Our ReVolt Pub

Following Phantom's request to improve lag management in version 14.0306, I've created this area so that we can discuss in private in a more convenient way than using PM.

We believe there's a chance for 0306's lag management to be better and make everyone love it.

To start with, I need to know more about lag management:
1. When a game is hosted with compatibility On, which lag management does it use exactly? Does it use the standard 1.00 management or is the game affected by the version's optimizations?

2. Is the lag management code something completely separable from the physics code or are both the same thing?

3. Is it possible to introduce the lag management code from 0306 into an older version of 1.2 to be tested?

Please, I would really need you to answer all of these questions and anything else you consider important to clarify my doubts and have a good start. I love when you explain things as clear as water. I've already have prepared a few reports on what are the weak points of 0306 and what things need optimization.

About the feedback, I need you to be clear what you want to hear and what you don't want to hear from us when we give a feedback. In my case I'll try to give details as accurately as possible but I never know what else you expect unless you tell me.

1. Off the top of my head since I have not checked the code for quite some time now, compatibility is mostly here to be able to run v1.2 together with v1.1, but it isn't meant to be a real unmodified v1.1 mode.

The reason for this is that keeping both new and legacy code for everything online-related would be deadly cumbersome... and removing old useless code is a pleasure as great as adding new code. Joking aside, cleaning up the code is necessary to make it easier to maintain.

All this to say that even with compatibility On, the lag management will likely use most of the v1.2 modifications and can't be used to compare new / old code (if that's your actual wonder).

2. For the local car I think we can consider that the lag management code is completely separated from the physics code, but when it comes to online gaming, the physics engine necessarily have to be cheated to compensate for the delay. It's not really the physics that are modified in fact, but after the regular physics code, extrapolation / interpolation are (and have to) applied to the remote cars / objects, according to the data received from remote players (check this post for a reminder about extrapolation / interpolation). It took time to figure this out, but I think that's quite the only reason left that could explain increased weapons effect in online races.

3. In theory it should be, but I think this will be a boring task and not really the way I'd be the most prone to proceed.

Anyway, if one older version were guaranteed to be glitch-free for example, I can (provided I have a source backup, it's not always the case for old releases) check each modification and reverse changes that are likely to have an effect on what is currently studied.

It's really good to see this privat topic about the 1.2 version problems and optimization. But the only thing what I can do currently here is test with Phantom. I really don't know much about the specific problems between the 1.2 alpha versions.

The only things that I really know is: the gameplay, the ballance between the weapons, the physics and collision on 11.0421 is better then on 13.0820, IMO.

Oh, and I can Phantom explain your patch developer and programmer words because he don't understand these really. I think I can better understand the connections than him.

RV_Passion @ Jan 8 2015, 09:59 PM wrote:the gameplay, the ballance between the weapons, the physics and collision on 11.0421 is better then on 14.0820, IMO.

Ok, so I can give my first advice for bug reporting in this topic. I know we're here to talk about online gaming, but whenever you mention physics / collision behavior difference, it's important to know whether the difference is noticeable:

- offline. You already notice the difference when playing offline. Obviously if the offline behavior is different the online behavior will be different as well.
- online only. You don't notice physics difference if you play offline, but you can notice them when you play online.
- online (offline no tested). I'd prefer avoid this category since it would be better, whenever you feel an online difference, to test whether or not you can feel the same difference in offline mode as well. But at least if you haven't tried offline let us know.

It would also be interesting to know whether it's mostly about:
- your own car physic / collision behavior
- the car / car collision
- the weapons behavior
- other.

I also have a request before we can start anything serious: can you guys make a kind of summary of the current situation. Something like, which builds are currently mostly appreciated / used or criticised / rejected, for what reasons? (including beta / alpha official builds as well as the test builds I provided some time ago). What was left unfinished in my latest Build (apart from the online code)? And about the rv1.2a14.0208 build changelog, what needs to be included / rejected compared to my latest test build G?

This is needed so that we can start to draw the rough lines of a roadmap. I haven't been working or even using the v1.2 for quite some time and my memory badly needs to be refreshed.

I know we're here to talk about online gaming, but whenever you mention physics / collision behavior difference, it's important to know whether the difference is noticeable:

Basically you can see and notice the differences offline and online, if you compare 13.0820 with 11.0421 for example. The best way to see the differences is make some online races with some players in one room with two different versions (13.0820 and 11.0421).

