PDA

View Full Version : Online hit detection bug?



jermin122
08-16-2011, 02:25 AM
I've been flying offline (quick mission, for gunnery practice) and online quite actively these days. Since the day before yesterday, I've encountered some hit detection problems, which mostly happened in Russian servers. The problem is that I missed a lot of very easy shots, such as a bandit is hanging 150 meters in front of me doing a shallow climb. I poured about 7 MK108 rounds onto it. But nothing happened. No fireballs, no smoke, no debris. Nothing at all. The bandit kept flying normally. That was weird, I thought.

I've scored hits in this scenario thousands of times in offline practice, so I am confident that I should have hit him. When I reviewed the ntrk of that sortie, I was surprised to see that I actually landed 5 MK108 shells on it. The enemy aircraft was surrounded by fireballs and black smoke. But after the fireballs and flame vanished, the enemy was like nothing had happened on him before.

However, in offline quick mission (I save trk after nearly every sortie for reviewing. And I didn't found this problem), Skies of Valor and RCAF_FB online servers, I don't have this problem.

Last night, I made a simple 1v1 coop mission which was essentially similar to the quick mission I used to practice gunnery. I was surprised that I experienced the hit detection bug again several times when reviewing ntrk.

At first, I thought it might be the enemy player was using some kind of mod. But since I still get this problem in my own stock server. This problem might be more complicated than I thought.

Has anyone experienced this before?

GratedLeeman
08-16-2011, 03:06 AM
Sounds like lag to me...

jermin122
08-16-2011, 03:14 AM
I had the same thought as you at first. But I pinged the Russian server with 1024 byte packet afterwards. There wasn't any packet loss or ping fluctuation. And it happened even when I flew with an AI in my own coop server. There shouldn't be any lag on AI.

Tully__
08-16-2011, 04:11 AM
Both forms of recording used in IL2:1946 had their issues, .trk and .ntrk
So far I've not seen a description of how they work in CoD, but if the record mode for .ntrk is essentially the same as it was in the original IL2 series, the events shown in the track are not always exactly what happened and in some cases simultaneous .ntrk recordings of the same event from two different players will show two different things happening.

The way the ntrk mode worked in the original IL2 is to record position, orientation, speed, state (gear up/down, flaps up/down, guns firing/silent etc..) and damage for each object in game at frequent intervals. On replay, the game engine extrapolates between samples to give the illusion of continuous recording. This means that a shot seen as an extrapolated hit when viewing the recording may have been a very near miss in the original event. Because it was a miss, no damage is recorded for the target object which makes it seem when viewing the playback that apparent hits caused no damage.

If this recording mode still works the same way, the fact that you saw no fireballs in the original online event but saw hits on reviewing the recording is not a contradiction, it's a consequence of how this recording mode works.

I would suggest that it was actually a combination of less than ideal ping and the shots actually missing by only a very small margin.

Edit: Jermin122, could I ask you to review http://forums.ubi.com/eve/foru...2709/m/1271032709#XI (http://forums.ubi.com/eve/forums/a/tpc/f/8071032709/m/1271032709#XI) and http://forums.ubi.com/eve/foru...1026566/m/7401097417 (http://forums.ubi.com/eve/forums/a/tpc/f/5181026566/m/7401097417) regarding signature / avatar file size and dimension limits and adjust your avatar image to comply please http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif

jermin122
08-16-2011, 04:57 AM
Sorry about the avatar. I've resized it to 64 x 64.

If ntrk is problematic, is it safe to use trk to analyze gunnery?

Erkki_M
08-16-2011, 05:36 AM
The recorded tracks do not replay the game exactly as it was played, even the damage doesnt isnt always shown right.

Secondly, in il2, all hits you score are true hits, you only need to hit in your own game, so there is no need to "lead the ping" as in many other games. Howver the Mk 108 has horrible ballistics with muzzle speed of only 550m/s. If you didnt see hits, you didnt hit, and you should double check your aim and sight zeroing distance/weapon convergence range.

Thirdly, the Mk 108 isnt quite the 1-shot 1-kill weapon it was in real life. Worst I have seen was a Spitfire that took at least 8 x 30mm hits (not counting 20 and 13mm hits at all), 9th and 10th made it smoke and 11th or 12th finally made it explode. I have on track too, low pings no packetloss and he was trying to evade me with no warping at all.

You will often score 1-hit 1 kill but depending on what you shoot and where you hit, that might not happen no more often than 40 to 50% of times. A hit will usually at least cripple the target, but I too often need 2-3 hits to make it stop fight back and the 4th, even 5th, hit to finish it off. That is one reason why many prefer the MG151/20, easier ballistics, over 3 times the ammunition and still pretty lethal if you can get your fire to connect.

jermin122
08-16-2011, 07:54 AM
Thanks for the advice, Erkki. I like your frag videos very much. I'm the one asked for your convergence settings.

This (http://sovstats.war-clouds.com/playerdetails.php?id=9627) is my stats page on SoV. I flatter myself that I was a good enough pilot to score those easy shots. I mean those are really easy shots. I score them every day. But I missed it when it happened. That's why it is weird. I will do more tests with a friendly AI.

Tully__
08-16-2011, 07:07 PM
.trk recordings record the player's control inputs (which is why they only work for single player). When played back the game plays the mission as it was played in the original, applying the player's recorded control inputs to the player aircraft and the normal AI routines to all the other objects. Provided all the recorded player inputs are applied at exactly the same time as in the original mission AND the aircraft responds as it did in the original (which it wont if there's been a change in flight model, hence .trk files only work in the version they were recorded in) the playback will exactly match the original mission.

Where you run in to inconsistencies with .trk is if the computer it's running on is straining to keep up with the sim. This can result in player control inputs not being applied until a few milliseconds late. This doesn't sound like much, but will accumulate through the mission and lead to larger and larger discrepancies until in a long mission, you'll see the player aircraft shooting at enemies that aren't there or trying to land several kilometres from an airfield (and possibly several thousand feet off the ground or underground).

The only way to see if .trk files work reliably for you is to try them out in a busy mission over a large city. If they work there, they'll probably be pretty reliable in most circumstances.

Additionally, you can record .ntrk recordings from a .trk playback if you so desire, though the extra data load may introduce the playback errors described above on marginal systems.