Posts for Chef_Stef

Posted: 10/29/2015 5:41 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

I've found that most people will get antsy if you make them sit through anything longer than about 15 or 20 minutes - it's too easy to get saturated. For example, unless you're a huge fan of Yoshi's Island I don't think sitting through a two-hour movie is going to go well, no matter how impressive it is as a TAS. Runs like that are great for fans of a series, but are they really the best for the general newcomer audience? I agree with moozooh's criteria and would actually go further and say we should aim for no more than 20 minutes max. For what it's worth, here are actual TASes I have shown people when I want to introduce them to the concept: [2016] N64 Super Mario 64 (USA) "0 stars" in 5:02.25 by snark, Kyman, sonicpacker, Mickey/VIS & ToT Fast, slick, packed with glitches, and tons of unexpected moments. One of the most accessible TASes since everyone's played the game, the tricks are broken but understandable (you can see what's going on in a backwards long jump, for example), and the gameplay itself isn't broken. [1438] SNES International Superstar Soccer Deluxe (USA) in 15:24.38 by Marcokarty The run is super funny and shows a complete mastery of the game to the point of toying with it. 15 minutes is the perfect length for this kind of concept. [726] NES Mega Man (JPN/USA) in 15:38.07 by Bisqwit & Finalfighter Yes, this is one of the older versions. For me this has the right balance of glitches and gameplay, compared to later versions which skip a bit too much. It's got good pacing both in regular gameplay (weapon progression) and glitch use (fewer glitches early on, leading to crazy glitch stuff in the Wily levels), and is one of the best examples of something that looks perfect. [711] NES Gradius (JPN) in 10:52.35 by adelikat This one plays out really nicely and is another example of showing off and toying with the game. Nobody I know has played this game but it's super easy to understand this TAS and is great entertainment.

Posted: 10/26/2015 5:02 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

I think a second quest run is worth doing. I'm not sure how well it would be received though since the levels are a lot more constrained. Are second-quest any% runs vaultable or do they have to be entertaining enough to make Moons? My other idea was some kind of 100% / high score run. I did one a long time ago that was way too long and boring (1:50 runtime). A properly optimized run would be around 1:20 but I suspect that's still too long for this kind of game. There's also the issue of whether to get max scores per level or not - doing so makes the most sense but will increase the movie time while not strictly being required for 100%. "100% + max score" doesn't sound like a vaultable goal either. I guess I could just put a run together and see what people think.

Posted: 10/25/2015 3:06 PM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

Really_Tall wrote:

There were a couple of places it seemed like keyclimbing could have been used: in 3-1 as soon as you reach the key, and in 5-5 to get over the wall at the end. Would it be faster, or even possible at all?

For keyclimbing to be possible you need to start next to the wall but facing away from it. While holding an item Mario can't turn around in midair so you have to actually stand and turn around next to the wall. In 3-1 until the platform starts rising it's moving away from the wall so it's not possible to position Mario for a keyclimb, and in 5-5 by the time you turn Mario around the platform has already fallen too far. These are good ideas, though!

True wrote:

Oh, another one of these overly-talky platformer Mario games.

If you're willing to fire up Bizhawk there's a solution for you! Maybe we should consider an encode of this too?

Submission Text wrote:

Shut up, Mario! Mario has some rather loud voice acting in this game. While watching this you may find yourself wondering how many more YIPPPEEEEEEs and WAA-HAAAAAs you can stand. Never fear, for there's a solution! Simply enable this Lua script and Mario's voice will be muted for the duration of the run! Turning Mario's sounds off will desync the actual submission so use this modified movie for syncing with the script turned on.

