at around 30 players. Or on EFront, I've not seen high enough player numbers to get this effect at all. And, for reference, offline, where my hardware doesn't seem too weak for tanking. And I don't see a correlation of this effect with lowering the FPS.

 

I think this is a seriously bad situation for a "online multiplayer flight (or tank) simulator". It's clear that something has to give when the client hardware is limited and depending on player numbers or map design, more processing is needed. So the FPS should go down. But my FPS stay around 30, no matter slo-mo or not. I've tried dynamic resolution factor, but it made no difference.

 

Due to the multiple ingredients for this, involving the game servers and map design, and on the other hand, the client, there is a lack of problem visibility/awareness for the map designer, and for the gamer.

 

I can detect the problem because I always drive the same tank. I would struggle to detect such a performance loss when flying.

The map makers are volunteers as I understand it, and usually don't have a zoo of different testing hardware of different specs. Usually they will have fast gaming hardware.

 

In a simulator, there should be:

• FPS should go wayyyy down before simulation fidelty is lost
• when this happens there should be some kind of warning
• a metric, from 0 to 1, would be good. It could be displayed with FPS
• this metric should be transmitted to the game server, so that the map designer can figure out that there is an issue.
   

In this way, the map designers would over time become able to see exactly which map design features are most expensive/harmful with respect to slo-mo.

 

[RossMarBow]

Yea it's a known issue.

 

 

Edited January 24, 2023 by RossMarBow

[stupor-mundi]

On 1/24/2023 at 5:03 AM, RossMarBow said:

Yea it's a known issue.

 

 

 

I am astounded this is not more of a topic, if people just say, or yeah, this is a known issue.

The upshot of what I have described, with factors from the game client, multiple factors from the game server/mission design, number of players in the vicinity, is, that performance of the simulated vehicles degrades in a quantitative way, from not all to very obvious.

Without necessarily any canary tell tale sign such as extremely low FPS.

So you get an illusion of a flight sim, you look at a particular vehicle, let's say a particular spitfire model, or T34 model, with a specific load out and fuel, and you expect/think it will be simulated in a consistent way, but the reality is that the spitfire of the player with the less powerful computer might have half the performance of the one belonging to the player with a top spec rig. But not all time, rather, varying with how many players join. ?

That's like having a boxing match where one player gets to have hard gloves, and the other has to use the official gloves, and you bring it up and people say, yeah it's a known issue. Or you buy some food and you find it's just a piece of plastic you can't eat.

That is the one thing an online mutiplayer sim has to do correctly, before any and all other features. Yet there seem to be not many threads about it. What's going on?

 

[CountZero]

a) most ppl dont expiriance it

b) ppl dont notice it or care about it even if it happends

 

 

[RossMarBow]

8 hours ago, CountZero said:

a) most ppl dont expiriance it

b) ppl dont notice it or care about it even if it happends

 

 

If you get caught abusing it in multiplayer you will get banned though.

 

 

 

Edited January 29, 2023 by RossMarBow

[[CPT]Crunch]

And the pool of players is way too small, nobody wants to drive even more away over things like that.

[stupor-mundi]

On 1/29/2023 at 2:30 AM, RossMarBow said:

If you get caught abusing it in multiplayer you will get banned though.

 

 

 

 

In my experience so far, it's something that, if you're on weaker hardware, you become the victim of. Whereas those on powerful hardware don't have to worry about it.

Trying to understand your post, if people abuse this and do get banned over it, well it sounds as if some people are able to induce slo-mo in other players, i.e. are able to control it.

So if there are players who abuse this, they must be players who are on sufficiently strong hardware to not be affected by slo-mo, but are somehow able to induce the slo-mo in other players. I'd like to understand what the mechanism is there, to be able to detect it if/when it might be going on.

 

If this kind of abuse goes on, does that not reinforce my argument that there should be a metric for simulation fidelty, which is not only displayable on the game client, but also transmitted to the server?

 

I have sometimes experienced extreme slo-mo when fighting against/ in close proximity to, other single tanks. I.e. it seemed as I approached closer to the other tank, the slo-mo got worse, without there being an overall increase in player numbers. Would that indicate a situation where someone abuses that?

 

 

On 1/29/2023 at 3:20 AM, [CPT]Crunch said:

And the pool of players is way too small, nobody wants to drive even more away over things like that.

 

TBH it seems to me to be the opposite, a new player joins IL2, at some point realizes there is slo-mo, and would stop using it as a result...

 

 

 

On 1/28/2023 at 6:02 PM, CountZero said:

a) most ppl dont expiriance it

b) ppl dont notice it or care about it even if it happends

 

 

 

I'd kinda like to know what the percentages are.

I see the point of letting go of a tedious point if I'm in a very small minority here.

OTOH, if there's a substantial fraction who have it, but don't notice it, that, to me, is pretty despicable, because then you might have a substantial fraction who have the advantage of it, know about it, and like it that way because they're fighting handicapped opponents...

