MTM2.com

EDIT:!:
The description given in this post is technically inaccurate in essential points.
The described effect corresponds rather to what may happen in some other racing games.
The correct analysis for MTM2 is contained in another post lower down the thread.
This post has been left unchanged to keep the wholce thread legible.
END EDIT:!:

In a recent tournament on the track Sandy Morgan Gulch (by Phineus) I had an incident with Drive2Survive. The incident is not specially related to the track which is named only to make it easier to follow the explanations.

RA's version of the story

Since I complained about the incident, D2S had to give his version of the story.

D2S' version of the story


Adding insult to injury ("he simply missed the corner" - phew!) was not meant to calm me down.
I went a bit loud, as moose do when they are irritated, lol.

A couple of days later I decided to give the events a second thought. I had raced D2S many times before and always appreciated him as a fair racer. Was it possible that he had changed that much?
I analyzed the situation using all the knowledge gathered from my recent trip to Australia. Maybe D2S got too much sun through the damaged ozone layer over there. Combined with an overdose of Vegemite for breakfast this could indeed be the explanation...or rather not? An even more compelling analysis of the incident involves the understanding of lag.

Lag
Because D2S is located in Australia and RA in Europe the time it takes for signals to go from one computer to the other is not negligible even though they travel at light speed (300'000 km per second). By cable it takes about 0.1 sec to cover the distance including all detours. Add to this various processing times at different nodes on the way and you end up in the 0.2 sec range.
(Things are even worse if the signal travels by satellite. Geostationary satellites are located 35’786 km above sea level. This means that the time up and down alone takes 0.24 sec.)

How does lag affect the game?
Let's say D2S computer sends a message to RA telling "at time t0 my truck is at position x0 and travels with speed s0".
At a speed of 90 MPH D2S will travel 27 feet (or one and a half truck length) before RA receives the signal 0.2 sec later (time t1). Because the game would be unplayable if the sent truck position would be displayed as is on the receiving machine, MTM2 tries to predict the truck position at current time based on the time elapsed since t0. The game assumes the truck continued in the meantime to travel at the indicated speed s0.
If you like maths, the simplified basic equation is estimated_x1=x0+(t1-t0)*s0
(simplified because most games assume rotations to continue as well).
<center> </center>
While fairly reasonable on straights (estimated_x1 close to real x1), the extrapolation miserably fails for the time immediately following the begin of a turn, as illustrated below.
<center> </center>
During a short time the extrapolated truck positions will thus be significantly wrong. But since on each computer collisions are detected based on the extrapolated positions of the other trucks, we can indeed have collisions existing only on one computer and caused solely by lag.

Conclusions
As improbable it looked both racers gave a true version of the events from their perspectives.
D2S did not touch RA and
D2S kicked RA down the bridge.

Nevertheless I am sorry for getting that loud, D2S.

Lag exists and plagues many other racing games. For the racer it materializes in the form of an invisible battering ram protruding out of his front bumper: the higher the speed the more it reaches out.
I think we should have some sort of a lag-aware driving code of conduct. Otherwise it is too easy to use the phenomenon to its own advantage. Some game series (e.g. GPL) went quite far in this respect.

What do you think?

just one thing, MTM2 actually lags worse on a LAN lol.

Ive done it direct cable to cable 1gbit ethernet. no go betweens, simple crossover cable. 0 LAT. ALmost every single hit resulted in a lag shot lol.

THats only thing i gotta say.

Slayer wrote:


That surprises me. We've ran MTM2 in numerous LAN parties with no lag at all in races with 6+ trucks.
Just banned custom trucks (repaints allowed).

That's a fair analysis.. I've been dealing with 'ghosting' trucks since day one.
Racing against Fila was the worst, him being in Romania and me in the US, and hooked up to a land line... One second he'd be beside me then the next second 30 truck lengths ahead, then behind me, then i'd pop off into oblivion. Passing is next to impossible for me. I'm used to it.