Posted: 10/16/2015 12:08 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

I did something like this a long time ago for DROD (2005 remake edition) and also Beret, a friend's indie game. I used Java.awt.robot for the input since that was the easiest way to hack something together. I got through the first level in each game before quitting. I'll admit my approach may not have been the best but I remember running into lots of issues. It was indeed a huge challenge to get playback to consistently sync (DROD was better here because it buffers input) and creating the movie got tedious really fast - there's no visual feedback on what you're doing (you have to guess what keystrokes to put in) and you waste a lot of time replaying the movie checking and fixing sync issues. The end product was satisfying but it sure was a headache putting it together. If nothing else I'd want to avoid subjecting anyone else to the same. Things might be better with a more standardized toolchain but it's hard to completely solve all of these issues without just going the emulator route.

Posted: 10/10/2015 3:51 PM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

Done with world 6! About 1450 frames saved over my cancelled run, 360 frames saved over EightBitGuy's other run, and 310 frames saved over the hypothetical best-of-each combination of the two. As expected there were a bunch of mini-optimizations I found that added up - just about every level had time saved. For brevity's sake I'll only list improvements where the strategy actually changed. -- 6-2b, I optimized the approach to the trash can, which was just enough to beat the upper brick cycle, ultimately saving 29 frames. -- 6-5b, optimizing the last ladder ascent saved enough time to activate the pusher much earlier, saving 18 frames. -- 6-MM, about 200 frames saved from a new complicated strategy to get the first mini into the red lasers much sooner, saving a laser cycle. I also manipulated the mini at the end to bounce left off the spring, saving time when approaching the toy box. The factors involved to get the minis where I want them is quite involved so I'll save that discussion for the submission text. Once I finish off the final battle I'm planning to do a quick pass through the run again and hex in any useful tricks discovered in later levels (such as springboard-clipping from 5-6b). Once that's done the run should be ready for the workbench.

Posted: 9/30/2015 3:56 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

World 5 done. 16 frames saved over EightBitGuy in this world. I didn't find any major strategy changes so the improvement is just a handful of frames. Non-trivial improvements: -- 5-5b, I optimized the part where the monkey carries Mario over the spikes, saving 2 frames. -- 5-6b, I discovered that with the right subpixel positioning Mario bounces off the last spring much faster, saving 3 frames. I'll be looking for other places this can be used, but it seems pretty limited. Somersaulting at the end saved another 3 frames over backflipping. -- 5-MM, optimizing the camera movement around the minis near the end saved 4 frames. Interesting non-improvements: -- 5-4a, vine-boosting saves 5 frames but the time savings get offset by the platform cycle at the end. -- 5-3b, I found a complicated strategy that makes the upper bombs blow up sooner, ultimately saving 6 frames. Unfortunately with this strategy you're also forced to collect all 3 presents, which loses a ton of time from the minigame after the level, so it's only useful for a 100% run. EightBitGuy's WIP ends after world 5 so from here I'll be comparing against his previous movie and my cancelled submission.

Posted: 9/19/2015 6:45 PM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

World 4 done. About 980 frames saved over EightBitGuy in this world. Non-trivial improvements: -- 4-2a, duck-turning with the key saves 2 frames. This let me beat an apparent frame rule with the conveyors, saving 8 more frames during the backflip. -- 4-2b, I found a faster way to ascend the donut platforms, saving 37 frames. -- 4-6a, a new trick allows pressing the yellow switch without having to go all the way around, saving 920 frames. -- 4-6b, somersaulting at the end saves 11 frames. The new trick involves starting a handstand and alternating left and right every frame at the apex of the jump. For whatever reason this levitates Mario in midair and makes him slowly rise at 9 speed. I discovered this while messing around in 3-4a trying to glitch Mario through the red blocks above the pounder. (Turns out it's possible, but it takes way too much time to execute.) 9 speed is extremely slow - about two pixels per second. For comparison, running is 272 speed and jumping starts out at 627 speed. On top of that, the handstand jump is quite short (about one block height, compared to the backflip's three-ish blocks) so it takes forever to gain any meaningful height with the trick. If there's a platform that backflipping / somersaulting can't reach, it'll almost certainly be faster to do the level normally than to spend 30-40 seconds getting up there with the trick. So while you can use it to ascend as high as you like, I don't think it ever saves time to do so. The more-useful aspect of this trick is that you get to hover in midair indefinitely. In 4-6a I use the trick to press against the underside of a donut block, which you can't do normally, in order to activate it from the wrong side. Unfortunately there aren't many situations like this in the game so I wouldn't be surprised if this is the only place the trick is useful.

