• Emulator: Mupen64 Anti-DS
  • Aims for fastest in-game times while also trying for fastest overall time whenever possible.
  • Takes damage to save time
  • Heavy luck manipulation
  • Completes bonus levels
Graphics plugin: Run synchs with either Jabo Direct3D8 1.6 or Rice Video Plugin 6.1.1 or later version.
NOTE TO ENCODERS: Run can be encoded with Glide64 Napalm WX if using some savestates and pasting segments together. I'll provide savestates and more info on this later. Since both Rice and Jabo has severe graphical issues this is definately the approach i recommend(Napalm has no graphical issues whatsoever).
Input plugin: TAS input plugin 0.6(2 controllers enabled) Sound plugin: Jabo DirectSound 1.6 RSP plugin: RSP Emulation Plugin
This run improves on Rising tempest's previous TAS, however it is overall slower because it beats 2 additional bonus levels because most poeple wanted to see those included. For this reason, run doesn't end at or ever show the credits, i tried to make this work but it wasn't possible. Option was between doing all levels and ending at credits and i chose to do all levels. So instead movie ends at fadeout of final level.
This is the first full TAS run where i've truly aimed for absolute perfection. This run accomplishes that goal in ways i could never have imagined when i started. Almost every level is completed in a time below what was believed to be the lowest possible time before run was started. I should also note that every level is a TAS record, i.e no level has ever been TAS'ed faster than in this run(not counting decimals). The improvements on individual levels compared to previous run are the following:
Level      | This run | Rising Tempest | Improvement
-
Dam          1:47       1:49             0:02 
Facility     0:49       0:52             0:03
Runway       0:33       0:35             0:02
Surface 1    1:44       1:48             0:04
Bunker 1     0:54       0:59             0:05
Silo         1:07       1:14             0:07
Frigate      0:56       1:08             0:12
Surface 2    1:10       1:16             0:06
Bunker 2     0:40       0:48             0:08
Statue       2:15       2:17             0:02
Archives     0:51       0:54             0:03
Streets      1:50       1:54             0:04
Depot        0:42       0:48             0:06
Train        1:40       1:44             0:04
Jungle       0:49       0:51             0:02
Control      3:51       3:57             0:06
Caverns      1:19       1:22             0:03
Cradle       0:32       0:34             0:02
-
Total:                                   1:21

Bonus levels against previous TAS records:

Level      | This run | Previous Record | Improvement
-
Aztec        1:18       1:29              0:11
Egypt        0:42       0:44              0:02
-
Total                                     0:13

The making of this run:

Originally this runs started when Jack(Rising Tempest) asked me if i wanted to be a co-author in an improvement run. Sometime after i started the Dam level and i simply kept going and finished it. At this point i hadn't had alot of contact with Jack and in the end i continued the run by myself. In the beginning, Simon used to do TAS'es alongside my own which was a big factor later on in perfecting alot of levels. We were able to compare runs and TT's and this made it possible to improve levels more than i probably would've done had i done it alone. When reaching the Surface 2 level Simon made his first contribution to the movie when i asked him to do the beginning of the level for me. On the next level, Bunker 2, i ran into some problems about halfway through the run and after countless attempts of trying to manipulate a guard holding a safe key, i gave up. At this point, Simon took over the whole run. He finished Bunker 2, and also did Statue and Archives levels. After this almost a year passed without any furhter progress. Mostly because this run aimed to beat all previous TAS records and the standing record on Streets was 1:51 by Simon which we thought was unbeatable with the control style used(because leaning can't be done which makes the Valentin strafing slower). However, roughly a year later Simon told me is was "about time" to continue the run so we made a try on Streets. Originally we planned to co-author the rest of the run but by coincidence, i ended up doing all the input for Streets. I also continued on with Depot and at this point, we decided that i'd do the rest of the game by myself. The times gotten here in the second half of the run were mindblowing both to myself and Simon and hopefully the rest of the communities following the progress. I managed to finish the run in only a few weeks after returning.