It would also be interesting to know whether it's mostly about:

-The car physics, car collisions and collision behaviors generally.
-The ballance between the weapons. With weapon ballance I mean, on 13.0820 you see often races with to much oils, stars, rockets, shockwaves or bombs. Its sometimes very difficult to come forward. Especially with to much shoot weapons where it comes with "Shooting Everywhere" situations.

can you make a kind of summary of the current situation. Something like, which builds are currently mostly appreciated / used or criticised / rejected, for what reasons? (including beta / alpha official builds as well as the test builds I provided some time ago).

Well, the only things that I can say about that is, on rvhouse is currently the most used version 13.0820. Only a few people use 14.0306, like me. How it looks on gameranger I don't know. To say which version is the most appreciated or criticised is very difficult to say because the most people just use the most used version. The only people who have said something about a version are me, dolo, stingbox or phantom. It seems like the other guys just accept the differences or they don't see a difference or they don't really care about the differences. Me, Phantom and Dolo like more the 11.0421 and we are the only ones that have criticised the 13.0820.

Thats all what I can say. If I had forget something to say, tell me that.

I know we're here to talk about improving the lag management in Re-Volt 1.2 and not about the physics/steering/weapon balance differences between old and new versions of the Patch but since you asked us to give a summary of the current situation in the online world, it's impossible to avoid it because it's the main point of frustration in the online players.

• Which builds are currently mostly appreciated / used or criticised / rejected, for what reasons?

Build 14.0314: Last time we used this was last year when I sent you the screenshots. Nobody uses this for normal races. I can re try testing it with Passion and others whenever you wish but as far as I remember I always had better results with 0306, so I don't know if it would be useful.

Build 14.0306: Top favourite for some players. Great track selector and best lag management so far for 2-3 players (no teleporting). Bad for crowded races and long racing sessions which is why it can't be the new standard and still needs optimization. I'll make a deeper review of this version below.

Build 14.0208: It's the top rejected version. Nobody uses it and the few who have tried it have downgraded to 0820 inmediately. I guarantee that 0.00% of the online players use this version.
The only 2 differences with 0820 are the time trial modifications and the Re-Volt music from hard drive. I've already given my thoughts about the time trial modifications and after a long discussion Huki finally understood my reasons and proposed a new scheme for the future. I beg you do not include any time trial modifications and keep the current 0820 system. The Re-Volt music from hard drive is obviusly something amazing. Nobody has complained about it so if including it in any of the following test versions is in your power, I'd say do it.

Build 13.0820: Current standard. I can't say it's the most appreciated but it's the most used for sure. In comparison to 0306, is relatively more decent for crowded races. It is also the most criticized because the lag is usually infernal on 12 player races, but people still play it. If you were on RVHouse, you could hear the same complaints about it every day: the lag, the teleporting, the strong impact of weapons, the tons of oils and bombs, the strange physics, the steering having less grip than old versions, that it feels driving on ice, people saying that aren't as skillfull with this version as with older version, etc.
And I'm not talking about Dolo, he doesn't even join these races. It's the general public. But you know how people are, they complain but they don't care, they'll play with it until something better comes to replace it.

Build 11.0421: In our search trying to find the origin of the collission issues, we started testing old versions from 2011 prior to any potential changes that may have affected the gameplay.
This is the oldest alpha release available in the site and it's one wonderful version, with Late Joining included and many bug fixes, which make it a better alternative than Beta 11.0208.
We mostly play on it with compatibility on because we noticed the lag management seems to be better than compatibility off, probably because this was one of the first versions to introduce Late Joining, so we guess the code wasn't optimized. The compatibility off tends to be laggier than compatibility on races. The lag management is certainly not as good as in 0306, there's the typical teleporting you would see in 1.1 where the car moves back and forth but despite this, the gameplay is still awesome.

Main reason why we love it is that we are sure that this version is bug free when it comes to collission / physics / steering / weapon balance. There is nothing that I can record and post it on a video, it's a feeling. Online races are simply more exciting on it, the steering feels different for all the players that have tested it so far. mmudshark has said that 0820 feels like driving on ice, while 0421 has more grip, and I feel exact the same. The contact with the surfaces, specifically the floor, feels better and therefore the steering is more enjoyable. Cosmo also insisted that the difference is evident but since we're online players exclusively, we're not on our territorry when it comes to offline races. That's why it's so hard to determine with the same accuracy as we can feel it online. Like Passion said, our way of testing works gathering some players in a room, and comparing races with one or another version. Always on Arcade mode of course, it's the only mode used by online players. If you ask us about other collission modes we can't tell, we don't play other collission modes.

Build 11.0208: Nah, after re-discovering the 11.0421 I don't think anybody keeps playing with this version.