Interesting topic RA

Privately in introducing me to this topic you have enquired as to whether I was still upset over things, based on my "shortspoken" responses to you when this was discussed between us over PM at the last tourney event. In fact I am not and was ready to make any range of jokes or light-hearted insults in response, but was wary of how they would be received (sun and Vegemite remarks clarify that now of course). I was greatly upset after your outburst at the end of Sandy Morgan Gulch, and that you would jump to and go away holding those views. You don't know it but immediately after our clash I was gathering evidence to understand for myself what had happened. After the tourney close and your remarks I was fired up and began preparing to send a missive built around my articles to cut you down to size. Being a finicky typist though, I had time to consider my own rage (lol) and came to worry that being confrontational about it might only achieve to fan the flames, given that you seemed ready to level that the whole world was against you that day. Eventually I decided to put everything aside and give both of us time to cool off, determining I'd only bring it out if it was still an issue you held to next time we met. When in fact you PM'd to explain your new understanding and apologise, I was cautious about picking my words to respond.

Okay, with that useless padding out of the way let me work through the analysis RA has given and present what I found for interest's sake.


It is misgiving I think to surround that text with a quite box, as those are not all the words I remember giving. I would not have agreed that I was behind RA, in fact I "knew" I didn't hit him because if anything he was behind _me_ and not visible on my screen.

My full story is something like this.

The race grid has pole and second positions side by side. RA and I were lined up with him on the left and I on the right. As we know with MTM2 starts the trucks are simply held in place and then released at a pre-determined time giving everyone a no-brainer perfectly timed start. What I saw as my truck started moving was RA's truck remain stationary for a further moment or so before it began to move, naturally with not getting the same jump he dropped back a little and out of my sight of my Chase Near view. Through the turn 1 left-hander I went wide around the corner so as not to cut across the bow or into the side of RA's truck taking the inside line into the corner. On the approach to the right-hander turn 2 I could see we were roughly side-by-side (RA on the left of the road and me on the right) but this time I had the inside line going into the corner and figured that turning onto the bridge I would slot back in front of RA again. I saw RA hanging wide at the entry to the corner just before I turned (losing sight of him) and flung the truck over the bump on the right side onto the bridge, carrying across to the left edge of the bridge (the diagrams RA presents is being quite generous in implying that I stick to my side of the road). I will admit the maneuver was with no regard to racing room but on my screen it looked like I would have jumped cleanly in front of RA without bumping into his side, and certainly not his rear. As it unfolded however RA was not behind me on the bridge but seemed to have come off at the corner.


That's definitely true as far as it goes. I'm not sure it's the correct explanation for what happened in our game however. Positioning error will occur when there is a significant delay between the arrival of packets. In such case, as you say, the game relies on predicting the course of the trucks based on their previous know velocity and position, until the next data packet comes along with the actual position from the other player's computer. If this delay is significant, then yes, the truck may continue in a straight line when the player has really begun turning, and so on. I think we have all seen instances of a truck 'jerking' rapidly as the error from prediction is corrected each time the latest packet arrives, or when a longer temporary delay occurs that an opponent's truck appears to drive off a bend or something, only to warp back on course a moment later. However often we can see our opponents trucks moving quite smoothly as well.

Latency is the time taken for a message to travel from my computer to yours as you have explained, but the way you have applied it is more like assuming that time is the delay between our computers receiving each packet of up-to-date information. In fact once the game starts, each computer should be firing off packets rapidly (once every frame I should expect, pending the throughput of the connection) and these should arrive at the other end in a fairly steady stream. There will be occasional hiccups and packets being lost or arriving out of order of course, but all things being equal the distance involved should not affect the regularity with which packets reach their destination.

Instead what latency causes is the games to be time-shifted. I mentioned that at the start of the race, I saw RA's truck stand at the line a moment while mine set off, so in my game I appear to jump to the front. For simplicity we can assume our PCs synchronise so the race start occurs at the same time for both of us in our respective parts of the world; however it still takes that period of time for the packet that says your truck has begun moving to arrive at my end, so I see you truck stand still until it gets here. Once it does, the updates to your position, speed and direction should keep coming at a steady rate, however your actions are now playing out on my screen a fraction of a second behind your actually doing them (and behind my game play too) - and likewise I imagine you would see my truck fall behind on your screen. This is how we have the situation that you say I was behind you on the approach to turn 2, and I say I was beside; you say I hit you in the rear and I say I was actually in front.

