Getting accurate replays..?

I play multiplayer quite often, and I'm a big fan of replays. I noticed that replays matches about never.

I thought at first it was the syncronisation problem: as in, what I saw in the replay was what really happened. While it did happen twice (Maybe a few times more, maybe), the syncronisation isn't the problem and the replay itself is.

Example: Yesterday, I played a 3v3 (One of the few where the lag was acceptable and everyone knew how to play). The game had been extremely tight, and 2 enemy players had left when they were nearly defeated (Like 1 shaken half-strength unit left only) after saying "I'm dead." All our team was still alive, but with barely any troops left, but we did manage to get the last remaining player to 1 Urban Cohort (One soldier). I sent my remaining 10 war dog handlers and joked about it, and the other player laughted too.. And when it came down to 3 handler vs 1 cohort, he/she said: "So much for your dog handlers!" before I charged with my 2 soldier general and few remaining Scythian Noblewomen to finish it, to which he/she said a big "Doh". We talked a little about the results, and we had the same "result pannel", based on what we were all talking about. That would lead me to think that everyone were playing the same game, synchronized.


It was an exciting match for all six players, and I was awfully happy of the contribution I made to my team. It would have been a very entertaining and helpful replay to watch a few times. The replay becomes more and more wrong as it comes... And when the other player said: "I'm dead", he/she still had five good units. My dog handlers were gone long before (They didn't stop routing in the early match), so the comments on both sides about them made absolutely no sense in the replay. There were units sitting there as if they didn't exist: Even if I had "dropped", no player would have done what I see in the replay.

Even in 1v1 (The replay looks closer a bit though, but still wrong), with the exact same result pannel and discussion in the game (Like: "No! Stop routing!", when his army wasn't routing in the replay, but was in-game). Or a 2v2 where when I caught his 20 cataphrac army in a pincer said a vocal "****".. And in the replay, there was nothing really happening at the time of his message (Not all units moved in for the pincer and it wasn't done yet). Or someone losing their general to art fire 5 seconds after replay start, which never happened in the real game. The replays aren't only negative for me either: Like one game I lost my general foolishly early on, and he survived the whole match in the replay (Though with only 2 soldiers).


Err... that was a bit too longwided :embarassed:... Anyway, the end result is that the replay is wrong, not the game itself (It does get unsyncronised sometime, but that appears mighty rare, mostly with people with different data files). Is there any known way to get accurate replays? I love watching them... but in their current state in multiplayer, they are completely worthless. I can't use them to see what the enemy did, to see how my allies faired or what I could improve in my play: They don't show what either me or my opponent were seeing.

It seems that the replay recalculate everything that happened, instead of recording the results. Like for example the first onageer shot killing the general in the replay, and only killing most of the bodyguards in the real game. The replay records the actions, but since the general is dead, the results become more and more different as the replay go by... Like, the general unit will still try to do the same actions as ordered in-game, but the number of men remaining will be different in the replay. So the further the replay is in, the less it starts to make sense on both sides. I doubt it, but... any way of saving accurate replays..?

Quote:

---

any way of saving accurate replays..?

---

Not within the game, until they patch the game to fix it. I imagine you could use a screen capture program like Fraps to make a video recording. Hope you have a large hard drive.....

To minimize file size the replay is not a video recording of what transpired in the game. Instead it is a recording of the unit types which began the game and their subsequent movement orders. (example: move unit A from point X to point Y). The rest is just the battle engine re-processing the sum of all these movement orders and re-creating the battle on your computer for you to observe.

I too have noticed the replays are highly inaccurate, and it is not related to desynchronization between machines. Play a game, save the replay, and then watch it immediately afterwards. You will see strange things happening. I've seen enemy units that I have routed, instead win their battle and rout mine. Then they just stand around (because they don't have any movement orders) unless another unit happens to cross their path. I've also seen matchups which I have lost in game, be won in the replay. I suspect, but have not confirmed, that if I watched the same replay several times that it would vary this way each time.

What this tells me is that the battle equation has so many variables involved in the calculations that it doesn't result in a discreet outcome (kind of a fuzzy logic is involved). I don't know whether this is by design (to add a fog-of-war or an element of randomness to the outcome), or by accident.

This problem existed in the first release of MTW and it was fixed in the patch.

The origin of the problem is the random numbers used to determine "hits" and "kills" (for example, did the archer's arrow kill the horsie?). Each individual battle uses a "seed#" to tell the random number generator (which is really just a list of previously generated numbers) where to start spitting out random numbers on its list.

Apparently the replays are not using the same seed#'s as the actual battle, so different soldiers will die in the replay than in the real battle. This will cause a massive ripple effect that changes the entire outcome of the battle as shown in the replay--because before long you have soldiers who died in the real battle who are kicking ass in the replay. . . .

Anyway, that is my understanding of how the problem is occurring. How and why this was allowed to happen AGAIN by CA is one of the quality-control mysteries of the universe.

Ahh, that sounds logical. it would explain a lot. Very good posts.

If what you say about the seed# is true, Dion, it should be fairly easy to fix. Just save one more number with the replay file and no need for a huge video recording.

If it has been solved before, it can be solved again (I just guess MTW and RTW was made by two different teams and there was just not enough crosslinking between them). I hope CA are aware of this now and can include it in the patch. ~:)

This sounds exactly like when I replayed a battle I had lost back in MTW 1.0, just prior to the patch. Then came the patch and I wanted to do a battlereport. The battle played out and I soon bagan to see things that shouldn't have happened, but they were still close enough but when the time came for the final showdown on a small inhabited hill I actually routed his (Shadewolf) forces... I was dropping my jaw.

So obviously the figures for the stats and chances of each man is different...