• What was left unfinished in my latest Build (apart from the online code)?
That great Track selector. The ability to switch from Normal to Mirrored or Reverse modes without quitting the game.
And improve one annoyance: You know that writing words inside the new Change track menu starts the track search. But if the player presses F12 key to chat while he is inside of the Change track menu, the words appear both in Chat and Track search causing an inconvenience for the player. The correct behaviour would be that pressing F12 key and start writing words when inside this menu shouldn't alter the text typed to search a track.

• And about the rv1.2a14.0208 build changelog, what needs to be included / rejected compared to my latest test build G?
Included: the ability to listen to the game soundtracks without the need of a CD or ISO Image.
Rejected: any modification to the Time Trial system.

Now, my observations about the 14.0306 are the following:
In this Patch players don't teleport as much as in 13.0820 because the 14.0306 tries to hide the lag as much as possible. Lag in this version is not easily detectable, and after re-reading your explanations about extrapolation-interpolation again, I can see it is connected with this.

But there is one problem with this version, the hiding of lag also has a side effect. It is sometimes noticeable that the position of the other cars isn't real, it's like the game is lying to you, creating fake positions of the other drivers to give you the fake idea of an unlaggy race. In contrast, old versions and even the 13.0820 show cars teleporting because they don't show a fake car position when data is not received, or at least is not as evident as in 13.0306.
The negative impact of this is strange things like what I'm going to show you now which are the reason why the 13.0820 standard hasn't been replaced yet for crowded races. The following belongs to a series of tests made with 11 players on 13.0306, everyone with all internet consuming applications off as I requested to them to reduce lag as much as possible.

Click on the button. From the other driver's perspective the electroshock never took place, but from the driver's perspective his attack didn't work.

Now look at minute 1:07, unexplainable bomb pass through the wall when the driver wasn't touched directly.

And now pay attention at how the bomb is passed after 2:48

Not seeing cars teleporting is fabulous. I mean, who wants to see cars teleporting? Everyone that has tested the 14.0306 can see that the lag management is better because cars don't teleport. But this is only suitable for 2-4 players and usually during a short period of time. In the first races of a session, the lag management is usually stable for some time.

The average online race tends to have 10-12 players from different countries in one room, it is usually long enough to race a minimum of 13 tracks, and has late joining on, which in my opinion is another cause of lag. I've detected that independently of the version, when a player is joining a game in progress the lag tends to increase. Races with no late joiners tend to be less laggy.

Seeing all this, I wonder what are the possible optimizations to 14.0306 that we can try.

Thanks, and very good job, that's exactly what I was asking for.

In terms of Phantoms PMs to me on RVL, just ignore and forget my last part from my post. My answers about exactly version or build question are rather dump.

In this Patch players don't teleport as much as in 13.0820 because the 14.0306 tries to hide the lag as much as possible. Lag in this version is not easily detectable, and after re-reading your explanations about extrapolation-interpolation again, I can see it is connected with this.

But there is one problem with this version, the hiding of lag also has a side effect. It is sometimes noticeable that the position of the other cars isn't real, it's like the game is lying to you, creating fake positions of the other drivers to give you the fake idea of an unlaggy race. In contrast, old versions and even the 13.0820 show cars teleporting because they don't show a fake car position when data is not received, or at least is not as evident as in 13.0306.

I understand this part can be very easily misleading, even myself I have to think twice every time to be sure I'm not getting it wrong, and I admit I'm not even sure I really get it right... anyway I have to correct one point I think you're misunderstanding.

Please read slowly... (it may sound redundant sometimes but I tried to make my point clear).

First, just as a reminder: Interpolation, when used alone, does not hide the lag. It just prevents a remote object trajectory to look jerky. If a player sends 6 packets per second and you update his car location 6 times per second only (direct copy) it will look exasperatingly jerky (not to mention we're late already because of the lag needed to receive the remote data). So instead of applying the new received location we can smoothly move the car from its current location to the new received data. It means the car position is delayed as it takes some time to reach the last received location. It implies that interpolation increases the lag even more.

And this is no good, but as we've seen we can't use direct remote location copy (too jerky, and already late anyway). That's why we use extrapolation. But now it means we have to guess what we have no way to know, that is to say the current position of the remote car. If the lag is 500ms, we just know where this car was 500ms ago. We also know where it was 650ms ago. And we have to mathematically guess where it is now, just from these data.

Our predication can be wrong, but chances are that is will be more accurate than direct remote data copy (with direct copy the car location is not only jerky but it is also always late, because the lag has not been compensated at all).