My own thoughts:

This run is far better than i ever expected it to be. And while some minor improvements do exist, I consider this run to be as close to maxed out as Goldeneye can get. Some levels like Streets where luck manipulation is extremely heavy had everything going right and the time is so fast that i think it would be extremely hard to even tie. I'm very proud of how the run turned out and i hope everyone can enjoy it as much as i do myself :)

Tricks used:

On Surface 1, i shoot through the window of a hut to make the guard with the key come out.
On Bunker 1, i use 2.3 control style to shoot during cinema, this to alert a guard so he can open two doors for me. I also shoot close to Boris in beginning so he'll get his hands up earlier and therefor start walking earlier.
On Silo and Aztec, i do little quickpauses to speed up any full pause coming later.
On Frigate, i manipulate every hostage to choose the closest escape point.
On Surface 2, i shoot a camera through the trees to save some time. This is done at earliest possible moment cause the camera is not loaded until this point.
On Bunker 2, we lure the clipboard guard early, which usually makes a safe key guard take a bad route, however, if you're fast enough, he'll take a good route, this saved 5 seconds.
On Statue, Trevelyan is shot so he can start talking earlier.
On Archives, 2.3 control style is used to get an earlier slap in beginning.
On Streets, a grenade is used to destroy a roadblock allowing Bond to run through it.
On Depot, 2.3 control style is used to destroy a screen during cinema, this is part of an objective and also attracts guards who open a roller door for you.
On Train, i use a weapon switching trick to use ZMG and Laser at the same time. This doesn't save anything but adds alot to entertainment.
On Jungle i once again use 2.3 control style to be able to throw mines instantly.
On Control, i shoot Natalyas legs many times so that she'll have a shorter distance to run when escaping.
On Cradle, i use a grenade to kill both Trevelyan and destroy a console at the same time. This completes both objectives on this level.
On Aztec, i lure 2 guards to open a glass door for me, saving a ton of time. I also input the guidence data in the end of the level instead of in the beginning.
On Egypt, i warp a solid stone door without even opening it!
I'm not gonna go into further details on each level because i could write endless details about them.

Possible improvements:

While most levels are executed perfectly, two minor mistakes were made in this run(and couldn't be helped since this was discovered long after run had started):
1. Control style 2.4 had been a better choice. It was discovered that bonds forward speed is the value that can accelerate after a certain time and 2.4 allows you to do this during cinemas.
2. Using aim crosshair for some shots was believed to not decelerate Bond if only held for a single frame. Further memory watching proved that this was false and therefor some speed was lost on some levels. This might still work with 2.4.

Special Thanks:

Simon Sternis, for support/help throughout run, especially for taking over the run when i gave up.
Jack Wedge, for creating the previous run which has been an inspiration and a source for comparising.
The-Elite community, for all years of developing strats and also for an overall interrest in the run.
Tasvideos community, for helpful ideas, help with memory watching and also great support throughout the run.

Mukki: Judging...
Mukki: Fantastic work here, no question. Despite the fact that a good encode is yet to be uploaded I may as well accept this now as an improvement to the currently published run. It is a no-brainer. Hopefully this can be published with a Glide64 Napalm WX encode.
Aktan: My next project....