Posted: 9/13/2015 3:01 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

World 3 done. 775 non-lag frames saved over EightBitGuy, the vast majority of which came from a new boss strategy. Non-trivial improvements: -- 3-2b, somersaulting to the moving platform allows backflipping without having to turn around, saving 11 frames. -- 3-4a, 2 frames saved from duck-turning with the key instead of just turning around normally. -- 3-DK, here I discovered a trick to hit DK without consuming the barrel. This lets me wipe out DK with just the first barrel, which is great because normally this is a very slow battle. 758 frames saved with this strategy. The new trick also works for 1-DK, so I redid the battle to include it (195 frames saved). I suspect it's usable for 5-DK too but I haven't checked yet.

Posted: 9/6/2015 12:08 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

World 2 done. Compared to EightBitGuy's WIP (again, discounting lag) I've saved about 30 frames in this world. Main improvements: - 2-1a, ceiling-boosting in two places saved a total of 4 frames. - 2-3b, it turns out it's possible to somersault directly across the platforms, which is 11 frames faster than backflipping. - 2-5a, 6 frames saved from manipulating Mario's position better when dropping off the vine. - 2-5b, a new strategy here allows vine-boosting an extra time, saving 4 frames.

Posted: 8/25/2015 2:10 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

I never said I wouldn't contact Prenz or anyone in the community. All I said was that the score records would be hard to apply to an any% TAS. I didn't say they were useless or anything, in fact I pointed out a couple of ways they could still be handy. I'd also appreciate if you would quit insulting me. I haven't done anything to offend you, all I've done is post a WIP TAS for a game I really like. If you see an issue with the WIP I'd love to hear it - that's why I posted it here in the first place.

Posted: 8/24/2015 11:51 PM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

ALAKTORN wrote:

Score records are still speedruns, though. The timer gives points. If you calculate all possible points in a level and subtract them from the score record, you could figure out the time with which the level is ended. For your TAS that time should never be faster than yours, obviously.

Yeah, but for a max score run you have to go out of the way to collect presents and kill enemies. In an any% run you're just beelining for the key, door, and mini mario. Even if you figured out the level time it's not comparable to this TAS since the routes are different. But, your suggestion would work for DK levels, where the any% and 100% goals are the same (kill DK as fast as possible), or levels where there isn't anything but the key. Again, those scores are a really useful reference for a 100% TAS. But it's hard to use them with an any% TAS.

Posted: 8/24/2015 4:18 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

The TAS aims for speed, not score, so they're not really comparable. Those would be a good reference for a 100% TAS, though.

Posted: 8/23/2015 10:21 PM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

Jerry sure is spastic, I don't think I've ever seen so much jittery movement in a TAS. Fun to watch, looks nicely optimized. Yes vote from me.

Posted: 8/23/2015 9:59 PM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

So I decided to give this game another shot. Here's a WIP completing world 1. Being secretive bit me last time so I'm planning to post WIPs once per world as I work through the game. I've switched to BizHawk for a variety of reasons (mainly its integrated tools). For whatever reason BizHawk has more out-of-level lag and less in-level lag than VBA so the overall movie times aren't really comparable. If you discount lag differences, in-level I'm about 60 frames ahead of EightBitGuy's WIP. The time-savers I found were: - 1-5b: Saved a frame with slightly better jumping on the trash can. - 1-6a: Optimized the upper area slightly better, which let me get through the bricks without waiting, saving 52 frames. - 1-MM: Did a handstand-to-somersault sequence while waiting for the minis, which let me reach the red switch faster, saving 5-7 frames (this level is so laggy it's hard to tell exactly).