[stupor-mundi]

To illustrate why I think having a simulation fidelity metric wouldn't be exessively hard to do:

 

I can tell whether I have slo-mo already when I lower the visor in my tank, or open the hatch.

These tank parts take part in the slo-mo, just like everything else that's relevant in multiplayer combat, such as tank overall motion, turret traverse, even bullet motion.

I expect in planes there will be comparable objects which take part in slo-mo, but I can't list any because I haven't flown in a long time.

 

The reason, I think, why visor and hatch take part in slo-mo, will be

* they can be seen

* they are part of the damage model

 

It seems to me that visor and hatch are not affected by gravity, i.e. opening/closing speed seem to me unaffected by the angle of the tank, when on a steep slope, for instance.

 

So I think, all that would be needed, would be to place an invisible widget on all vehicles, that has a predetermined opening/closing speed, that is registered as having its state transmitted in multiplayer, and that would be caused to open/close every 30 seconds or so (by a timer).

 

Then all you have to do is time the actual opening/closing speed.

 

Simulation fidelty metric = proper opening speed / actual opening speed.

 

Not too hard to do I think.

Edited January 30, 2023 by stupor-mundi

[RossMarBow]

Put the bong down bro

 

You slow yourself down, but of course you end up going faster afterwards.

Edited January 31, 2023 by RossMarBow

[dgiatr]

9 hours ago, stupor-mundi said:

To illustrate why I think having a simulation fidelity metric wouldn't be exessively hard to do:

 

I can tell whether I have slo-mo already when I lower the visor in my tank, or open the hatch.

These tank parts take part in the slo-mo, just like everything else that's relevant in multiplayer combat, such as tank overall motion, turret traverse, even bullet motion.

I expect in planes there will be comparable objects which take part in slo-mo, but I can't list any because I haven't flown in a long time.

 

The reason, I think, why visor and hatch take part in slo-mo, will be

* they can be seen

* they are part of the damage model

 

It seems to me that visor and hatch are not affected by gravity, i.e. opening/closing speed seem to me unaffected by the angle of the tank, when on a steep slope, for instance.

 

So I think, all that would be needed, would be to place an invisible widget on all vehicles, that has a predetermined opening/closing speed, that is registered as having its state transmitted in multiplayer, and that would be caused to open/close every 30 seconds or so (by a timer).

 

Then all you have to do is time the actual opening/closing speed.

 

Simulation fidelty metric = proper opening speed / actual opening speed.

 

Not too hard to do I think.

You are right about that , its a known issue that i think Devs are aware of this.

 

My opinion is that it has to do with how game engine works and that IL2 is mostly a single threaded application which loads too much one single cpu  thread each time and a little some other threads, that's why clients CPU single thread performance is the key factor that affects too much that slow-motion issue. I experienced it myself many times while flying online exactly where and when you described and had to buy a 13th gen cpu in order to deal with it ( apparently that shouldn't be a solution ). I have seen many complaints from other players about that but i think that IL2 game engine should be redesigned for that issue to be solved.

 

Same problems also appear in single player mode with AI planes.

 

Strange thing is that one year ago this problem didn't exist... its very obvious with last game versions, i don't know what changed and if someone can go back to fix it, probably new game versions are more cpu demanding...

 

Unfortunately i don't see anybody to bother....

Edited January 31, 2023 by dgiatr

[CountZero]

14 hours ago, stupor-mundi said:

 

In my experience so far, it's something that, if you're on weaker hardware, you become the victim of. Whereas those on powerful hardware don't have to worry about it.

Trying to understand your post, if people abuse this and do get banned over it, well it sounds as if some people are able to induce slo-mo in other players, i.e. are able to control it.

So if there are players who abuse this, they must be players who are on sufficiently strong hardware to not be affected by slo-mo, but are somehow able to induce the slo-mo in other players. I'd like to understand what the mechanism is there, to be able to detect it if/when it might be going on.

 

If this kind of abuse goes on, does that not reinforce my argument that there should be a metric for simulation fidelty, which is not only displayable on the game client, but also transmitted to the server?

 

I have sometimes experienced extreme slo-mo when fighting against/ in close proximity to, other single tanks. I.e. it seemed as I approached closer to the other tank, the slo-mo got worse, without there being an overall increase in player numbers. Would that indicate a situation where someone abuses that?

 

 

 

TBH it seems to me to be the opposite, a new player joins IL2, at some point realizes there is slo-mo, and would stop using it as a result...

 

 

 

 

I'd kinda like to know what the percentages are.

I see the point of letting go of a tedious point if I'm in a very small minority here.