nfq
Player (92)
Joined: 5/10/2005
Posts: 1204
Warp wrote:
However, does this also affect the running speed, or does it only affect how many frames per second the screen is updated? (In the vast majority of 3D games the framerate does not affect the playing speed.
It probably affects the playing speed. I noticed once when I slowed down the game that I walked much slower. Wyster or someone could easily check this with the speed RAM address... just walk without lookdown first, then load a state and walk the same thing with lookdown, and see if the speed increases.
Active player (301)
Joined: 9/2/2006
Posts: 504
Speed addresses are constant and not dependent on lag i think. However it might be possible to check distance traveled/number of frames.
Joined: 7/2/2007
Posts: 3960
My understanding is that lookdown is used because the game's in-game timer counts lag frames in addition to normal frames -- it's effectively a real-time clock instead of the usual frame counter. Thus, anything you can do to reduce lag will get you a faster time. We're just not used to this applying to real-time runners as well as to TASers. :) I doubt it impacts your running speed in RAM, which is almost certainly expressed in units per frame to avoid having the physics depend on how laggy an area is. That way madness lies. Your description of framerate vs. playing speed certainly doesn't apply entirely to Goldeneye. You can easily overload it and cause slowdown (not just low display framerate) just by causing a couple of explosions. Even if the physics updates are decoupled from the display updates, the game is clearly only willing to wait so long between display updates.
Pyrel - an open-source rewrite of the Angband roguelike game in Python.
Active player (301)
Joined: 9/2/2006
Posts: 504
Sounds very logical ^^ Thx for clearing this up. So if the only reason you get faster times is because timer counts lag frames, does that mean that a run aiming for real-time can skip lookdown completely? If so than that would definately be the goal for the next run. Too bad the only other minor improvements is 2.4 and some very minor strat changes on 2 levels or something... not enough to warrant an improvement run imo.
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
Derakon wrote:
My understanding is that lookdown is used because the game's in-game timer counts lag frames in addition to normal frames -- it's effectively a real-time clock instead of the usual frame counter. Thus, anything you can do to reduce lag will get you a faster time. We're just not used to this applying to real-time runners as well as to TASers. :)
I didn't really understand that. If getting from point A to point B takes the same amount of (wall clock) time regardless of the framerate (meaning that with a lower framerate you simply "jump" longer steps between frames so that the overall speed remains the same), how would it affect any in-game timer if that timer counts real time and not frames? (If it counted frames, it would actually be advantageous to have as a low framerate as possible.) I do understand, however, that a lower framerate could make your running speed slower by some small fraction if the timing routines in the game are less than perfect.
Joined: 7/2/2007
Posts: 3960
You missed the second part of my post, where I said that Goldeneye does not have a perfectly realtime physics system. You can absolutely cause massive slowdown (where things take more real time to happen) by e.g. looking at a couple of explosions. I wouldn't be surprised if the same thing happens to a lesser extent if you make the game draw lots of things, by not looking at the ground.
Pyrel - an open-source rewrite of the Angband roguelike game in Python.
Active player (405)
Joined: 3/22/2006
Posts: 708
Sorry for the late reply. I haven't been active with this site recently. Just wondering why this run obsoletes the last one, since it's actually slightly longer (according to the tas bot).
sgrunt
He/Him
Emulator Coder
Joined: 10/28/2007
Posts: 1360
Location: The dark horror in the back of your mind
This run completes all the levels, not just up to Cradle.
Active player (301)
Joined: 9/2/2006
Posts: 504
Finally got around to uploading the run to youtube: http://www.youtube.com/watch?v=a_jF9PzGOxI
Joined: 1/22/2008
Posts: 319
Location: Brasil
where can i find this emulator Mupen64 Anti-DS ? i was using mupen64-rerecording-v8 with all the recomended plugins and is got desync in the very begginin of the first stage
Run..Run...Run.....
Editor, Experienced player (851)
Joined: 8/12/2008
Posts: 845
Location: Québec, Canada
wicked wrote:
where can i find this emulator Mupen64 Anti-DS ? i was using mupen64-rerecording-v8 with all the recomended plugins and is got desync in the very begginin of the first stage
Here.
Joined: 8/27/2009
Posts: 159
Location: California
I voted 'no', even though I knew I'd be in the minority cuz almost nothing gets rejected here. While this is all good and optimized, it's just not fun to watch because of the need to look down. I understand it's the fastest way to play, but I feel like I'd be more comfortable voting 'yes' if there was a run with the entertainment tradeoff of actually being able to see what's happening instead of looking down and to the wall. Not that this isn't an impressive run, and not that I don't appreciate the work that went into it, but it's just not what I'd want to be the only example of a Goldeneye TAS.
Active player (301)
Joined: 9/2/2006
Posts: 504
While this is all good and optimized, it's just not fun to watch because of the need to look down. I understand it's the fastest way to play, but I feel like I'd be more comfortable voting 'yes' if there was a run with the entertainment tradeoff of actually being able to see what's happening instead of looking down and to the wall. Not that this isn't an impressive run, and not that I don't appreciate the work that went into it, but it's just not what I'd want to be the only example of a Goldeneye TAS.
I'm aware of the issues with this run and i have already started a new run which aims to increase entertainment and being viewer-friendly(no lookdown) along with some improvements here and there. If you're interrested in following the progress WIP's are available in the Goldeneye topic.
MTA
Joined: 4/3/2012
Posts: 68
i vote YES this was an awesome run