missing info Replay Track Broken

[Doughguy]

Well if its easy to reproduce and the spit is borked in that regard anyone incl OP should be able to recreatw a track with said behaviour pretty much within 10 15 minutes...

[LeCuvier]

15 hours ago, NineLine said:

If you see another example, please feel free to share, and what you might be seeing that is not playing back right, I will try the Spitfire some this weekend and see what I can see as well.

Some of these issues can be easy fixes, others... not so much. But as always, happy to take a look if you supply all the info.

Thanks, and enjoy your weekend as well.

I've been following this for a long time, flying mostly Bf-109 and FW-190. These missions are generally dogfights with a substantial amount of turning. Each time I looked at the track, I found it went "off-track" after a few turns. On the other hand I flew a mission with the Hornet yesterday, and the track was absolutely fine.
Seeing all these posts today, I decided to produce a track. I flew a mission with the Bf-109 vs a Mosquito. I killed the Mosquito and saved the track. When I viewed the track, I found myself shooting into the blue sky turning around aimlessly and in the end was shot up by the Mosquito. I then replayed the same track again, and the action turned out very similarly, but somewhat different. And if you replay the track you will probably see something different again. The duration of the replay is always exactly the same as that of the original mission, so the tracks are truncated at different points in the action.
I attach the track, plus Tacview recordings of the original mission and of the two replays. Tacview shows clearly the differences in the course of action, between actual mission and replays.

Bf-109 contra Mosquito.trk Tacview-20250906-Mission 2 Replay 2.zip.acmi Tacview-20250906-Actual Mission 2.zip.acmi Tacview-20250906-Mission 2 Replay 1.zip.acmi

[LetMePickThat]

17 hours ago, NineLine said:

If you see another example, please feel free to share, and what you might be seeing that is not playing back right, I will try the Spitfire some this weekend and see what I can see as well.

 

I stumbled upon this very thread after facing a similar issue multiple times with the MiG-29A, so I'll share all the data I saved here to keep everything in the same place.

Flying a simple 1v1 dogfight between my MIG-29A and a M2000C, I found out that the track diverges very rapidly, to the point that there is nothing in common between the original flight and the replay. I've tested this multiple times, and the outlook of the combat sometimes changes dramatically. The AI in replays completely ignores the flight path and actions it took during the real flight, and develops a mind of its own. A coherent dogfight turns, in the replays, into a mess where my plane still follows the original trajectory, but is fighting against (and firing weapons at) thin air, whereas the target is now actively trying to get a shot out of my plane. This is repeatable for each flight/replay combination. I've seen multiple instances of the 2000C shooting me down in the replay whereas I won the fight during the flight.

For testing purposes, I flew a 1v1, which voluntarily kept long to generate data, and won in the end to illustrate the change in outcome that often happens in replays. I saved a bunch of files for diagonistic:

• ORIGINAL_TRACK is the track saved immediately after the flight. Logically, it should show me winning the fight (it does not).
• REPLAYED_TRACK is the the track produced (and saved) by playing ORIGINAL_TRACK (it should show the issue, and have me not winning - but it's unreliable anyway).

Because neither tracks can be trusted - that's kind of the problem - I also saved two Tacview files:

• ORIGINAL_FLIGHT_TACVIEW is the Tacview file produced during the original flight. Because of the track bug, it's the only file that is trustworthy when it comes to how the flight was actually performed. It shows me downing the 2000C at 095858.
• REPLAYED_TRACK_TACVIEW is the Tacview file produced during the replay of the flight. It has nothing in common with ORIGINAL_FLIGHT_TACVIEW save for my trajectory and weapon employment, and I can be seen fighting ghosts in it. At 095858, I'm shooting at nothing whereas the 2000C is closing in my 6, and tries to shoot me down at 095902.

 

As one would expect, the longer the mission, the larger the difference between the replay and the OG flight. I've had very short dogfights (sub-30s) replay most the time the same, because errors and changes don't have the time to propagate. Still, even on such short flights, sometime the difference is massive (for instance, I had a frontal gun engagement turn from a complete 2000C destruction in the OG flight to a single flaperon loss in the replay. In that replay, the 2000C shot me down afterwards because I flew a straight line - there were no ennemies anymore to worry about in the OF flight at this point). Longer dogfights never replay correctly.

As it stands, tracks are unusable as a mean to save a flight. Tacview is the only tool that can store trajectories and results in a reliable fashion.

 

1 hour ago, LeCuvier said:

These missions are generally dogfights with a substantial amount of turning. 

Yeah, the problem seems to be markedly worse when maneuvering.

 

ORIGINAL_FLIGHT_TACVIEW.zip.acmi REPLAYED_TRACK_TACVIEW.zip.acmi ORIGINAL_TRACK.trk REPLAYED_TRACK.trk

16 hours ago, Doughguy said:

Well if its easy to reproduce and the spit is borked in that regard anyone incl OP should be able to recreatw a track with said behaviour pretty much within 10 15 minutes...

There you go. You now have two example of the issue.

Note that this isn't Spit-specific, it's a sim-wide problem. I could replicate that with any plane and the two examples provided include the 109, the Mosquito, the 2000C and the MiG-29A.

Edited yesterday at 12:32 PM by LetMePickThat

[Doughguy]

55 minutes ago, LetMePickThat said:

 

I stumbled upon this very thread after facing a similar issue multiple times with the MiG-29A, so I'll share all the data I saved here to keep everything in the same place.