OTOH, if there's a substantial fraction who have it, but don't notice it, that, to me, is pretty despicable, because then you might have a substantial fraction who have the advantage of it, know about it, and like it that way because they're fighting handicapped opponents...

servers can always lisen to what devs sugested last time and try with less slots then 84, game right now cant handle 84 ppl servers, and counting on devs to see what they mesed up to make slow.motion happend more offten, when their focus is new game is not great idea... but i still think not many ppl expiriance this as if they do there would be more complains about it, i remenber when bullet data was reporting at wrong time, and there was imidiatly mas of complains about it and after long time it got looked at and fixed... but it was obvious and happedn to many so it took mass of ppl comlaining about it for devs to look at it, for this slow down i do not see mass MP ppl comlain about it.

[RossMarBow]

Server admins don't care about this (topic in general).

The only time they care is if enough people start complaining about a single person.

 

They then have the audacity to complain when the devs do the same thing.

 

You will never get a majority of the community understanding complicated issues like this. They are just too dumb.
And multiplayer servers are ruled by the lowest common denominator.

 

The only solution is for the devs to actually care about multiplayer (they have already said they don't) and place much stricter controls on multiplayer servers to insure quality.  

 

Multiplayer stutters cause identified - Complaints - IL-2 Sturmovik Forum (il2sturmovik.com)

 

@stupor-mundi read this thread - devs said it was too hard to fix, and multiplayer admins said it wasn't important

This is really just a dead topic.

Edited February 1, 2023 by RossMarBow

[RossMarBow]

18 hours ago, CountZero said:

servers can always lisen to what devs sugested last time and try with less slots then 84, game right now cant handle 84 ppl servers, and counting on devs to see what they mesed up to make slow.motion happend more offten, when their focus is new game is not great idea... but i still think not many ppl expiriance this as if they do there would be more complains about it, i remenber when bullet data was reporting at wrong time, and there was imidiatly mas of complains about it and after long time it got looked at and fixed... but it was obvious and happedn to many so it took mass of ppl comlaining about it for devs to look at it, for this slow down i do not see mass MP ppl comlain about it.

If you hang out of the multiplayer discords. People are continually running into this issue. Usually people trying to run bomber formations suffer the most from this. For everyone else it doesn't really matter unless someone is obviously abusing to slow their plane down for reversals. 

Edited February 1, 2023 by RossMarBow

[CountZero]

12 hours ago, RossMarBow said:

If you hang out of the multiplayer discords. People are continually running into this issue. Usually people trying to run bomber formations suffer the most from this. For everyone else it doesn't really matter unless someone is obviously abusing to slow their plane down for reversals. 

Ill say what i say to my tm8s when they complain to me about something, IF its not on forum it do not mather, devs cant here what we talk on ts3...  if they cant get bathered to post about it on forum then it dosent bather them that mutch.

[stupor-mundi]

18 hours ago, RossMarBow said:

Usually people trying to run bomber formations suffer the most from this.

 

Interesting point, makes sense to me, when comparing to tanks. Lots of vehicles/planes in fairly close proximity to each other, which otherwise with planes you don't get as much. And the bombers have many stateful sub-assemblies such as turrets, which are relevant in multiplayer combat and the state of which will be constantly transmitted.

 

18 hours ago, RossMarBow said:

read this thread - devs said it was too hard to fix

 

Interesting point in that thread about the number of spawn points.

I get very varying amount of slo-mo in the different missions of AAS, I'm going to keep this in the back of my head when looking at those, to see if that correlates generally with the outcomes I see. A mission which has particularly little, sometimes completely absent, slow-mo for me, Tessy, also has few spawn points.
Whereas for example Villersbocage which gives me loads of slo-mo has many spawn points, most of which are disabled at any one point of time, but I suppose the disabled spawn points will be just as harmful. So I'm intrigued.

It's interesting because to the player, the non-map-designer, the spawn points don't appear to 'do' anything, in the way of consuming client cpu or server cpu, but apparently they do something...

[RossMarBow]

3 hours ago, stupor-mundi said:

 

Interesting point, makes sense to me, when comparing to tanks. Lots of vehicles/planes in fairly close proximity to each other, which otherwise with planes you don't get as much. And the bombers have many stateful sub-assemblies such as turrets, which are relevant in multiplayer combat and the state of which will be constantly transmitted.

 

 

Interesting point in that thread about the number of spawn points.

I get very varying amount of slo-mo in the different missions of AAS, I'm going to keep this in the back of my head when looking at those, to see if that correlates generally with the outcomes I see. A mission which has particularly little, sometimes completely absent, slow-mo for me, Tessy, also has few spawn points.
Whereas for example Villersbocage which gives me loads of slo-mo has many spawn points, most of which are disabled at any one point of time, but I suppose the disabled spawn points will be just as harmful. So I'm intrigued.

It's interesting because to the player, the non-map-designer, the spawn points don't appear to 'do' anything, in the way of consuming client cpu or server cpu, but apparently they do something...

They all provide radar and all update at the same time.

[stupor-mundi]

10 hours ago, RossMarBow said:

They all provide radar and all update at the same time.

great coding, devs.