1 2
5 6 7 8
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
zaphod77 wrote:
Wow. i didn't think there was any way to make a game truly hostile to tasing without any BS tricks. but seems they managed it.
True :) All SMB games are like that, sadly, even if the end result is extremely satisfying and impressive. Link to video
Collision?!
I have good news and bad news. The good news are that a new strategy was found completely by accident after grueling hours of streaming thanks to the help of CyclopsDragon; a collsion so crazy I have no idea why it has a reason to exist. This collision is an upwarp caused by hitting very slightly the edge of the platform for some reason. While this isn't a full clip through the platform, it clips through the lower end and allows Aiai to be instantanely warped up. This new stratrequires you to slow down much less than the previous strat. The bad news.. Unless a way to boost into the wall corner and somehow get enough speed, this is much slower than the previous method. The way shown in this video doens't clip off the left wall: if there's a way to clip off this wall then execute the strat, it will be faster for sure. If this strat turns out not be able to be executed.. I'll call it a day, see if any improvements are doable in the second half of the strat, then be done with this level. The WR is alderaly held by the TAS of this level (by three frames) and if improvements aren't real, they're not worth pursuing.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Link to video Big news! After about an hour of routing, testing and streaming, a new collision using the new start (that slows down a tiny bit at the start) has been found. The biggest thing is that it allows us to get past the first set of platforms without slowing down!!! This is huge: it may shave off in the double digit of frames. I'm currently looking at a way to get past set 2 of platforms without slowing down for more than one frame: but, for the first time in a really long time, there's hope for saving a TON of time. Crossing fingers for World 1 under 100k rerecords.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Link to video
A solid summary of this strat
Rerecord count: 4229 User movie #26486672734294737 What was thought an unbreakable strat, an cycle-dependant level that would never be improved, just got improved by one frame, and does it feel much more than just one. The entire strat has been iterated, redone and redone over the course of a little more than one month, and my efforts have finally come to fruction. This strat features two perfects clips so crazy I'd never thought to see them. The first is showcased just above, the second is right here. Due to redoing the entire level, it now also looks much cleaner, and I can finally sleep in peace.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Moving on in backwards order to step two: Progenitor. This level is fun to TAS, but it's optimisation hell. My main objective here is to abuse one main thing: what we refer to as the Fladdermus bounce. You can see it in action here: www.youtube.com/watch?v=oezy7jesCaQ The Fladdermus spot exists in this railing in a very specific spot, and I still don't know specifically why it happens here out of all the places, but it seems to be a misaligned spot (?). This spot, when landed on, will bounce you much, much higher than landing anywhere else on the railing. Fladdermus uses it in his video to shoot straight for the end of the level, and the current TAS doens't use that, because of a less wide angle, resulting in less speed, resulting in not enough velocity to cross the gap. But if this can be succesfully implemented, the time savings would be huge, probably more than an second in time saved. This is much faster than the strat I use, and I'm currently trying to accomodate my angle and speed to try to get it. It also looks extremely impressive.. so I'm not gonna give up on this one.[/userfile]
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Only now I realise trying to do these levels in order is a terrible idea, considering World 1 is arguably the hardest world in the entire game. The order in which I do levels will be much more liberal from now on, starting from trying to apply certain strategies I test in my free time. But before, the greatest collision of all time: https://youtu.be/3WAGGCXU93o For the record, this is one of the lowest bounces you can get in the entire game, executed by a very precise clip. It gathers a whooping 243 speed! but is unfornately completely useless.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
https://youtu.be/U6gG9X4K-1c Rerecords: 1038
Instances of collision
The most impressive thing for me may be that I have fully TASed this level in around an hour. There are a lot of notable instances of questionable collision, the obivious one being the clip THROUGH the staircase, which is why the entire strategy works. Bonus: bypassing the first death barrier ( https://youtu.be/d9n1lhq53Dc ) gives you a good view of how incredibly polished this game backgrounds are for an early DS game.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
https://youtu.be/9rASt9RtXTg Rerecords: 1582 Frame count: 295 Infinity is Entwined but harder. Thankfully, the collision is still bad. This took a shitton of time to find the collision that zipped Aiai forwards on the very specific frame, but it was worth it.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Now begins the fun levels.. Road to Nowhere has so much potential and is so complex I could write a book about it. 51.86 is the SMB1 TAS world record, and considering how broken collision is in Touch&Roll and the fact that T&R is about 30% faster, I'm aiming ultimately to beat or be around this time; that's when I'll be satisfied with this level. The current world record is 35.55 on this level. I've tested some collisions for an hour or so and here's what I've got so far, with 689 rerecords. The bounces are unoptimal, they can be manipulated a lot better, and everything else is massively unoptimised, but it should give you a good idea of what I'm aiming for. (In case you haven't noticed, this beats current WR by 6 seconds.) https://youtu.be/jd8pnPiiH7o
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Link to video Rerecords: 1275 Frame count: 584 .. Wow. This is one of the biggest improvents from IL to TAS in the entire run, and will probably stay so. Road to Nowhere is officially one of my favorite levels. I have no explaination why this strat works, it just does.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Receding Curves is a rather particular level, known wide and far for its rather particular collision. I'm currently testing things out ATM, but I'm very close to a viable strategy. Here's a testrun I made of the TAS strategy (826 rerecords): https://youtu.be/_n96icynf0Y This is currently .10 (6 frames) behind the human record, though I'm almonst sure to surpass the record with a bit of tweaking. The major improvement is finding a better frame to slow down at the start, as well as slightly better collision, which would push it beyond .50.