So using extrapolation to hide the lag is not lying to you much more than direct copy. Direct copy uses the accurate position of the car, but 500ms ago, so it's also some kind of lie (and can lead to the same kind of aberration you're mentioning).

The weird collisions, bomb pass through, etc. are not side effects of hiding the lag, there are merely the side effects of the lag, only the lag. I you have understood the points above you'll agree that hiding lag though extrapolation is in fact rather likely to lower these issues than to increase them. But when the lag is too great, it's become impossible to predict the current remote car position, that's why extrapolation has to be capped (to avoid blatant location aberration).

I agree though, that since extrapolation tries to hide the lag, it's kind of hiding the fact that something is going wrong and you're more surprised of the weird weapon reaction or collision than you would be if the remote car were actually jerky (if you know that something is going wrong with the car already, you will not pay much attention to a weird collision).

I've detected that independently of the version, when a player is joining a game in progress the lag tends to increase.

Interesting info.

Thanks for this great explanation. Great job.

If there's any other info you would like us to provide or any other way we could be of help just mention it.

Thanks Phantom. In fact I'm just back home today and still didn't have the time to really get started (not to mention my free time is concurrently taken by the Blender plugin, this issue and some other parts with Huki for the future of v1.2...).

I think the next step will be only when I'm ready to send you a new build to test (maybe at the end of the week but I can't promise anything yet). Until then I'm sorry I can't ask you much since you've already done such a good job with the preliminary report.

Allright, see you at the end of the week.

Good luck with your next build jigebren.

Btw Phantom, if we have jigebrens new alpha build should we test on gameranger with some players?

Yes on gameranger will be better as soon as we get something.

Yes, sorry guys I've been much busier than expected (in real life, with the future OpenGL build, with the ORP board...). As soon as I can get enough hours to concentrate on this issue I'll let you know. I've already had a look, but no time to code yet.

Don't worry jigebren. We can wait, even if it means something inactivity here. It should be just don't discouraged you because it tooks more time. Your work will not disappoint us.

Just a question: I presume the more players are in a session, the more the lag effect become noticeable, right? So have you, from experience, noticed some practical min-max values on the number of players? (when you know that in the current session there's actually player with bad connection, I mean not the max valuer you were already able to reach in the perfect case scenario).

For example, you may have noticed that you almost expect no issue below 6 players, not matter the players who joined in... And that above 10 players you're quite sure it will turn bad at one time or another. In that case I'm interested by these 2 values.

Or does it mostly depends on which players are in the session? That is to say: who joined in is actually more important than how much players joined in.

Testing this! Will be back with an answer with accurate numbers. Very very good question.

Have you already make tests Phantom?

RV_Passion @ Jan 30 2015, 07:11 PM wrote: Have you already make tests Phantom?

As you noticed, I haven't been online this week. Got too much work to do, sorry. I guess we can do more specific tests for this question this weekend.

jigebren wrote:For example, you may have noticed that you almost expect no issue below 6 players, not matter the players who joined in... And that above 10 players you're quite sure it will turn bad at one time or another. In that case I'm interested by these 2 values.

From experience with the 0820, 6 is already a high number for Re-Volt. In a 6 player race I'm very sure it will turn bad at one moment or another in the 0820 if a player with unstable ping joins. Sometimes not in the first race if he/she isn't having a really bad ping, but it will be noticeable in the 2nd-3rd race. For the 0306, which I think your question is aimed for, I haven't really had much experience in the range 4-8 players and this is why I'd like to have specific tests.

On 0306 I've tested a lot with 2-3 drivers and with 12 players:
• On a 2 players race it's usually lag-proof even if players with bad connection start the session with the rest (by lag-proof I mean no teleporting and almost no weird escenario like "shock not working" or "fake bomb pass").
• On the range 6-12 players I expect hilarious things to happen. Maybe no teleporting on 0306, but many situations like weird bomb pass or electroshocks working later. In the videos I posted before you can see just a bit of this, recorded by Giacomo who is from Italy and has great connection and also lives near from Dolo who was hosting. From my perspective it was a bit worse than what's shown in the videos. In the videos it looks like there's zero lag (because it has zero teleporting).

To be honest I'm not entirely sure if it depends more on which players are in the session or simply in the quantity of players, I always thought it was a bit of both things because more players in the room means more chances to find players with bad connection. Do you have any tips that could help us determine which of these 2 variables manipulate the other?

I'd appreciate some hints here on the last thing I've said. I've been doing the tests, but ended up with no relevant/accurate info. I mean, I don't want this info to be irrelevant or end up producing a wrong consequence caused from a wrong way to perceive info.

