Half Life TAS

(user is banned)

2009-06-24 01:35:22 am

Don't think! feeeeeal

Ya, I think it'd be interesting. How do frames actually work in HL though? I mean you can slow the game way down with host_framerate, here it seems like there's an astronomical number of frames in just one jump. I'm guessing this actually changes the number frames per unit real time. That would actually explain the physics change people swear by. So instead of things being updated at only 60 fps, they're updated at like 600 fps, thus things like speed can grow at increasing rates per second, like interest compounded quarterly vs infinitely. Over a long period of the time end result will be substantially higher.

What is the standard "frames" (like game updates) per second of HL anyway? I'm guessing this is somewhat dependent on your "fps" video graphically speaking, which would explain why different FPSes allow you to fire weapons much faster and jump higher, ect ect.

I'm sure a lot of frame precise bugs could be found, stuff like the edge glitch would be very practical.

Since it's a TAS you might even want to do some stuff like save abusing to get more ammo; if you kill someone and grab his ammo right away, then save, load the save he will being dieing again and drop his ammo again. I'm not sure if it can be used more than once but probably can be.

Ya, I think it'd be interesting.  How do frames actually work in HL though?  I mean you can slow the game way down with host_framerate, here it seems like there's an astronomical number of frames in just one jump.  I'm guessing this actually changes the number frames per unit real time.  That would actually explain the physics change people swear by.  So instead of things being updated at only 60 fps, they're updated at like 600 fps, thus things like speed can grow at increasing rates per second, like interest compounded quarterly vs infinitely.  Over a long period of the time end result will be substantially higher.

What is the standard "frames" (like game updates) per second of HL anyway?  I'm guessing this is somewhat dependent on your "fps" video graphically speaking, which would explain why different FPSes allow you to fire weapons much faster and jump higher, ect ect.

I'm sure a lot of frame precise bugs could be found, stuff like the edge glitch would be very practical.

Since it's a TAS you might even want to do some stuff like save abusing to get more ammo; if you kill someone and grab his ammo right away, then save, load the save he will being dieing again and drop his ammo again.  I'm not sure if it can be used more than once but probably can be.

Scepheo

2009-06-24 01:56:34 am

As you're gonna headshot every frickin' enemy out there, I doubt you'll need much ammo. Most games currently run at 60 FPS, because it's rather smooth to the human eye (contrary to lower FPS, which "flickers" unless somewhat advanced/heavy motion blur is applied to each frame) and higher FPS costs both more CPU and the difference is almost unnoticeable.

However, the change in framerate could indeed have an effect on the physics (which is caused either by bad programming, or compensation for lag. Or both) so you'd have to find a way to avoid that happening.

Spider-Waffle

(user is banned)

2009-06-24 02:30:32 am

Don't think! feeeeeal

so 60 fps is standard for console and PC as well?

And for ammo it'd be for things like m4 grenades mostly, well can't really think of anything else actually. With the vast amounts of health door glitch, and tons of free m4 nades, you could really save a lot of time.

Rake

2009-06-24 08:32:57 pm

60fps is standard because that aligns with the refresh rate of most LCD monitors (there are a few that go higher)

error1

2009-06-24 09:07:00 pm

Not a Zelda 2 refrence

well a crt monitor runs at 85 hertz, lcds can do more then that. Many games have vsync as optional, meaning you can run it at what ever frame rate you want and just not display all of them. I know that quake clients have a max fps of 72 ( Dementjev got his runs pulled from the site recently for playing at 85fps ).

OlikA

2009-06-25 09:28:38 am

http://www.fortress-forever.com/fpsreport/

Same applies (nearly) to every HL mod [and HL of course].

fuzzerd

2009-06-25 11:52:56 am

most people I know of tend to use 100 fps as the standard for half life mods. this is mainly because its the max you can get with out developer 1 and the physics start to change drasticly after that. (also I notice that the firing rate seems to even out around 100 fps) also the developer limit does a much better job of limiting your fps then fps_max (just be sure to use fps_max 101 or the fps_max will limit you and give you an unsteady framerate. you'd have to have your fps ALWAYS at 100 in order to even think about tool assisting this game I think.

Spider-Waffle

(user is banned)

2009-06-27 03:51:53 pm

Don't think! feeeeeal

Well if you could set up frame advance, wouldn't fps become, well, irrelevant?

Kabuto_

2009-06-27 03:59:28 pm

If you advance by frame you need do define how long a frame should be in in-game time. For consoles this is always 1/50 for PAL and 1/60 for NTSC but for PC there's no such thing. You can use monitor refresh rates (60 Hz being common, but 75 Hz, 85 Hz, 100 Hz, 120 Hz and many others are also used) or by the game's internal refresh rates (mostly variable, unpredictable, in real life heavily depending on the player's hardware). If the game developers themselves defined a maximum frame rate (e.g. 100 Hz for Halflife) you can use that. Theoretically a speedrunner could choose a variable frame rate in a TAS by not only giving input but also how long this frame should take in in-game. A drawback is that this would be hard to translate into a video, especially on frames that take exceptionally long (e.g. 0.2 seconds), unless recording a demo from player input and then playing back that demo would smooth that out.

quadrazid

2009-06-27 04:10:26 pm

Svart Lyser Tronen

I dont think thats a problem.
If you record a demo with 5fps, you can watch the demo with 100fps if you want to.

