Joined: 6/10/2013
Posts: 41
Location: The Netherland
New improvements with Knuckles
Angel Island 1: still 0:42.24
Angel Island 2: 0:25.23 (Old any% 0:25.59 & Old 100% 0:25.45)
Hydrocity 1: 0:35.50 (Old any% 0:38.24)
Hydrocity 2 death: 0:07.25 (Old any% 0:08.51)
Soon you can download the gmv when i'm finished Marble Garden 1 & 2
Joined: 10/6/2011
Posts: 1690
Location: RU · ID · AM
LTRP wrote:
Nobody really cared for TASing Sonic 1 any more. Which is why he got the record by himself.
You are certainly not right. The previous Sonic 1 TAS was already optimized to unbeleiveable level. Tee-N-Tee improved it, and not only by adding new tricks — improved precision was also in order.
P.S.: this topic is about S3&K
S3&A [Amy amy%] improvement (with Evil_3D & kaan55) — currently in SPZ2
my TAS channel · If I ever come into your dream, I’ll be riding an eggship :)
Joined: 8/8/2005
Posts: 296
Location: NSW, Australia
Holy awesome! Always knew ring runs could look great in TAS but this is quite outstanding. Should have kept a count on the amount of times Knuckles clipped into a wall :P
Joined: 10/6/2011
Posts: 1690
Location: RU · ID · AM
Those Marble Garden runs are brilliant, just like all your other runs. I am really sure you have an amazing talent at this. Also these runs are a good proof that tool-assisting can be used not only for pure time attacks, and unordinary goals may be very fun to see. That being said, personally I think that you can even let yourself add some time-loosing tricks, like killing enemies just for fun, like you did in your old Ice Cap 2 run, for example. Since it’s a superplay and not a speedrun, such things should be allowed, as long as they look really impressive and fun…
P.S.: apologies for a great delay with Sega Atlas encode, it’s all because of a lot of IRL things to do… Also I had 2 TAS WIPs (one my own and one helping got4n) and, until yesterday, had a really bad network connection. This issue seems to be resolved now:
Of course, it’s not as good as I could have if I lived in a city (a good illustration is a speed test from my company’s office), but it’s certainly a lot better than 3G which I used until yesterday. 43 times better for uploading. Also I have ability to increase the speed if needed (as long as the ADSL technology allows it). I was asking the ISP about possibility to connect via FO or ethernet, but they told me they are allowed to do it only for companies… Sorry for some off topic, just for you guys to know what is happening behind the scene… :)
S3&A [Amy amy%] improvement (with Evil_3D & kaan55) — currently in SPZ2
my TAS channel · If I ever come into your dream, I’ll be riding an eggship :)
Thanks again everyone, these comments really help me stay motivated.
WST wrote:
That being said, personally I think that you can even let yourself add some time-loosing tricks, like killing enemies just for fun, like you did in your old Ice Cap 2 run, for example. Since it’s a superplay and not a speedrun, such things should be allowed, as long as they look really impressive and fun…
I agree things like that can be interesting also, but for this run I decided not to include them. I wanted to keep my goals as well defined as possible since I'm still hoping I might be able to get this published here someday. And don't worry about the atlas encodes, I mean I want to see them but there's no rush. I'm just glad you're making them at all.
This one is not new, it is just not useful for Knuckles -- you can't finish the act from the place you end up at because Knuckles doesn't trigger Sonic's boss and you are boxed in a place for which there is no escape.
Joined: 6/10/2013
Posts: 41
Location: The Netherland
marzojr wrote:
This one is not new, it is just not useful for Knuckles -- you can't finish the act from the place you end up at because Knuckles doesn't trigger Sonic's boss and you are boxed in a place for which there is no escape.
Yes, i see it now, that is just sad. I will use the old zip for Knuckles
So I know it's a holiday and I know what i'm about to link is just a crappy youtube encode, but I'm wondering if anyone would be willing and able to critique my latest tool-assisted ascent/descent of the mountain in IC1? I used a (now old) recording I did of IC1 in one of my first posts here asking for advice. In that video, it took 1:09 in-game from start to finish. In this attempt, I managed 6 seconds faster in 1:03. Here is a breakdown of each aspects of that:
0:12 - Reach the base of the mountain after climbing straight up the wall
0:49 - Collect 50th ring and begin descent
1:03 - Land at the bottom of crevasse
Link to video
In other words, it took 12 seconds to climb the wall, 37 seconds to reach the high point of the slopes, and 14 seconds to reach the very bottom. The strategy I used was basically to stop and rev up a full spin dash (~14 frames each time) whenever my ground speed fell to ~1500, which is close to the normal running speed (except for on a couple of very steep slopes where the process of stopping manually alone takes probably twice as many frames - in those cases I just rolled to a stop naturally). It seems to me there is surely an opportunity-cost there, and I guess my question is what is the best way to get the most of it? I'm really not sure if my method is the optimum way of solving this gameplay puzzle, of climbing up the slopes as quickly as possible, and I don't even know how to test this properly.