Former player
Joined: 4/18/2015
Posts: 168
Location: Canada
xy2_ wrote:
Link to video Rerecords: 1275 Frame count: 584 .. Wow. This is one of the biggest improvents from IL to TAS in the entire run, and will probably stay so. Road to Nowhere is officially one of my favorite levels. I have no explaination why this strat works, it just does.
That was beautiful :') Might be my favourite level you've done so far. It was so fast and extremely complicated. This level is a good example of why SMB TAS will always be interesting. The actual time difference between the WR and this is also amazing. Good job!!
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
A new strategy for Prohibition is currently being researched by blendra and Silverboxer, and in the meantime, I've tried my hand at TASing this strategy: https://youtu.be/3Sywf17OtHI This is 2 frames behind the WR, but it has alderaly been beaten (we got .93 non-assisted shortly after.) Reaching the .95 barrier is 2 frames, and there's alderaly one down.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Link to video Rerecords: 2045 Frame count: 184 The year is 2016. Stupid levels are being smashed, and all is well. The strategy used here is a completely new strategy to smash a long-standing record, which is able to grind out just one frame out of the human .95 WR. The key? A brand new bounce! After doing the classic start (boost, then bounce on the second ramp) I go left just enough, then redirect myself to slightly right (196,0) before hitting the third moving block, and grind max speed out of it as soon as it is hit; this was originally blendra's and Silverboxer's idea and they devised the strat. This TAS is just a slight variation of it. Once the bounce from the new strat is done, I go and bounce between an edge and the third red bar, bouncing me from the edge that is to the right into the bar and giving me a very low bounce straight into the goal.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Link to video Movie file While dicking around and failing to get any progress done (mostly entering into walls) I've started doing a TAS of this game on MRGP - Monkey Race Grand Prix. This completes all the courses in Grand Prix mode. I chose this mode instead of time trial because of the presence of items and other monkeys, making it a lot more interesting in my opinion. Fortunately, this mode is much easier to optimise than normal gameplay, as this time you have control over the monkey and not the field. As such, movement isn't terribly complex to optimise (and I've gotten pretty good at it at this point.) Many times you'ill see me bump instead of cutting corners, and that is because bumps waste almonst no speed. One interesting thing about this run is that we have 5 laps. This might sound it will make a boring and repetitive TAS, but each lap ends up being different because of interaction with monkeys/items. Manipulating monkeys (movement and behavior based on your position) is a difficult ordeal but gives you an extra around 15 speed for each bump, so I abuse it whenever I can. Items are a whole different beast. This game has zero RNG: this works against us here, as instead to decide item box drops the game bases itself on a global timer that runs upon startup of GP. Manipulation of this is borderline impossible, as the timer is long. This means I have to make do with what the timer gives me. However, items are pretty interesting as a lot of items are useful. The two items I used in this track are bananas and ice cubes. Starting with ice cubes, I didn't actually 'use' ice cubes, but another monkey launched an ice cube at me. This is because the ice cube is an offensive move, and it has an interesting effect: it gives you a huge spike of speed (more than double normal speed) but at the cost of being able to change direction only by bumps. This is the second fastest form of movement possible, although its boost lasts on a wider period on time than a banana boost. Bananas are the other item abused here. They have the property to massively increase your speed upon landing on one for about a second, then applying a short, almonst irrelevant slowdown. The main attraction of banana boosts instead of ice cubing is that since banana boosting retains your speed. As such, they can be combined with other means of movement! The most notable exemple in the run is at the third lap, where I use a strategically dropped banana a turn earlier to banana boost immediatly after exiting a booster, spiking my speed up to almonst 390 and allowing a much faster wall, saving more than a second.
