Here's the blended version.
Here's the point in the demo rec you use as a splicing point.
I believe this is approximately the starting point you use for the shadow play half of the splice.
I went back and looked at the times for each sequence and they seem roughly the same, honestly this issue could easily be something that happened during the encoding phase. If one video was encoded improperly and inadvertently sped up or slowed down it would make a large enough discrepancy to where if they were "synced" together it would be off. I personally lined up the actors in the scene (if you go back to my original post you see that I tried to line up the bots actions / movements / something tangible and compared the x-hairs that way) to test the position of the crosshair in one clip as opposed to the other one. Another reason why the synced video is so different is that the shadow play recording is at a higher field of view. Again, go back to my original images and it's very readily apparent. You could have synced the exact same gameplay at two different F.O.V.s and still had a problem making them appear to align properly.
First, I don't think anybody has said to wit that the replay system has unerring accuracy. Everyone knows that there's a "crouch" bug that's easy to encounter that makes the viewpoint seem lower than it is. Second, it's very important to note that the demo or replay system processes, parses and distributes information the same way as the engine does. It records the x / y position of the crosshair and the movement of actors in relation to it from the master engine located at the server. When your x / y crosshair coordinates line up with an actor and you fire that's a hit. In theory this means that even though the exact view the player experiences won't be apparent, the view that the engine takes into account does. If the demo spectator system was completely flawed and broken it would essentially mean that the game was unplayable and that there is an implicit problem in the engine itself because the replay view is what the server "sees" during a game. Nothing would ever line up and nobody could ever land any shots because of how outrageously out of whack the game's engine would be. It's funny because even in your fundamentally flawed videos they do actually sync up nearly perfectly when the sniper bullet is fired.
Ultimately though aimbots don't behave in "human" manners, they hook into the engine and use that information to instantly process the fastest angle to a target based on the cheater's preferences. If the engine logs linear player input that happens to move at the same speed as an actor that's exactly what it will show, I find that information a lot more pressing than what the client sees. As we have access to the developers we can find out from the horse's mouth if the information presented to a replay viewer is an accurate representation of what the server registers.
@Rev Here's where cheats come into play. Typically with an aimbot the crosshair moves to a spot on the enemy player's skeletal mesh and "locks" there for a brief period of time. Whatever movement the opponent player makes the aimbot is going to follow exactly because it's following the movement as it's being placed into the engine. When you see people complain that the bots in single player mode are way too good at tracking weapons this is why - people see and react to an opponent's movement but the bots are being relayed the information instantaneously from the engine can react to it just as quickly. Human beings don't work that way, you can predict to a certain extent but you have to be able to use your spatial reasoning and see where your opponent is moving to continue aiming. In order to recreate these shots you would have to snap towards a target and then upon reaching a certain spot of your choosing relative to that target, would have to slow down your mouse to match the exact same speed of the enemy in a single fluid motion. Not only that but in that first clip you would have to flick towards them aiming at their hip and when you reach their hip you would then have to perfectly decelerate your mouse to match the opponents movement speed. Upon decelerating to their exact speed so that you could continue tracking the pin point precise location you want you would then have to track your mouse in such a way as to continue following them perfectly while drawing a perfect straight line downwards to the hip again as the model is falling from the sky. First and foremost it's , much less while you're tracking a moving object in a perfect manner. Imagine you had to draw a perfect straight line downward while I was moving your mouse pad to the right fairly quickly. That's what linear aiming means - a computer program is only concerned with plotting the fastest point to a target, it doesn't know that it shouldn't do so in a perfectly straight line because it's humanly impossible. Aimbots also have certain features such as being able to specify what part of the body you want to aim at, if you want to aim at the nearest target, the target with the lowest health, the "aim angle" you want to set (I.E. whenever you get your crosshair within 5 degrees x / y of your target it's going to take over for you), and many more.
Another problem is that the replay view actually was used the clip at the end - but the rest of the clips were made via a spectator's live stream. In fact the clip that used the built replay system was originally captured by a spectator with shadow play. Ultimately it was edited to that fashion by a notable player whose video editing skills are as formidable as his flak cannon but who used the replay system as his source. In both versions of said clip however the same linear and mechanical aim was exhibited.
Here's a shadow play recorded by novaz and spectated / shadow played by myself recently. They sync up extremely well. Even if the demo/replay system was inherently flawed and utterly broken (which again will be verified via code) it appears as though the spectator system records information with a very high accuracy.
Something more important to consider is the fact that in the clips which many find questionable the player is moving in more or less a straight line, as opposed to the comparison video which shows a player making numerous and successive wall dodges and ground slides. When you're trying to scientifically recreate an event you want to have as many of the same circumstances as you originally had (i.e. the control) in order to eliminate the most potential confounding variables. I feel like someone walking in a straight line and snapping / locking onto rigid points of space on an enemy model controlled by a human for prolonged periods of time (in F.P.S. standards at least) and recorded by a spectator using shadow play is not dismissed by said player making exaggerated movements while mouse aiming at a bot that is standing completely still and then personally encoding that replay. It's something akin to finding out that a kite with a key tied to it will attract lightning, then going outside on a beautiful summer day to recreate that event.
I also find it funny that another important part of the clip - reacting to blimpo going up a lift as if he's able to shoot him through a wall - has not been addressed.