That's partially because of some on-going confusion I have with regards to the finer details of movement speeds in this game, even though I've read http://tasvideos.org/GameResources/Genesis/SonicTheHedgehog.html several times over the past months. Specifically, about rolling it says:
"...causes the player to decelerate on a level surface. Rolling allows the player to gain speed faster on downhill slopes than running does; however, rolling caps the players x-velocity at 4096 subpixels (16 pixels) per frame. Interestingly, when on ground, both deceleration and y-velocity are based on the player's speed variable, not velocity, and the player's speed variable continues to increase as normal on a downhill, even when the x velocity cap has been reached. Rolling does not cap y velocity."
So, as a couple of side questions, I'd also like to inquire: Could somebody maybe just rephrase this for me, please? What's the difference between x-velocity and ground speed? I recall reading once in some TAS/speedrun notes that, when going downhill, "in most situations" rolling is faster than running - well, when exactly is it not faster, please, and what about going uphill? And finally in case qwerty or any of the other good Knuckles players should happen to be about, I would like to make sure of my jump-climbing technique for Knuckles: is the optimum time to begin gliding, 2 frames before his y-velocity = +96?
Thanks
When you are on the ground, your speed is completely determined by your "inertia" -- the third value for speed on my HUD -- by floor angle and by whether or not you are rolling. When you are not rolling, you have that
X speed = inertia * cos(angle)
Y speed = inertia * sin(angle)
When you are rolling, X and Y speed are still determined by the above formulas, with the exception that X speed is capped at 4096 subpixels/frame. Thus, you can never move horizontally faster than 4096 subpixels/frame when you are rolling, but you can when running. Since camera position is used to start several events, rolling downhill loses efficiency compared to running because the camera can scroll up to 6144 subpixels/frame in S3&K (for S1 and for S2, camera moves at most 4096 subpixels/frame, so you lose less time).
When you are rolling, you accelerate faster when rolling downhill than you do when running downhill; you also decelerate less when rolling uphill than you do running uphill. If your X speed is less than the rolling X speed cap, you will benefit from rolling downhill; if not, you will waste speed because you could be running faster.
Joined: 10/6/2011
Posts: 1690
Location: RU · ID · AM
TheYogWog wrote:
I would like to make sure of my jump-climbing technique for Knuckles: is the optimum time to begin gliding, 2 frames before his y-velocity = +96?
You may be surprised, but actually all of the jump-clibing patterns (excluding the very shortest one) give approximately the same vertical speed, which is about 1.3 pixels per frame (compared to 1.0 pixel per frame when climbing mormally). The most significant part of the benefit is usually in the top of the wall, when jump-climbing allows you to land on the top and charge the spin dash earlier, but as I see from watching your recent runs, you took that technic yourself.
As I see from my attempts, the best speed (~1.36 pixels per frame) comes from the pattern recently used by me and Qwerty. Here is the pattern in assumption that you are climbing the wall located to the left of you.
I used that pattern in my run, but then I noticed that feeuzz slightly modified the input for some reason (I cannot tell you why — better if he tells it himself), but, as I see from quick measures, his change did not affect Knuckles climb at all — looks like he simply beautified the input by removing unnecessary pressings… Or maybe he went ahead of me and had to redo it :)
Even better jump-climbing speeds can be got from abusing a nearby opposite wall, but I am not blind and see that you already used it :)
Sadly, I cannot say anything about Ice Cap 1 beginning (running up the slope to collect the rings) — would be good if the experts at optimizing posted their comments on it. But, I beleive, the idea is like with jump-climbing — maximizing the average speed allover the slope.
S3&A [Amy amy%] improvement (with Evil_3D & kaan55) — currently in SPZ2
my TAS channel · If I ever come into your dream, I’ll be riding an eggship :)
you also decelerate less when rolling uphill than you do running uphill. If your X speed is less than the rolling X speed cap, you will benefit from rolling downhill; if not, you will waste speed because you could be running faster.
I didn't know that about uphill deceleration. The rest of that pretty much confirms what I was thinking but wasn't sure, so it's nice to know. And WST it's also a relief to know the climbing methods are not very particular as to which one is the best. As for everything else you guys wrote, I will try to keep that in mind, thank you both for clarifying. Honestly I'm not 100% clear on everything yet but definitely better than before, and I think what you've said will sink in more and make more sense to me with time/experience.
I did some more testing just now and found that doing a full spindash whenever my ground speed fell below 2,000 yielded best overall results. The ideal range is probably in the low 2,000's, anything from 2,000-2,300 would probably be about the same. Anything much lower or higher than that starts costing more frames than it's worth. Sometimes it also depends on the angle of the slope. I may be wrong of course, too, but using this method I managed to improve my time by 2 and 4 seconds on the ascent and descent respectively. That's probably as good as I'm willing to aim for, for now. I don't know how I could improve it much more.
Of course, it’s not as good as I could have if I lived in a city (a good illustration is a speed test from my company’s office), but it’s certainly a lot better than 3G which I used until yesterday. 43 times better for uploading. Also I have ability to increase the speed if needed (as long as the ADSL technology allows it). I was asking the ISP about possibility to connect via FO or ethernet, but they told me they are allowed to do it only for companies… Sorry for some off topic, just for you guys to know what is happening behind the scene… :)
Here is the gmv: http://dehacked.2y.net/microstorage.php/info/1278573735/qwerty-sonic3knuckles-hydrocity2improvement.gmv