Taechuk
He/Him
Player (63)
Joined: 12/4/2017
Posts: 66
Is the TAS still in progress? Because I just re-picked up that game and became interested in helping TASing this game.
Taechuk
He/Him
Player (63)
Joined: 12/4/2017
Posts: 66
I've been doing my TAS on my side. Did it fairly optimised, but I couldn't beat some WRs, so I'll need to do these again for more optimisation. I'm at the end of world 11, and I think they only count the bananas you've collected when you end a world. Might be wrong, but from what I know, you can't use practice mode to get all 2006 bananas. That's sad, because I estimated that you could get all 2006 bananas by repeating 10-5 (bonus stage with 50 bananas that can be done in around 8 seconds) and do that for a bit over 8 minutes. But you can't do that because practice mode doesn't work, and since it's a bonus stage you can't die and repeat the level.
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
Wow, I totally missed this post, my bad. Yes, the TAS is not in developpement, I just do some ILs whenever I feel like. Had been busy with other projects, other TASes and school most of the time. You have encountered the final world unlock problem. My plan for it was to just start with a SRAM-anchored movie with all 12 worlds unlocked and do all the worlds like the other SMB runs, because honestly this restriction is stupid - I don't want to watch more than 15 minutes of banana grinding in the middle of an otherwise fast paced run. One good thing is that the game has no RNG as far as I saw during my attempts. So this is why I only TASed individual levels, and then when doing the final run, all that would have to be done is to simply stitch them up using hex editing, and you'd have a full run. If an improvement was found, no need to restart the whole run, just edit the ILs. This is also why the movie file is available for each level, so that would I be unable to finish a TAS of this game (which is what happened) others could just copy paste and then try to improve my input if possible. Could you post your current TAS on userfiles, along with tools/scripts/etc you used? I would be interested in taking a look at this.
Taechuk
He/Him
Player (63)
Joined: 12/4/2017
Posts: 66
Yeah, just uploaded it. Feel free to look at it. http://tasvideos.org/userfiles/info/48869176322987241 I didn't use any tools/scripts. I did use another TAS (https://www.youtube.com/watch?v=TmLcN9_7jp0 it's not that great, but it's not terrible) as a base until around world 8, where I started looking for the world records. Oh, and you actually don't need to finish a world to save the bananas, you just need to beat world 11 then collect bananas. Still, no practice mode allowed...
Editor, Experienced player (854)
Joined: 5/2/2015
Posts: 696
Location: France
This TAS is impressive for a relatively low time worked on it. Some levels are not optimised at all (which I can understand if your point of reference was the first 10 worlds TAS) but there are other which are surprising. The best one was certainly Backslide, congratulations on getting that collision. It beats both WRs and previous TASes combined on 19 levels (6 ties, 13 WRs beaten). Out of the 110 levels currently TASed, humans or other TASes beat your TAS on 91 levels. I have put up a sheet of your times vs. previous TASes and human WRs: https://docs.google.com/spreadsheets/d/1Id58dEgl1iDiBxiaIkdcbS6_tRskOM9XUItfy84xPMs/edit?usp=sharing A yellow time means TAS wins, an green time means human win, and orange means a tie. I've linked to my TASes as well as other TASes for reference (only if they beat either human times or your times.) If you wish to try them out in the emulator, the movie files for all my TASes should be linked.
Taechuk
He/Him
Player (63)
Joined: 12/4/2017
Posts: 66
Thank you very much! Yeah, I made that TAS in a single month, so of course most parts aren't that optimised, and that other TAS isn't too great. I litterally picked up the game and said "I should do a TAS of that game" so, I didn't have much experience with the game and that's the second game I tried to TAS on DS. Anyway, thanks for the google docs, now I know which level needs optimisation. I'll do my best to beat those records.
Taechuk
He/Him
Player (63)
Joined: 12/4/2017
Posts: 66
Tried improving 1-2, didn't have much success, so I tried 1-3. Link to video 530 rerecords and I tied the WR. Got a bit lucky, but I'm sure there's room for more improvement. Edit : Actually nevermind. Saving a frame off would take soooo much time if you actually can.
Taechuk
He/Him
Player (63)
Joined: 12/4/2017
Posts: 66
Link to video Progeniitor 50.96 1226 rerecords
Taechuk
He/Him
Player (63)
Joined: 12/4/2017
Posts: 66
Link to video That was a really lucky bounce. Anaconda 53.76 828 rerecords
1 2
5 6 7 8