I'm not necessarily refuting your analysis that lag and the game prediction could have been and is a factor, I'm just saying there is another effect which I think in this situation is the greater consideration. If you need more convincing, check out this:

In the round after our tourney race, I was matched to race Kmaster. Given your complaint that I had bulldozed you off the road I approached this race trying to be very courteous and leave him room. After what I thought was a cleanly run race I asked him if at any point we collided, which to my chagrin he answered yes. I saved a replay of the race, and asked him to likewise save a replay on his computer and send it to me (which he did, thanks K) so I could compare how the game played out for each of us. Unfortunately, when I raced RA neither of our computers captured the collision in the replay, but Kmaster and I had more luck with this one as we were close all race and it saved more laps.

Here are the files:
Replay from D2S
Replay from Kmaster

You will also need to mount this pod file in addition to the Sandy Morgan track to make the replays work since for some reason they mis-record the track sit file name (all the pod contains is a renamed sit).

The replays each show the entirety of laps 2 - 4 which was the race finish. It's a bit confusing because the replay saved on my game has done the classic thing of swapping the trucks; Kmaster in Stinger actually appears as Black Stallion and vice versa for me (Kmaster's replay preserved the associations of the trucks correctly). However, I can identify who's driving is whose because on lap 2 I made a mistake to fall off the bridge and had to make the slow high jump back onto the road, which let Kmaster through to the lead. Thanks to this there's lots of overtaking action in the following laps.

Watching the replays revealed that where I thought I was giving sufficient room to Kmaster and executing clean passes, I was actually giving him a hammering at more than a few points. If you watch it for yourself I'm sure you'll find the differences striking; however I picked out one incident to illustrate with screenshots. On the approach to the second loop corner, I was side by side with Kmaster (as I saw it on my screen) with the inside running into the turn. Much when like when in the race against RA, I turned in at high speed to slot neatly in front of the other truck - at least so I thought, and as far as I saw on-screen. This is the screenshot from my replay, riding with Kmaster (remember the trucks have been swapped) which shows me clearly and cleanly in front.

<center>
Riding with Kmaster in replay saved on D2S' computer</center>

After the overtake however I saw (via the track map) Kmaster slide off the corner and he answered in the affirmative when I asked if I had pushed him. I slowed to let him catch up and we continued.

Now viewing the action as scribed by Kmaster's replay:

<center>
Riding with Kmaster in replay saved on Kmaster's computer</center>

My truck appears much further back in terms of road position than I saw it on my computer. My higher corner entry speed and line taken now travels through the position occupied by Kmaster's truck. His computer detects I hit him which sends him off the corner. I never even felt the bump because in my game, his truck never contacted mine.

When I saw this great difference I was horrified and I seriously hope it's not that bad all the time and against all competitors. Bottom line is that, at least when I am playing against the Chilean one, latency on this day meant the difference in truck positions seen on each screen was of the order of two to three truck lengths. It will vary for other players and the conditions of course, but I have little doubt this is virtually identical to what RA saw occur in his game. In fact if I had deliberately aimed at ramming RA off the road, I would quite probably have missed due to this


Yep. Two accounts of the same story can both be right.

Latency works both ways. Just as I was running into the back of people when I saw myself alongside or in front, likewise I might appear clear behind on someone else's screen and on my screen they are running oblivious into my side or back. In each case the 'crash-er' may not even get a glimmer of the impact.

I don't see how much can be done about it. For my part I will be trying to be more aware and give room to other players in my rear quarters as it seems likely I may not be 'finished' with driving around them but of course its hard to keep track of that. Just gotta accept that sometimes you're the windshield, sometimes you're the bug...

Sorry for misquoting you D2S.
Since direct communication was ...errr difficult that day I got your statements from second hand.

Seems I made some pretty wrong assumptions in my analysis which was based on similar articles written for GPL. Maybe MTM2 does not display forecasted positions after all.
I need to look further into this. Thanks for your replays with KM which will help getting a clearer picture.

I just read this thread. I saw it last week allready but it was to long a read for me lol. But i got curious and read it anyway. Some remarks on what i think of this...


I think a lag-aware driving code of conduct is very difficult in mtm2. I play GPL as well and i know that in GPL you have to be very aware of this lag thing.
Why ? Easy... because GPL (and MTM2) are older games. They date back to the late '90s. When playing online they don't calculate the lag time in themselves like newer games do. If you play GPL offline you can hit an opponent (ai ofcourse in that case) without having to much damage. But if you hit a real life opponent when racing online it always results in a hughe crash. because GPL does not calculate the lag. One moment you are not touching eachother and the next your cars are almost in eachother. GPL thinks you must have slammed hard into eachother and gives you a horrific accdident. Same goes for MTM2. More modern games (like LFS) calculate the lag in and know if a crash was that hard or not.