Posted: 8/18/2015 2:27 AM

Post subject: Re: TASing anime

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

sack_bot wrote:

Wait, I thought that this was about making some sort of tool assited speedrun of existing anime. And no, I don't know how that would work either

How about this? Link to video

Posted: 8/15/2015 3:19 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

For Scribblenauts I think we want levels where Twitch can suggest crazy things and actually have them be accepted by the game. Then we have both "Twitch is crazy for coming up with this" and "this game is crazy for accepting this" which is perfect for a game showcase like this. Idea #1: Use open-ended levels where you're given a category of objects to summon. These are the least-restrictive so they would motivate a variety of words from Twitch chat (and many of them have a good shot of actually working!). A few level ideas: -- 2-7, "What does the kid want to be when he grows up?" - Potentially any profession will work here, in the TAS we summoned criminals and they worked. -- 4-4, stock a grocery store, in the TAS crazier stuff like "slop" and "sea slug" still worked. -- 5-2, summon items for a horror movie. -- 6-3, fill a garden with plants, in the TAS we summoned poisonous plants and a "gnome". -- 6-12, decorate a house. I remember this level being particularly loose, it accepted all of the lolcats as legitimate "decorations" for whatever reason. -- 10-1, populate mars with a bunch of people. Like 2-7, the game accepts lots of sillier professions. -- 10-3, fill a classroom with school supplies. Idea #2: Pick levels needing a bit more focus. Twitch chat would have to hone in on stuff that actually solves the level rather than just throwing out random words, but there's still some wiggle room. Some interesting levels that fit this category: -- 5-4, do stuff for 3 customers at a hair salon -- 6-2, give a sequence of 3 gifts to family members -- 6-10, satisfy 4 monsters in sequence -- 7-6, appease an alien, create a water source, create renewable food sources, create an air source. -- 8-8, "army boot camp", progress through a series of mini-challenges -- 9-4, fireman training, same concept as 8-8 -- 10-2, policeman training, same concept as 8-8 Thanks to the TAS I'm biased towards noun-based solutions so there may be other levels that work perfectly well when you take adjectives into account. I'm sure Twitch chat would abuse adjectives like crazy anyway if we put this in front of them. A side note about bots, they're going to have to be pretty sophisticated to handle the situations in this game properly. Think about placing summoned items into the field (they can't overlap anything), picking an ambiguous word (does it select the entry for fish-food or fish-animal?), emptying the field if the meter is full, etc. Manual intervention might be needed in some places.

Posted: 8/24/2014 3:10 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

I really enjoyed this run. I was surprised at how many antics you could fit in while just running forward. My favorite levels were 3-1, 7-4, and 8-3. While watching this I kept saying "Mario, stop freaking out so much!". Maybe he just gets possessed from time to time? Easy yes vote.

Posted: 8/20/2013 1:06 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

Looks like that was the problem. I had left it in the default setting of Enabled because of the scary warning about glitches if it was turned off. Disabling it solved the issues I was having - frame advance is now locked to input frames, which is perfect. Thanks for helping out - any idea why this setting affects frame advance like this?

Posted: 8/19/2013 4:49 AM

Post subject: Strange/inconsistent frame advance behavior

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

I'm running into a problem where Dolphin's frame advance is acting inconsistently. Dolphin's frame advance seems to be capable of running in two different modes: Advancing one render frame at a time (from what I've read elsewhere, this is equivalent to an input frame) and advancing "half" a render frame at a time (i.e. you have to advance twice to get to the next render frame). I haven't found any setting to control this, and it seems to switch from one mode to the other for no reason. I've been trying to do a testrun for Super Monkey Ball 2 and this is really getting in the way (since "twice-per-render-frame" mode is pointless and requires a lot of extra button-pushing). Some examples where I've run into this: 1. I recorded a movie that navigates through the game menus. Once the movie finishes, Dolphin frame-advances once per render frame. I then hex-edited 10 input frames from the movie file. After playing back the modified movie, Dolphin frame-advances twice per render frame. 2. I completed a testrun TAS of the first level with Dolphin in once-per-render-frame mode the whole time. Once the next level loads, however, Dolphin switches to twice-per-render-frame. Does anyone know why this is happening? Does this have anything to do with the game itself? If it matters, I'm using a development build from about 3 weeks ago. I also tried the latest build and the behavior is the same. I can provide the example movie files if it helps.

