The Elite Forum
The Big Three Plus One => GoldenEye 007 => Topic started by: Herppa on August 26, 2020, 01:34:15 pm
-
with Calle's statue untieds, I was wonder how much full lookdown actually saved. He says he got a 217 fail with only a sideboost. Is this because of perfect lines and val strafes or full lookdown? Problably both but I want to know how big of a factor it is.
We have been using the same amount of lookdown John Keleta was using in 2002. I think its time to find if there are better alternatives.
-
From what I know look down saves a 1 second a minute, but those statue unites are just a sick combo of trevor shot, perf FR, perf lines. There is a lot of stuff going on in that level too so look down is like needed!
-
From what I know look down saves a 1 second a minute
I have also heard this multiple times and have seen side-by-side comparisons on dam, but I am curious to see where this stat comes from and if it has actually been tested properly on every level.
It surely isn't the same amount on every stage (differing amounts of things to render, differing distances visible, transparent textures, animated guards vs static objects, which weapon is drawn) and I would like to know if looking straight up would save the same amount?
Also I recently saw alkamaass video on pausing being slower in lookdown, which confused me
-
A second per minute is a fairly good estimate though it's quite complicated in actual gameplay.
The Alka video has nothing to do with lookdown affecting the in game timer. And Henrik's tests essentially showed that only looking up affected the pausing animation's speed.
https://forums.the-elite.net/index.php?topic=18518.msg381786#msg381786
I have also heard this multiple times and have seen side-by-side comparisons on dam, but I am curious to see where this stat comes from and if it has actually been tested properly on every level.
It surely isn't the same amount on every stage (differing amounts of things to render, differing distances visible, transparent textures, animated guards vs static objects, which weapon is drawn) and I would like to know if looking straight up would save the same amount?
In theory, aside from the times you need to shoot or load the guards or objects, it would give you the best framerate by rendering as few polygons as possible at all times and would possibly slightly improve your in game time. This has generally been known for years and we've tried pretty damn near everything; I think most of the top level runs are making good use of optimal look angles.
Players have messed around with lookup and it's used on some levels already, but generally the floor is more useful for navigating, plus seeing a guard's legs is usually enough to know which animation they're doing and what line to take past them. If you can only see their head or no part of them you'll be stuck or backboosted constantly by troll guards you can't see. If the level requires any pausing then lookup is bad as it extends the animation vs planar/lookdown. Then there's the matter of some levels requiring shooting or keeping things loaded in spite of heavy lag. And in many cases you want to keep guards loaded around corners or on the other side of doors since they can pop into a disadvantageous spot if you unload them, or similarly warp doors they're meant to open.
Fortunately there is 20 years experience by hundreds of players to go on. It's actually shocking just how subliminally knowledgeable top players are in many cases, and how much there is to learn from existing runs.
One example I'd put forward would be the spiral on Caverns; it's obvious that doing it in left strafe (looking at small portions of wall) renders far fewer polygons compared to how everyone in the history of the Elite has done it (looking at the entire huge room). In theory, anyone who can master the level in left strafe should be able to cake a couple free untieds. However it probably saves next to nothing while the slightest graze against the corners of the path will ruin your run, and doing it in left means you cannot see any of the dozens of angles you have to cut precisely.
-
A second per minute is a fairly good estimate though it's quite complicated in actual gameplay.
The Alka video has nothing to do with lookdown affecting the in game timer. And Henrik's tests essentially showed that only looking up affected the pausing animation's speed.
https://forums.the-elite.net/index.php?topic=18518.msg381786#msg381786 (https://forums.the-elite.net/index.php?topic=18518.msg381786#msg381786)
Yeah that's something different but it's too much difference to be ignored. I meant to look at this a while ago when alka first showed it to me but I haven't :p . Henrik's tests will still be right generally but it looks like extreme lookdown may be bad.
If the saving of doing left strafe spiral is significant I can actually see someone doing it. You can still see the corners on the outside wall of the spiral so it's maybe doable.
Back to lookdown it's interesting that Calle said that he had to look down more to save the run, I'd like to hear his thoughts on this. Generally who knows what what saves where really - we need some sort of rom with splits on it to find out.