Quote from Spider-Waffle:

Well if you could set up frame advance, wouldn't fps become, well, irrelevant?

Im not sure but
If you play with 10fps frame by frame, the gap between the frames will become much bigger then if you play with 100fps?

Spider-Waffle

(user is banned)

2009-06-28 03:17:48 am

Don't think! feeeeeal

This seems like it has potential. I can tell you right now, there's A LOT of strats that only a TAS could do that would change overall strategies and planning a lot. Like for example, on the cliff side level, you could m4 nade jump all the way into the exit pipe right out of the pipe that takes you into that level. I'm pretty sure this is impossible by a human, but with TAS like air accell I think you could do it.

Scepheo

2009-06-29 12:02:03 pm

The FPS should still be determined, as, apparently, the in-game time 1 frame takes has an effect on the physics. So you need to find a way to get a frame to take a fixed amount of in-game time, regardless of the real time it takes (as that may well be half an hour, when the TASer decides to take a break).

moozooh

2009-06-29 01:55:56 pm

Hail Discordia!

How does Doom define its "frames"? (I think they're called ticks in its case.) There is a possibility to interpolate playing speed (and possibly the amount of frames rendered per second) endlessly, but I believe the demos sync fine at any speed.

Apjjm

2009-06-29 05:03:13 pm

What?

This might help if your looking for a way to make the game get the same physics results each run.

http://gafferongames.com/game-physics/fix-your-timestep/

You may need to force/make the timestep passed into the function the same - figuring out what it is? well, i'd hazard a guess as to the maximum update rate the game will run the physics simulation at (perhaps?). Then all that would happen is the update would just be run less frequently, as opposed to compensating for a change in frequency. How you could go about this i don't really know.

Hope that helps, and sorry in advance if it doesn't.
Not a half-life runner myself, just thought i'd share what i read through the other day.

DJS@lazy2login

2009-07-02 11:33:59 am

So what you guys are talking about is that you are gonna do a script/cfg file that you are gonna exec the cfg file that will be like:

up
+lookup
+attack
-attack

and so on?

That would be cool.

Scepheo

2009-07-02 12:38:20 pm

It'll probably be more like recording a demo very, very, very slowly. And I don't know about the others, but I'm not gonna do it as I don't know anything about HL's coding.

Kabuto

2009-07-02 12:45:35 pm

I'll try to learn a little about HL coding but don't expect results soon Sad

Spider-Waffle

(user is banned)

2009-07-04 01:49:23 am

Don't think! feeeeeal

I was watching through some of my HL run and I'm thinking up notable improvements a TAS could make like every 30 seconds. grabbing the hazard suit for example, You can grab it through the barrier without even opening it so you don't have to wait for it to open. I didn't do this because without the extra room to jump I couldn't jump back over the railing and it was like .2s slower. I think a TAS could make this jump over the railing without the suit chamber opened, saving over 1s.

Then there's ways you get out of the test chamber without using the alien or taking damage. On and on.

Eldar

2009-07-04 01:54:43 am

That'd be interesting to see a whole run with all those improvements.

OlikA

2009-07-07 07:42:17 am

I did some experiment, and i couldn't find better than this:
pausable 1
bind mouse3 "pause; wait; pause"
bind p "pause"

Press spoiler cause i don't know that anyone should find this useful and i don't want to waste precious free space.

-----------------------------------------------------
So my method is that when i press the pause button [mouse3], i press the other buttons i want to in that very moment. (literally, pressing the buttons in the same time)

Here's a table that shows, how many times you need to press mouse3 to perform a +jump
fps_max 10 : 7 times
fps_max 20 : 14 times
fps_max 30 : 21 times
fps_max 60 : 42 times
fps_max 100 : 66 times (and not 70)

I recorded fps_max 20 and fps_max 100, but in demo playback you can't skip frames due to the pauses (you can play it, but can't do frame-by-frame). And as far as i "feeled" it, there were no real time difference between the two jumps. I have no guarantee for this.

Other things...
...when you are in pause state, you can't "free look". you should use scr_ofsx scr_ofsz scr_ofsy binds.
...changing the fps_max for example to 20, and you playback the demo on fps 100, it will be look crappy.
-----------------------------------------------------

And why I use this [pause; wait; pause] method and not host_framerate and others:

cause host_framerate pauses only the client. so, if you press the door, and you set host_framerate 0.00001, you will float in the air, and the door will open during 5 secs. And what about the 10 secs runtime, 9 mb dem files...
I also tried to get constant fps : no chance, there will be always +-0.01 fps change, no matter how hard i tried it.