Now why is there such a code in GPL and not in MTM2. Several reasons if you ask me.
1. most important is the damage model. When you have a crash in GPL (as i said before, even a small rub turns into a big crash) your car is seriously damaged. So badly damaged that finishing your race is almost impossible. When you play with "real damage" you can't even do a shift-R to fix your car. And in GPL there is no pits where you can go to get your car fixed. So after even a minor crash there is a 90% chance that your race is over. Nobody likes that ofcourse so GPL racers became very cautious because nobody likes to see their race ending after a crash. GPL is one of the best games out there if you ask me. It's only this lag and damage thing that is wrong with the game imho.

2. In GPL you have time to be cautious. A descent race takes at least 45 mins so you have time to wait for a clean pass. In MTM2 a race lasts approx 10 mins so you have to get by the guy in front of you asap.

3. Now american racers don't take this the wrong way pls lol. American racers are ... (how shall i say it) ... a little bit more aggresive. How often don't you hear : "Rubbin' is racin' " and stuff like that. Nothing wrong with that but it's completely different with how european racers race. Just look at the difference between F1 and CART or FIA GT and NASCAR. In Europe most of the times rubbin' is no part of racin'. But since we all play a game that actually situates itself in the states (because that's where monster trucks come from) i have adapted and use that style as well. You won't see me rubbin' in other games tho lol.



I'll keep this in mind D2S... If you are not slamming into me you are trying to




I don't see how to change the way we behave on an mtm2 track. We'll just have to live with lag i guess. Sometimes it works on your favour, sometimes against you. To me it's easy... sh*t happens. Live with it

Did we ever race?


lol

My "favourite" lag situation was in a tourney against a fellow SLO racer. We both managed to win the race on our PC


When the synchronizing happened the results was that i won but he couldn't believe it and after watching his replay i was a full truck behind him while on my PC i was half a truck ahead.







Rumbling with me was the worst however. I never knew why i was such a bad rumbler until i learned that if i want to win anything i am supposed to lag shot the opponents

Because any other way it didn't work. If i didn't lag shot the others i would get a lag shot myself so it was basically impossible for me to play a rumblet the proper way.

old game timer glitch

when online, when racing, hit the alt key, which freezes your game, wait like 30 or 40 seconds, and hit the alt key again, hit F4 to go back to full screen. Now if you don't lose any time, or actually catch up a bit, you will win because your total time is lower than everyone elses, without the hassel of figthing for positions

Yeah I know its cheap, but, meh, its neat.

BTW, if anyone does this during a tourny Ill personally pull your eye balls out through your butt.

Never knew that!


lol

Here we go for analysis part 2
Experiment
With D2S and Mumhra we ran a scripted 1 lap race on Sandy Morgans Gulch.
On specified segment it was agreed which driver would drive at constant speed while another would place himself
just ahead of him with respect to what he sees. Positions were swapped to observe each pair of racers.
The experiment was ran twice.

Findings
Replays show what the respective racer saw on their machines.
MTM2 does not display predicted truck positions. Apparently it displays what it gets at the time it receives it.
It is not clear what happens until the next position update is received.
From what can be seen in races when someone gets disconnetcted I conjecture the truck continues its trajectory based on the current speed, throttle and steering. (The truck follows the terrain for sure).
Winners circles show for each driver the time according to the respective local machine.
For this reason Winners circles always match.
Relative differences between trucks (i.e. differences in lead times) are of about the size of the ping.
In MTM2 the ping is shown in the latency display in the multiplayer screen.
(To actually read it you must edit the monster.ini in the MTM2 system folder and change the line "latencyDisplay=2" to "latencyDisplay=1". Otherwise you will only see colored bars. )