-
Yeah just like calle's claim that his excessive lookdown on S2 agent helped save the 0:46 since it had the slowest warp so far for a 0:46. I thought my warp last night was slower but his still is multiple frames slower after comparing. So idk maybe he has a point I'd like to see the stats too tbh. Also I don't agree that we've been using the same lookdown, everyone has their own style and looks down a different amount each area. Just there's only so many angles so of course many peoples runs will still look similar
-
Back to lookdown it's interesting that Calle said that he had to look down more to save the run, I'd like to hear his thoughts on this. Generally who knows what what saves where really - we need some sort of rom with splits on it to find out.
I thought I mentioned this but I guess I left it out of my post. There's no cycle accurate N64 emulator at present where we can exhaustively optimize what look angles save the most on console using a TAS like toolset. We can basically optimize lines and inputs though warping isn't quite right on emulator, but most lag management stuff is console only. The best we can do is use tools to count polygons in a scene to get an idea where to look, use disciplined analysis of captured video to compare example runs (with inconsistency run to run and lacking an in game timer), or the wondrous splits rom on Everdrive which has many advantages.
Similar to people posting runs with input display, I think having example runs on the splits rom by top players is a must even though most players don't have access currently. Doing detailed analysis of video is a pain in the asshole and needs to be taken with a grain of salt either way since it's real time and anything chosen as a split can vary a lot on a recording.
Having it set up to split automatically means there's no way to misinterpret the action, everything is accurate to in game time and consistent between players, and the primary analysis can take place within the flow of gameplay rather than taking longer than the runs themselves. Anyone can set it up and experiment with the hopes of coming up with a useful contribution.
-
The biggest time saves in my 2:17s are a faster Trev shot and making sure Val sees me as early as possible (which is actually trickier than you'd think). That along with better strafing is what made me feel confident in getting these times. When I started getting very consistent with everything I implemented more lookdown. I can't say exactly how much it saved in my SA run, but it clearly pushed the run to a very high 2:17.
The thing about lookdown on Statue is that it's not needed everywhere. You have to get a feel for the frame rate and when it can improve by looking down. The return is usually very laggy, but you can avoid that by looking down. However, this doesn't work very well when there are guards blocking or creating lag from just being in the vicinity. Sometimes you wanna look up to make sure the lines are good, but like I said, you gotta get a feel for when it's not affect the frame rate. 35.xx boostless returns are possible with good lag management. I got one the other day.
Not going for safe plays, like a slower Trev shot cause you're on good pace, is what is needed to get the possible untieds that are left. I believe in 1:09:59 happening some day.
-
1:09 wil 100% happen. So many big strats can be used for more untieds. (Gateshot on s2 for example) Its just a matter of time at this point.
-
Similar to people posting runs with input display, I think having example runs on the splits rom by top players is a must even though most players don't have access currently. Doing detailed analysis of video is a pain in the asshole and needs to be taken with a grain of salt either way since it's real time and anything chosen as a split can vary a lot on a recording.
Having it set up to split automatically means there's no way to misinterpret the action, everything is accurate to in game time and consistent between players, and the primary analysis can take place within the flow of gameplay rather than taking longer than the runs themselves. Anyone can set it up and experiment with the hopes of coming up with a useful contribution.
Splits data from top players/runs would be really powerful. Need more runs on Ted's splits roms - plus he's also made versions for Eng/J/PAL so we can get away from the 'feel' of versions and have some actual science on which is better where.
If I can use the splits roms to help get Archives 16 (I was 1-2 frames faster to the top of the stairs on PAL, so decided to play on that) imagine what a good player could do with that info :pimp:
-
I did pretty extensive testing on this on runway, and it does vary even run to run, but I think 1 second per minute is about right. I think lookdown saves roughly 0.2 seconds vs a look up & juggle run, but the first ~8 seconds of the level are level are the same with both techniques, so the savings is about 0.2 secs over the last 14 seconds of the level in this case.
I think as Calle said, it will vary level to level (and section to section on each level).