Replay is not the battle actually fought?

[sarjenius]

Ok this sounds strange but i noticed it now the second time in a row. I was testing some cavalry charge setups and was surprised that they work often when the setting is very hard and the amount of units is not too high.

In one battle i did about 6 or 8 charge attackes with the cavalry against spearman and other cavalry effective units which reformed while my cavalry did the same in some distance - so it was rise and repeat until all infantry units were routing. The battle took about 6 minutes.

Then i loaded the replay of this particular battle. I was seeing the first minute exactly as i just played it. Then all of a sudden the narrator told me the enemy is fleeing the field and we have won. I just saw ONE charge with my cavalry, then the enemy routed and fled the battlefield while my men reformed. I was puzzled and loaded the replay again. To no avail. My 6 to 8 charge battle which lasted about 6 minutes was reduced to one charge - one minute replay.

I tested it with another new battle and alas .. the same.

Anyone experiencing the same weirdness?

[Ignoramus]

I had this happen in Medieval: Total War. I was attacking the Golden Horde in Kiev and won the battle, but the replay showed that I lost.

[sarjenius]

Even the battle statistics are "forged". In the actual battle the units have gained like three experience chevrons. In the replay statistics only two units got one.

[Ringeck]

RTW classic. Sometimes, the only thing similar about the replay and the battle was the armies and the terrain. The game got creative with the rest!

[Ares]

Ok this sounds strange but i noticed it now the second time in a row. I was testing some cavalry charge setups and was surprised that they work often when the setting is very hard and the amount of units is not too high.

In one battle i did about 6 or 8 charge attackes with the cavalry against spearman and other cavalry effective units which reformed while my cavalry did the same in some distance - so it was rise and repeat until all infantry units were routing. The battle took about 6 minutes.

Then i loaded the replay of this particular battle. I was seeing the first minute exactly as i just played it. Then all of a sudden the narrator told me the enemy is fleeing the field and we have won. I just saw ONE charge with my cavalry, then the enemy routed and fled the battlefield while my men reformed. I was puzzled and loaded the replay again. To no avail. My 6 to 8 charge battle which lasted about 6 minutes was reduced to one charge - one minute replay.

I tested it with another new battle and alas .. the same.

Anyone experiencing the same weirdness?

Indeed the same problem existed in RTW prior to the 1.2 patch. Hopefully this issue will be fixed in a patch soon.

[Myrddraal]

The reason for this is that replays are basically a stored set of commands. The theory goes that if the AI issues the same commands, and you issue the same commands, the same battle should happen.

But there's a random element, and if you play a battle where one of those random elements got lucky (like a unit didn't rout despite the odds against them) then in the replay they probably will. This problem has been in the game since STW.

It seems to me they could increase the accuracy of replays quite drastically by adding not much data to the replay files (like unit rout times), or occasional unit strength checks (i.e. if a unit has lost too many men, make them invulnerable for 6 seconds then check again)

[Comrade Alexeo]

The reason for this is that replays are basically a stored set of commands. The theory goes that if the AI issues the same commands, and you issue the same commands, the same battle should happen.

But there's a random element, and if you play a battle where one of those random elements got lucky (like a unit didn't rout despite the odds against them) then in the replay they probably will. This problem has been in the game since STW.

It seems to me they could increase the accuracy of replays quite drastically by adding not much data to the replay files (like unit rout times), or occasional unit strength checks (i.e. if a unit has lost too many men, make them invulnerable for 6 seconds then check again)

One wonders if it would be better if CA or some ingenious modder recorded replays as movie files of some sort...

...although this might make things too large - how large is the average replay file as it is now anyway?

[Faenaris]

A couple of hundred kilobytes, I think. Recording a movie would make the file a lot larger unless you use one heck of a compressor.

But I agree with you, Comrade Alexeo. Movies would be handy and if you have diskspace to burn, I say: Give us movies!

[Singleton Mosby]

Even the battle statistics are "forged". In the actual battle the units have gained like three experience chevrons. In the replay statistics only two units got one.