Phantom @ Jan 31 2015, 03:55 PM wrote: Do you have any tips that could help us determine which of these 2 variables manipulate the other?

None I can think of, If was asking just in case you noticed it from experience.

Anyway I've just compiled a new test build yesterday. I have not merged yet the features of latest official release, it still remains closer to my last E / F / etc. builds. The time to pack / upload it and you'll be able to try.

I hope the weapon behaviour will be improved since I've now understood a possible issue in the way it was managed, and I have had a new idea for the lag management that could be both simple and rather stable/efficient. I was quite excited about it so if you can get me some feedback as soon as possible I'll appreciate it.
After a test I realized this idea was unfortunately wrong, so I'm already a bit less excited...

Then in case there's no major issue I'll be curious to see if it still works with a lot of players in a room. But this part can wait for a few days if needed.

As soon as possible! Yes!

Ok, here it is.

Re-Volt_15.0203.7z

As my last builds it's just the exe file so it should be used on top of a v1.2 install to work (not necessarily the latest one though, I think).

I just hope it works as I personally had several issues while trying to try it... but they were coming from my own GFX card.

Jig are you sure you haven't sent us the same build you sent back in 14.0314. It's strange, the file size is exactly the same (1.288.704 bytes), version in frontend says 14.0314, it has the same debugger (the weird numbers at botton of screen) and it has the same bug than 0314 had: Lamp poles dissappear when you drive near them (can be seen in Hood1, offline and online). Is this the right build?

Yes I'm sure, but you were right to worry.

Size may not change between build because of padding inside the exe, I presume (if you open an exe file in an hexadecimal editor you'll find several places filled with 00 - which were BTW very convenient for binary patching at the time of WolfR4 ).
I have not updated the version number inside, and I still don't know where this lamp pole bug come from (but if it doesn't occur in last official build then I may find it when merging Huki's code, otherwise I'll keep it for later anyway).

Well, in fact I was going to tell you to rather forget about this last build, I have found several mistakes that make it quite irrelevant. I'm preparing another one, I'll let you know when it's ready.

Oh interesting. I understand.

Mini-bug found by Santi which also exists on 0306 and 0314:
When you enter in the game, online or offline, and you go to choose a track in the Frontend, this tracks appear with Lenght Unknown. Only the first track. Changing track displays lenght correctly.

Also, why are the debugger numbers present and how should I use them? Is it relevant to pay attention to it?

Offtopic question:
Do you specialize in some part of the code and Huki in another one, or you both are able to work with any part of the code independantly?

Ok, time for a new build. This time I have already tested it myself a bit to avoid shooting in the dark as with the previous one.

Re-Volt_15.0205.7z

There's a small changelog included.

The main change you may notice is that remote cars will be drawn in ghost mode when the lag is too big (>500ms). I had this idea since the current code will be more likely to avoid weird remote cars behavior like flying or jerky direct-location-copy. This way should make clear when a player is lagging, and you know you can't expect a normal behavior with this car.
Let me know what you think, and if the 500ms value sound appropriate in practice...

I'm also curious to know if you feel the weapon behavior is better now.

Just for info if you run it in DEV mode you should be able to display some extra info about remote players with the Enter key (the one on the numpad), like the current ping, or the delay that the new code is currently allowing for lagging players to avoid extrapolation blatant issues.

In case one player has a bad connection and there's a lot of players in the room, he may not be able to send 6 packets per second to all players. That case is not supported, let me know if you think you faced it. Using the extra info panel mentioned above, this player should see number 0123456789 increasing somewhere on the top-left (that's the number of remaining packets in his Send queue).

EDIT:

Phantom wrote:Also, why are the debugger numbers present and how should I use them? Is it relevant to pay attention to it?

The numbers in place of the radar? There's were here to try to fix the number of lap issue. I've removed them in this build. The most relevant info you can get are in the Numpad Enter panel in DEV mode.

Phantom wrote:Do you specialize in some part of the code and Huki in another one, or you both are able to work with any part of the code independently?

It depends, for example at the beginning Huki was the only one to play with the online code, as far as I'm concerned I was not much interested into that part. But for this issue I'm now taking it in charge while Huki concentrates on the new libraries for the OpenGL build. So yeah, it really depends. But it's usually better when we specialize otherwise we have to discuss / explain ever change and it takes a lot of extra time. And when you're working on a feature it's not always convenient to have someone else modifying it on the same time.

Oh man, to bad I can't test with Phantom during this week because late shift.

Too bad actually. Let me tell you as soon as you have a little feedback... I'm really curious.