Analysis
Races start on non-host computers once they receive the race starts message from host (or rather the message "race starts in 3sec").
Because the local time is shifted by the lag at start, non-host machines will display the host's truck at about the correct place in time (w.r.t local time), i.e. at correct relative position as long as latency remains about constant.
There is a mechanism compensating computer clocks running out of sync. This was not an issue in our scripted races but was affecting (in a minor way) the race between D2S and Kmaster.
In particular, if the local truck passes the finish line in front of the host, it usually means the local truck really finished ahead of the host.
Even though the relative position of the host's truck is correctly displayed on the other computers, its trajectory ahead is already driven. The same is true for the other players, but there is an assymetry with respect to the host. On his computer the other trucks are displayed with a time lag corresponding to the latency (or ping time), i.e. the time for a signal to do a full loop.
We could verify this on the replays of the scripted runs by comparing the CP crossing times at various CPs and then comparing the time differences:
<center>
<a href="http://mtm2.com/~forum/images/rocketalceslagpic3.jpg"><img src="http://mtm2.com/~forum/images/rocketalceslagpic3.jpg" width="600"></a>
</center>
The ping between RA and D2S was about 0.5 sec (i.e. 0.25 sec one way).
D2S starts his race once he got the signal that RA (the host) got green light, i.e. 0.25 sec after RA actually started. RA sees D2S moving once he got the corresponding signal from D2S, i.e. 0.5sec after his own start. On RA's machine it looks like D2S is a very slow starter.
On the famous straight at a speed of about 90MPH D2S placed himself just ahead of RA.
At the same place RA sees himself 2 truck length ahead of D2S.
Number crunching on the replays' data showed that the difference in lead times between D2S and RA indeed correspond to the latency of about 0.5 sec. This means that whenever D2S saw himself ahead by 0.2 sec, RA saw himself leading by 0.3 seconds.
The same effect can be observed between Mumhra and D2S with about the same latency (0.5 sec) , and for Mumhra and RA with the much lower ping of about 0.1 sec. I created a synthetic replay <a href="http://mtm2.com/~forum/images/rocketalceslagpic3.zip">rocketalceslagpic3.zip</a> by superposing the replays recorded on my and D2S' machine from start to CP2. In accordance with the above, both images of RA's truck (the host) are at about the same place while the images of D2S' truck are separated by several truck length. Switch on names ('N') to Watch the replay. Also observe the rubber-banding effect when D2S almost stops near CP2.

The important thing is that in close racing one cannot expect the other to take avoiding actions because the other's trajectory is already driven for the latency time. You can think of a shadow trailing behind your truck that may hit the other trucks on their respective owner's computers.

I had an incident with Kmaster where I passed him while he was recovering from a spin. I passed him cleanly but due to the high lag my shadow hit him on his machine. Because of this his truck was accelerated in such a way that he bumped into me on my computer! Stated otherwise, I became the victim of a chain collision I induced.

Rule of thumb
Multiply the latency by 5 to get the number of truck lengths by which relative positions vary at a speed of 70MPH.
Examples:
ping =0.5 -> 2.5 truck length
ping =0.2 -> 1 truck length
Use this information to know when to expect "special effects".

Many thanks to D2S and Mumhra for their support in running these experiments.

[ edit ] side scroll

Very interesting and well executed and written up. Thanks! I'm glad to see a scientific analysis of MTM2 Lag.

I guess we'll have to start designing track with jersey barriers separating the lanes!

Hopefully we'll be able to use a near-no-lag internet in the future.

The last time was September 2003 on a track called Elevation... You and N8 beat me senseless. We need to do it again sometime soon :-)

========

rocketalces you're insane. lol ... seriously, nice work.

Didn't remember

Wow - pretty intense stuff here! Rocket, you've come a long way since LaPaLu, partner lol.

I don't have anything remotely scientific to add to this awesome thread, but after reading through all the post I thought I'd add simple basic and simple for a change of pace lol. With the exception of racing, lag makes this game a heck of a lot of fun. Rumbling is often times extremely exciting, hilarious, unpredictable and even frustrating if you're the one being tossed like a salad on any given day lol. It's great in RnR and CnR as well, sometimes working for or against you and your team, again, adding an element of unpredictability whereby sometimes would-be underdogs emerge as the victors. I know some have said that if it weren't for lag, this game would be a lot better. I think that holds true for racing purest, but for the other facets of this game, well . . . they just wouldn't be quite the same in my opinion.

Lag and all I love this game, and I can't imagine that sentiment ever changing. Just wanted to add my 2¢ lol.

PS - Hi everyone!

Wow, way to make lemonade Cope.