Flying a simple 1v1 dogfight between my MIG-29A and a M2000C, I found out that the track diverges very rapidly, to the point that there is nothing in common between the original flight and the replay. I've tested this multiple times, and the outlook of the combat sometimes changes dramatically. The AI in replays completely ignores the flight path and actions it took during the real flight, and develops a mind of its own. A coherent dogfight turns, in the replays, into a mess where my plane still follows the original trajectory, but is fighting against (and firing weapons at) thin air, whereas the target is now actively trying to get a shot out of my plane. This is repeatable for each flight/replay combination. I've seen multiple instances of the 2000C shooting me down in the replay whereas I won the fight during the flight.

For testing purposes, I flew a 1v1, which voluntarily kept long to generate data, and won in the end to illustrate the change in outcome that often happens in replays. I saved a bunch of files for diagonistic:

• ORIGINAL_TRACK is the track saved immediately after the flight. Logically, it should show me winning the fight (it does not).
• REPLAYED_TRACK is the the track produced (and saved) by playing ORIGINAL_TRACK (it should show the issue, and have me not winning - but it's unreliable anyway).

Because neither tracks can be trusted - that's kind of the problem - I also saved two Tacview files:

• ORIGINAL_FLIGHT_TACVIEW is the Tacview file produced during the original flight. Because of the track bug, it's the only file that is trustworthy when it comes to how the flight was actually performed. It shows me downing the 2000C at 095858.
• REPLAYED_TRACK_TACVIEW is the Tacview file produced during the replay of the flight. It has nothing in common with ORIGINAL_FLIGHT_TACVIEW save for my trajectory and weapon employment, and I can be seen fighting ghosts in it. At 095858, I'm shooting at nothing whereas the 2000C is closing in my 6, and tries to shoot me down at 095902.

 

As one would expect, the longer the mission, the larger the difference between the replay and the OG flight. I've had very short dogfights (sub-30s) replay most the time the same, because errors and changes don't have the time to propagate. Still, even on such short flights, sometime the difference is massive (for instance, I had a frontal gun engagement turn from a complete 2000C destruction in the OG flight to a single flaperon loss in the replay. In that replay, the 2000C shot me down afterwards because I flew a straight line - there were no ennemies anymore to worry about in the OF flight at this point). Longer dogfights never replay correctly.

As it stands, tracks are unusable as a mean to save a flight. Tacview is the only tool that can store trajectories and results in a reliable fashion.

 

Yeah, the problem seems to be markedly worse when maneuvering.

 

ORIGINAL_FLIGHT_TACVIEW.zip.acmi 214.15 kB · 1 download REPLAYED_TRACK_TACVIEW.zip.acmi 196.58 kB · 1 download ORIGINAL_TRACK.trk 3.09 MB · 1 download REPLAYED_TRACK.trk 3.33 MB · 1 download

There you go. You now have two example of the issue.

Note that this isn't Spit-specific, it's a sim-wide problem. I could replicate that with any plane and the two examples provided include the 109, the Mosquito, the 2000C and the MiG-29A.

Cheers.

I guess since i only play pvp i have less trouble with tracks/replays not working, whereass pve does create issues.

My guess is that ai still can pull off some stuff regular players can not hence there is discrepancies later in the track?

 

There are few dedicated mp dogfightservers for mp only which could be used for testing.

But it kinda appears the flightmodels sorta could be the root? Or PvE respectively.

Edited yesterday at 01:28 PM by Doughguy

[LetMePickThat]

1 hour ago, Doughguy said:

Cheers.

But it kinda appears the flightmodels sorta could be the root? Or PvE respectively.

 

I suspect that the AI isn't tightly controlled during replays, and has some leeway it shouldn't have. This causes errors compares to the original flight, errors that will propagate with time, leading to vastly different outcomes.

[Fear-Factor]

I am asking myself how the dev use this "debug review system"?
For example, the interesting timestamp is around 10 min.... the dev saw a problem and want to review that moment..... do he watch all the 10 mins again and again?
I mean.... a simple rewind function would be after all the years something great... isnt it? Not only for devs. We have also longer track replays and no rewing function is just pai.... uncool.
 

[buceador]

I have responded to this post earlier, but just to add that I experienced the same overwhelming frustration when I first came to DCS, I would pull off a fantastic dogfight win only to discover, to my dismay, that the track 'replay' differed wildly from reality.
Yes, one of the myriad of track breaking variables is the module in use.
ED has said many, many times that the track replay system was never intended as a 'proper' replay system as in MSFS / X-Plane (et al) and until they have the development time to address the anomalies we simply have to live with the altered reality that is track replay.

[rob10]

22 hours ago, Fear-Factor said:

I am asking myself how the dev use this "debug review system"?
For example, the interesting timestamp is around 10 min.... the dev saw a problem and want to review that moment..... do he watch all the 10 mins again and again?
I mean.... a simple rewind function would be after all the years something great... isnt it? Not only for devs. We have also longer track replays and no rewing function is just pai.... uncool.

The devs are using a completely different software to analyze this (my guess is looking at the behind the scenes stuff going on rather than primarily "watching" it, which is why they want a track rather than a video), so I'm sure they can do very different things to what users can.