Just for info the Lamp Pole bug is now fixed. A stupid typo in an little optimisation I had brought to the models loading that took me hours to point out.

Also, avoid to save Replay with that build as there'll be issue with other builds.

I have also merged the whole last Alpha code with my current build so the next build will be a lot closer to the official one. And if we get any interesting results they could be merged sooner in the official (along with the track search feature and camber, etc.).

Ok, Test Results time.



We've tested with 3 and 5 players today. In both tests there were no cars teleporting back and forth, also no flying cars, which is excellent. The position of other drivers was more accurate and smoother than a typical 0820 race. It's similar to 0306 in this aspect. The weapons worked correctly, electros and bomb pass worked 99% of the times. We've tested specifically these 2 weapons because are the more jerky in 0820.

The 5 player race had some little issues.
The only issues we had were the following and all happened in 1 session (13 races, 3 laps each):

1.a - This is not an issue, this is a doubt. We notice the ghosts appear quite frequently when a race ends. My guess is that either a player is always minimizing the game for 1-2 seconds when the race ends or the connection always weakens at the end of the race. I'm not sure yet, this needs some test.

1.b - This is not an issue, just a suggestion: About the beginning of the race, since there's no connection yet while we're "Waiting for player", all cars blink like ghosts except my car. It happens all the beginnins of the races. It is pretty logical that this happens at the beginning but I wonder if you can make this pre-race ghosting not occur for the players that are ready (because it is not necessary and it's quite weird seeing cars ghosting at this moment even when everything's okay). The less ghosting we see, the better right?

2. This is an issue. In one occasion the old radar bug came back. What you see in the picture is me being 5th in reality but the radar showed Sete many meters behind me and therefore it shows me 4th. You can see from my F3 camera Sete's real position.





Note: GameRanger has a ping detector. It always shows Sete with a ping of 470 to me. Maybe this is the reason why I see him ghosting almost all the time and he sees me ghosting practically all the time. For me all the other drivers looked okay without teleportings and only occasional ghosting. Later Setemeny commented that all cars constantly flash/blink for him.

3. Again Sete. In one occasion he looked like driving under the floor. I am always controlling Sete with F3 camera as you can see. And you can see he's not there lol. Also from my F3 camera you can see he should be right next to me, but the radar shows a complete different value.







4. In one occassion (Museum 1) I had to restart the race because all cars started the race except Kosztya. He couldn't move even after 3,2,1,GO. After I restarted, some cars looked like if they were still racing the previous race and attacking each other for some seconds, then they stopped and their game was restarted. It's like it occured a delay in the restart order. This picture was taken after I restarted and while we were waiting, somebody exploded. I need more tests to see what was the cause of this. I've no idea how to reproduce it.



5. The last races were worse than the first races. Yes, it's like if the lag is acumulative or something. Usually the worst things happen at the last races and not the first. Sete's car was okay during the first races. It's not until the 4th-9th race when the strange things like driving under the floor or Kosztya's not starting start to happen. There was no big crazyness in any of these tests, nothing huge or very annoying but minor things like these.

Judging by today's races I can answer your previous question by saying that the player who joins is actually more important than the number of player. Because most of these issues seemed to be triggered by one person, Sete. And that got worse and worse with every race.

However, Jig I must congratulate you because your efforts are definitely producing changes. Despite these little issues, this build is already better than 0820 by far and I think better than 0306. Because of the stable car positions like 0306 and the fact that weapons seem to be less jerky. Now we still have to face the Fire Test when a chance for a 12 player race arrives. I'll let you know as soon as this happens.

But overally the lag management was better than a typical race with 0820 and these improvements you created have the potential to replace the 0820 standard I think.

Ah, little note: did you know the Pole lamp bug doesn't only happen in Hood1? It happens in almost every stock track with different objects.

Thanks for the test and report.

In order of what came to my mind... The ghost feature can actually be a bit hacky, I have added so that you can see whether it's worth keeping or not, not really as a well polished feature. Currently I would not really concentrate on these issues (except if they were revealing another interesting info). Your feedback about it is already interesting anyway.

About the radar bug we (unless I have not read Huki's own changelog carefully enough) have never really been able to fix it, and I'm afraid I no longer have much new ideas to improve this part of the Re-Volt design...

>> GameRanger has a ping detector. It always shows Sete with a ping of 470 to me
Ok, so not totally surprising you have some glitch with him. I'd be interested in a few other ping values BTW, so as to see what can be considered as "regular", and as the "min" and "max" player-to-player ping values when playing over Internet.