I'm wondering what are you guys trying, and what are you doing, and what are your results...

I did some experiment, and i couldn't find better than this:
pausable 1
bind mouse3 "pause; wait; pause"
bind p "pause"

Press spoiler cause i don't know that anyone should find this useful and i don't want to waste precious free space.
[spoiler]
-----------------------------------------------------
So my method is that when i press the pause button [mouse3], i press the other buttons i want to in that very moment. (literally, pressing the buttons in the same time)

Here's a table that shows, how many times you need to press mouse3 to perform a +jump
fps_max 10 : 7 times
fps_max 20 : 14 times
fps_max 30 : 21 times
fps_max 60 : 42 times
fps_max 100 : 66 times (and not 70)

I recorded fps_max 20 and fps_max 100, but in demo playback you can't skip frames due to the pauses (you can play it, but can't do frame-by-frame). And as far as i "feeled" it, there were no real time difference between the two jumps. I have no guarantee for this.

Other things...
...when you are in pause state, you can't "free look". you should use scr_ofsx scr_ofsz scr_ofsy binds.
...changing the fps_max for example to 20, and you playback the demo on fps 100, it will be look crappy.
-----------------------------------------------------
[/spoiler]

And why I use this [pause; wait; pause] method and not host_framerate and others:
[spoiler]
cause host_framerate pauses only the client. so, if you press the door, and you set host_framerate 0.00001, you will float in the air, and the door will open during 5 secs. And what about the 10 secs runtime, 9 mb dem files...
I also tried to get constant fps : no chance, there will be always +-0.01 fps change, no matter how hard i tried it.
[/spoiler]

I'm wondering what are you guys trying, and what are you doing, and what are your results...

Spider-Waffle

(user is banned)

2009-07-07 11:24:04 pm

Don't think! feeeeeal

I never knew what scr_ofsx scr_ofsz scr_ofsy actually did. I would think you'd want +left, +right, +lookup, +lookdown binds for aiming/bhoping.

I was thinking you might as well make some kind of hook that can turn your view at the optimal speed for you to accelerate, and one that "vibrate" accelerates you in a straight path at optimal acceleration. Then like a hook that handles jumping at the absolute soonest possible instant of touching the ground.

OlikA

2009-07-08 03:18:58 am

Here is a list of the cvars with descriptions. http://scripting.elxdraco.net/cvarlist/

I think that you don't need a bunch of scripts, i could easily perform crazy jumps just by "advancing frames" (pause; wait; pause). However, you can use the console also, to "build up" cl_yawspeed, so there's no need to restart the script 934923 times, you can develop it "just-in-time". Maybe even save/load works too (with pre-pause)

You should experiment with
bind mouse3 "host_framerate 0"
bind space "host_framerate 0.001"
to figure out, what should be the very-very perfect, fully optimized jump/bunny.

Is there any summary or "collected in one page" thingy about all the HL bugs? cause it's 89 page long and... well... you know... it's f***ing long.

Spider-Waffle

(user is banned)

Spider-Waffle

Edit history:

Spider-Waffle: 2009-07-09 02:44:13 am

2009-07-08 09:44:46 pm

Don't think! feeeeeal

Oh so you were saying to use scr_ofsx scr_ofsz scr_ofsy too look around while paused? I guess that could be useful in rare instances.

I don't have the same host_framerate .00000001 problems you were having, everything is slowed down. Did you have server and client separated somehow? I'm playing in HL 1.06.

I'm talking about using programmed input assistance. Like something which sets your cl_yawspeed to the perfect value based on what your speed value is.

OlikA

2009-07-10 10:14:29 am

New game > training [version: 48/1.1.2.1/2.0.0.0 4554 insecure (70)]
And these are the effects of host_framerate:

---------------------------------------
host_framerate .001 : slowdown, works great
host_framerate .0001 : "player" totally frozen, everything else ultimate slow-motion
I can't move by the move keys (@ smaller framerate than .001) and gravity has no effect on me, only worldbrushes (new game, hard: train pushes me forward)
---------------------------------------

I checked the pause; wait; pause method, and according to the delta_stats, it's advances frame by frame.

oh, assistance, good idea. it won't be hard to make it, actually i already did a minimalistic application, which shows my speed [kernel32.dll memoryread] and that i'm in air [same], and tried to preprogram keyboard input [user32.dll keybd_event], but i'm new to this.

Spider-Waffle

(user is banned)

2009-07-10 03:41:44 pm

Don't think! feeeeeal

Ah awesome, I take it you have some coding background though. The only problem with pause is that you can't pre-pause during loading. Maybe a mod could be made which lets this happen. Host_framerate .000000001 can be done during loading though. So you say trains still move with the host_framerate pause?

Do you have an IM btw, my aim is SpiderWaffle?