Posted: 7/24/2013 3:55 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

I really enjoyed this TAS. This is a game I played a lot as a kid and it was really interesting to see it cleared so quickly. This TAS has a lot of fun stuff - crazy resource management, fast action, and complete domination of enemies. For some reason I kept thinking of an ant farm as I watched all the peons running around... Some of my favorite moments: * All the silly voices, and how they often overlap each other when you're giving rapid orders ("Ahoy! Ahoy! Ahoy!" "Very-very-very well." "Yes-yes-yes-yes master."). Also... "Zug zug"! * Any time you trimmed your army down for the purpose of resource management ("you guys have served me well, now it's time to die") * Other silly things (like slaying critters during idle times) * Taking advantage of the enemy AI to get friendly fire hits or just make their attacks miss constantly Some levels were a bit long but there was always something going on. I was certainly entertained the whole time. A big yes vote from me!

Posted: 7/18/2013 1:29 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

Okay here's a dropbox link: https://www.dropbox.com/s/3rs51mc8l8fgj98/scribblenauts_word_groups.zip (thanks for the tip, I almost never host files online). Regarding the script - I thought of that idea as well, but the problem is that you also get regular letters in the garbage data as well. Filtering those out means you might see stuff like "ACAR" in the list, whereas leaving the garbage characters in means that those words automatically get flagged as fishy ("@3 *ACAR"). This was meant to be a quick hack to get the data out of the ROM anyway, so give me a break! :D Edit: The link is now posted in the submission text as well.

Posted: 7/18/2013 12:11 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

The T wrote:

Is your version of this list going to be available anywhere? :)

I uploaded it here: http://www.sendspace.com/file/sythlb The numbers at the beginning of each line correspond to the object identifier (this is what we used to determine which words mapped to which objects). Everything in one group maps to the same thing. If the number ends in "02" it's an ambiguous word (such as "Chicken", which could be the food or the bird) so those words are put into their own groups. For whatever reason, certain words are stored at slightly different offsets in the ROM's data table than others, which means that occasionally you'll see some garbage characters with a word (or you'll see a word truncated slightly). It should still be easy enough to tell what the actual word is. One neat thing is that similar types of objects often have adjacent identifiers, which means they show up in this list right next to each other. For example, if you search the list for "Hockey stick" you'll find things like "Golf club", "Bowling pin", and "Baseball bat" right there as well. This was very useful for us while planning the run. I should also note that some words that appeared in the ROM weren't actually summon-able in-game (such as "Face flannel', one of our candidates for 9-2). Luckily this seems to be pretty rare...

Posted: 8/19/2011 8:03 PM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

I've really enjoyed watching your WIPs in the thread the whole time, and the newest bit (the last dungeon, etc.) was great, too. It's good to see yet another Zelda game represented here at TASVideos! Yes vote, obviously.

Posted: 8/16/2011 3:02 AM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

Any updates at this point? You're really close to finishing off the game, and I've really been anticipating this improvement for a long time.

Posted: 4/29/2011 3:07 PM

Quote

Chef_Stef

Editor, Experienced Forum User, Published Author, Active player (467)

Joined: 5/23/2006
Posts: 361
Location: Washington, United States

I'm voting 'yes' because the piano section was really entertaining, and the rest of the run (minus some of the memory sprint sections) was still a great watch. I think the style of the drawing sections is different enough from the currently-published Brain Age TAS, mainly because there are more varying game modes to play around in. For example, the abstract scribbles in the word scramble game were nicely contrasted with the more-complete drawings in other stages. I do think that there's more potential here, and I would hope that it's possible to create an even more entertaining TAS for this game. I liked one poster's suggestion that you start from a predefined save with all games unlocked, since you could pick out all of the best ones that way.