>> Later Setemeny commented that all cars constantly flash/blink for him.
Can't he run in DEV mode and take a few screenshots of the "Numpad-Enter" panel I mentioned previously when he is lagging and there's several remote players in the race?

Point 4. is bad, but even on my side I can't count the number of times my two computers lost sync eg. on track restart, etc. during my test with artificially degraded network (or when I wildly close Re-Volt on one of the computer). The online session management was rather Huki's job and is therefore a part I have never really looked at yet.

Point 5. is interesting. Again, screenshots of the Enter panel may (or may not) give some clues...

>> Judging by today's races I can answer your previous question by saying that the player who joins is actually more important than the number of player.
That could be rather particular to this build though. Because if a player has no sufficient upload bandwidth to send 6 packets per second to each other players his car will likely look worth for everyone else than in previous builds. On the other side, if one player can't receive data fast enough, it should not prevent everyone to send data to other players (contrary to previous builds). I don't know yet what is best...

To conclude with the Pole Lamp glitch, yes it actually occurred on every instances. Each instance has a bounding sphere so that we can do a fast test: if I don't see the sphere I don't see the instance at all. A typo I made in my modification made the sphere smaller for each instance, so they were likely to be rejected from drawing even when still partially visible.

Ok I'm going to keep testing and bring you new test results as soon as they are available.

2 things to consider:
1. The new ghosting feature is actually great. It's a brilliant idea. You don't need to and you should not remove it. It's really worth keeping it. Everyone in the room appreciated it, the only complaint was with the fact that it occurs at the beginning and we think it would be nice if the ghost blinking could be 'hidden' before the race starts and start blinking after 3,2,1,Go and not before.

2. The lag management has already been optimized a lot thanks to your precious work in this build.
But there is the chance that it cannot be improved further and we will have to accept it. It's not bad news, because comparing it with old versions of the lag management it's better than any. We can say we currently have the best lag management we could ever have.

So we can say that this thread worked its purpose which was improving things on the lag management compared to 0306.

Now, I'd like to ask you something. You know that when I made the review of the current situation I mentioned that in the multiplayer ambient we often prefer to play with older versions because the game style in an old version (things like contact with the surfaces, floor, walls, etc. and the steering which are only important for online players) are more enjoyable than in newer versions. Particularly we mentioned the 11.0421 because we're sure it's bug-free when it comes to this aspect. But I also mentioned that the lag management in 0421 is certainly not as good as in 0306 or other new versions because it has the typical teleporting where the car moves back and forth and a primitive late joining system, but despite this the gameplay is still awesome.

So I take the audacity to ask it to you:

Is it be possible to combine an old version like 11.0421 with the new lag management code from your build? You can see more details about my review of the current situation on my 2nd post of this thread. Even without the many features of the new versions, it would be interesting to see a build offering the old physics and steering sensation but with these lag management optimizations from the last build. We could even sort this problem out.

@Passion: Can you imagine it? The old physics with less lag. It would be the dream of the online player.

I leave it to your consideration Jig.

Phantom @ Feb 8 2015, 04:13 AM wrote:Is it be possible to combine an old version like 11.0421 with the new lag management code from your build?

I checked the changelog and well... I'm afraid it almost sounds impossible.

There's definitively something to check about this, but it's not easy, all the more as it appears to be very subtle. If at least we're absolutely and definitively sure that one version doesn't have any issue as far as the physics behavior is concerned it's already a good point. Now we would also need a version where we're sure the glitch is definitively present, to narrow the changelog window.

Otherwise it sounds like an impossible mission...

Oh, BTW can you always mention full version like 11.0421 instead of 0421, that way when I'm reading I already have an idea of how old the build is (yes, I don't remember all builds just from their last number )

jigebren @ Feb 9 2015, 09:11 PM wrote: Oh, BTW can you always mention full version like 11.0421 instead of 0421, that way when I'm reading I already have an idea of how old the build is (yes, I don't remember all builds just from their last number)

Yes, sorry.

First, I'll look at each build again to look for the origin. Last time I did this I ended up in a huge confusion because the changes were so subtle that it's only easily noticeable when you compare a very old version with a very new version, not when you compare 2 very close versions.

Now that I'm thinking, if merging sounds imposible can you take alpha 11.0421 and start adding certain lag optimizations? We might be able to:
• Know if the lag management optimizations were the cause of the different feeling.
• Or if these optimizations didn't affect at all and it was part of another physics difference.
• Finally get a version that it's free of collission/steering/feeling/car behaviour differences like 11.0421 but with a better lag management.