I have seen the same....battles in which no troops on my side where lost (horsearchers) but at the end it showed I lost 27 of them

[alpaca]

The reason for this is that replays are basically a stored set of commands. The theory goes that if the AI issues the same commands, and you issue the same commands, the same battle should happen.

But there's a random element, and if you play a battle where one of those random elements got lucky (like a unit didn't rout despite the odds against them) then in the replay they probably will. This problem has been in the game since STW.

It seems to me they could increase the accuracy of replays quite drastically by adding not much data to the replay files (like unit rout times), or occasional unit strength checks (i.e. if a unit has lost too many men, make them invulnerable for 6 seconds then check again)

Actually I think they could simply record the random seed(s) used, then theoretically, if you issue the same commands again, the exact same battle should unfold again before your eyes.

[Kobal2fr]

A movie wouldn't cut it.

Nevermind the space issue (and that's asking a lot. Raw movie data is *huge*. 20 mins of unencoded video would probably take up something like 5-8 Gigs. Yes, that much). Think about it : why do you watch replays in the first place ? I do because I want to see the battle from, say, the ennemy point of view. Or from really up close to bask in pure high poly goodness. Or to watch more closely a part of the battle I didn't pay enough attention when fighting the actual thing (what the *other* flank was doing, where those horse archers had been running away to etc...). Can't do that with just a movie of what you already saw and did.

No, storing minimal data and letting the game engine sort it out for itself is the only way. The idea of saving the random seed is interesting, but I wonder how feasible it is - after all, there is a *lot* of die rolling involved in even a single unit to unit melee...

[Myrddraal]

I do think that recording a little, if not much, extra data would go a long way. Just storing the rout times of each unit and making them fight until that point would be on heck of a step forward.

[sarjenius]

Games like Company of Heroes are able to save a replay file with the exact happenings in the battle that was fought. The largest files are about 2 MB .. and that is in an online game with a duration of 15 minutes.

[alpaca]

A movie wouldn't cut it.

Nevermind the space issue (and that's asking a lot. Raw movie data is *huge*. 20 mins of unencoded video would probably take up something like 5-8 Gigs. Yes, that much). Think about it : why do you watch replays in the first place ? I do because I want to see the battle from, say, the ennemy point of view. Or from really up close to bask in pure high poly goodness. Or to watch more closely a part of the battle I didn't pay enough attention when fighting the actual thing (what the *other* flank was doing, where those horse archers had been running away to etc...). Can't do that with just a movie of what you already saw and did.

No, storing minimal data and letting the game engine sort it out for itself is the only way. The idea of saving the random seed is interesting, but I wonder how feasible it is - after all, there is a *lot* of die rolling involved in even a single unit to unit melee...

Random seed is not the same as random number. A random seed is usually generated not too often. For some games even only once, on initialization. From this random seed, the actual numbers that are created, are calculated in a deterministic way. So even if we talk about a new random seed every second or so (which is a lot really), that wouldn't make up for a huge amount of numbers (well 3600 numbers per hour is pretty much nothing, that's about 12kB)
I'm actually pretty sure that is how CA intended it, but I guess they forgot some new seed generation somewhere ;)

[Kobal2fr]

Yes, I know what a seed is, but I guess I wouldn't be good coder material - I keep thinking in human terms, not on code ones, so I imagined every unit to have its own seed and all of its own "random" numbers determined before battle individually, when of course, now that you mention it I realize its more efficient to have just the one seed feeding numbers on every single die roll but changing it ever so often so that a battle is not determined by a seed giving only good numbers to this side and botches to that one...

Being no coder though, while I can grasp the concept of the seed, I have no idea what it actually *is*, code wise or data wise. Is it a single number ? A mathematical expression ? A set of variables ?... Time to hit the wikipedia.

[Cowhead418]

Yes, in Medieval Total War I defended a castle and won a close battle. My general's unit was a major key in the victory. However, in the replay, my general was killed early on (in the actual battle he survived) and I lost a close battle. I thought it was actually pretty cool.