Some interesting info is that the different feeling is perceived by players with totally different contollers, keyboards, joystics, driving wheels. Especially the more stable steering in 11.0421 is something we admire.

We keep comparing 11.0421 with new versions and even if we want to deny it, the game play feels damn excellent on 11.0421. Races are more enjoyable and exciting, our performance as drivers is more stable. We're total and definitely sure that whatever caused the different driving feeling, it's not present in this build.

Phantom @ Feb 10 2015, 09:28 AM wrote:We're total and definitely sure that whatever caused the different driving feeling, it's not present in this build.

Yes, but as long are we're not 100% sure that the subtle differences are actually offline physics differences and not online lag management, it'll be hard to progress... And the way to proceed is not updating the 11.0421 lag management, but actually playing offline only and see whether there's noticeable changes or not, and with which build(s).

Once again I have the feeling we're mixing these two issues. I'm not against treating both in the same thread, but if we're focusing on the physics subtleties of your own car, then for a while we have to forget about the online racing, even if that's what you enjoy the most...

Oh, and about online racing, do you have other regular player-to-player ping values from Gameranger?

jigebren @ 11 Feb 2015, 06:55 PM wrote:for a while we have to forget about the online racing, even if that's what you enjoy the most...

Yes. I know, I know. Playing offline to detect feeling differences in physics is the next way to proceed even if that's not what we enjoy the most. It's hard to believe that somewhere in the middle of the development it happend and I'm not able to detect in which build exactly by the changelog. Detecting which build and which feature(s) would be the easiest way to fix it. If only I could manage to do it..

I've been feeling a bit depressed lately about this to be honest. I know that the different feeling in the current versions of v1.2 is the cause, and seeing that it is impossible to fix it increases the sadness and kills the huge hope that I usually have.

Passion and other players feel the same. For us there's no comparison in the fun that the 11.0421 game style provides and seeing that with new versions we don't have the same feeling and it is unfixable makes the whole situation a bit depressing. I know we should try not to lose hope and think positively that it will be fixed at one moment or another but.. it's hard..

Do you have other regular player-to-player ping values from Gameranger?

Regularly 90-100ms with people from my same country or nearby.
Higher than 350ms with people from other continents, including North America.
Sometimes Down, which I don't know what that means.

And the way to proceed is not updating the 11.0421 lag management

I know you're right. This is just my idea in case the issue is indeed absolutely and definitely unfixable in new versions. In that case we older players could at least enjoy the excellent gameplay from 11.0421 between us with a bit less teleporting. But I do know it's not the right solution.

I tried to play a bit lately with older version to compare the feeling with newer one. Honestly I can't say I'm definitively able notice anything different (I mean, noticing a difference in such a way that I could be perfectly sure whether it's the "good" or the "bad" gameplay/gamestyle in a blind test). Even though I do believe you have something to complain about, it's hard to get rid of any placebo effect when it comes to "feeling".

I also recently checked old discussion about it, unfortunately:
1. from the beginning the distinction has unfortunately not been made between offline and online behavior, which make almost all bug reports irrelevant so far.
2. nobody can really say: this build has no offline issue, but the next one had it, so we can be sure it has been introduced in this next one.

Maybe a reason for point 2 could be that there's not only one cause for your concern. Anyway I think Dolo said that one of the test builds had the proper gameplay back again, I don't understand why you still not happy with my latest build then...

Last point I wanted to mention, which may or may not be quite relevant here. Going back to the time when I was working on the so called "GoGo bug" and "Tuxabug" issues to fix them in WolfR4... they were related to the timer management in Re-Volt (Re-Volt makes some roundings to avoid overflowing the 32bits timer variable), during some test with Arto we came to the conclusion that some different values of cpu frequency can led to slight but noticeable different timer speed in Re-Volt. In some cases he clearly noticed he could artificially make better lap record.
More recently (even after the first complaints about the Physics), I finally found out another nasty bug that was likely the real cause of what we saw with Arto (as at the time we were unable to totally understood it).

I'm almost sure these 2 modification to the timer only contribute to make Re-Volt closer to the way it has been designed and the way it originally ran on a 1999 computer. The first was mandatory (to avoid go go bug, etc.), the second occurred after the Physics complaints. Maybe it could have make the Physics a bit slower that before it were fixed though, which you feel is a bug, while the previous builds were the ones actually bugged. I mean, even if you run the original v1.1 Re-Volt on your computer, you don't have the behavior of Re-Volt in 1999 (the two bugs I mentioned above are the perfect example of that). You would have to get a 1999 computer to really see how it originally felt, and I presume nobody really remember that...