In the wake of the
Scribblenauts TAS that was just published yesterday, I have a couple reroutes and minor optimizations to contribute. Many chunks of the input are untouched from the previous submission (including most of the tutorial and the world 3 grind) as I believe it to be optimal and the touchscreen input is a bit tricky to work with in a TAS environment. Below are things I looked into, and timesaves for anything that actually saved time.
in these notes, the A and P before a level number stand for Action and Puzzle respectively
University
Most of tutorial was untouched by me, although after my initial submission Samsara saved
00:01.02 in 0-2 by skipping the last prompt and getting the Starite immediately after breaking the plank.
In 0-9, I summon a Classroom instead of OJ for reasons that will become apparent in the next section.
In 0-10, it's
00:00.53 faster to use a Nothing (black hole), compared to an AFV clip in the previous submission.
World 1 (The Gardens)
The way the grind in A1-1 has to be done to work is to use words that generate multiple objects, because they are bugged and don't increase the par. This ekes out some extra money each time you do the level. During advanced mode, you're forced to get the Genius and Prodigy merit once each, so I summon two objects that I want to use later while that is forced to happen (I switch to Study in particular because in order to get past 4000 money I need a bit extra from summoning the new object. Study is a bit faster than Classroom so I continue to use it once the merits are no longer a concern). Each merit has to fade into view in the list, which takes
00:00.27 each; however, overlapping merits in this fashion is much better than getting them separately as the "merit get" menu does not need to load at all.
For a long version of the various revisions of this route I've gone through, see below:
My original submitted idea for A1-1 grind was to use a Living Room (abbreviated to the two-letter LR), because it's very fast to type. However, two days after submitting, I realized that it's an unnecessary timeloss to get the Entertainer merit every time (this happens because the Living Room contains a TV and a Console, and the merit triggers when you summon two entertainment objects). Despite this, the route saved
00:09.54 over the previous submission. Some of this timesave is from not using the white OK button on the level and world select, as this loses a few frames (maybe like 4-8 each) in some cases, but the majority is just because both P1-1 and P1-2 are slower.
My second idea was to use Study (office), which contains Computer, Filing Cabinet, and Clock. I had to do a bunch of finagling to actually get Maxwell to reach the Starite, but it worked out in the end. This route saved
00:05.50 over route #1.
After seeing the whole 2nd route through, I tested the 3rd route. Classroom is an object that contains Table, Chalkboard, and Apple. The apple is of particular interest because it can be used for the chef in the tutorial, bypassing the initial All New and New Object merits in A1-1. Unfortunately, this makes me lose out on 25 Ollars, just barely putting me below the required 4000, but I solve this by summoning Study during advanced mode where you're forced to get merits anyway. This route saved
00:00.62 over route #2.
The next day, I realized that you could get away with summoning Nothing in 0-10 if you summon AFV during advanced mode, while a merit is already guaranteed to happen. AFV is actually faster than both Study and Classroom (
00:00.23 faster than study), but can only be done one time because of the 25 ollars from All New + New Object making up for the 20 penalty from summoning an object that counts towards par. Funnily enough, it seems like getting the two merits loses
00:00.53, which cancels out the timesave from the tutorial, but the level itself is faster at least. Once Study is summoned for the first time and it's no longer a new object, there is no reason to not keep using it, since it is slightly faster to type than Classroom and saves
00:00.15 each time.
I'm not 100% certain on every individual timesave figure from each revision listed above. However, altogether I managed to save
00:13.31 on world 1 alone.
World 3 (The Peaks)
World 3 was almost entirely untouched.
I attempted to grind on both A3-7 and A3-11 to see if it would be faster (A3-7 because you don't get merits after doing it except for advanced mode, and A3-11 because you get more money per completion). Unfortunately, neither of these were faster, although 7 was maybe a bit more promising than 11. It may still be faster to do 3-7 once, if you use Nothing during the tutorial, but it didn't seem like it from rough estimates (you still get a new object merit no matter what if you do it this way, since you'll get it on A10-1).
Below are the methods I used to grind 7 and 11, in case they are of interest:
World 10 (Mish-Mash)
A10-1: I tried saving time by touching the Starite as Maxwell is falling down. While I was able to touch the starite, this did not make him pick the star up faster and actually lost an extra frame.
A10-2: saved
00:00.30 to doing the AFV clip from the start without moving Maxwell.
A10-3: replaces A10-4 due to being
00:01.37 faster (I think 50/50 due to fewer merits and due to having a slightly faster solution). I did attempt to optimize A10-4 further with a freeze ray strategy but this lost time. A10-6 seems less laggy and could potentially be faster than A10-3 if not for the fact that the hat that Maxwell wears at the start of that level counts as a new object and gives merits. If there were some convenient way to summon a top hat earlier in the run, this would be another routing optimization to look into.
In my original submission, the combined timesave upon entering A10-11 was
00:11.33. I originally did a container glitch using Bilbo and Bag to save a few letters (i.e. the same strat as the previous submission but better words), and pulled ahead by 2 frames even with a slightly slower setup. However, a day after submitting, I realized that the jetpack might have just enough airtime to get above the gates and path to the Starite from there. I spent some time trying to get the earliest possible final input; this requires panning the camera left for the entire duration of the level and tapping the Starite in the corner as Maxwell lands on the gates, while giving him a good landing angle so that he can stay on top of the gates while in autopilot. With this revised strat, I save
00:01.14 in this level for a total of
00:12.47 saved over the previous submission.
As of the most recent revision, the timesave appears to have jumped to
00:21.11.
Further timesaves that could happen
- Even more tutorial optimizations
- Maybe a reroute in world 1, 3, or 10; if my notes for world 1 are any indication of how chaotic it can get
- Minor lag optimization
- Using Scribblenauts Collection...
Regarding Scribblenauts Collection
In RTA, I consider Scribblenauts Collection to be the superior choice due to the button controls enabling far more consistent and less frustrating movement, as well as introducing the double jump glitch, both of which allow for more efficient grinding in the two grind levels in this route. Another major difference (although unimportant for RTA any%) is the fact that air travel now has unlimited airtime. There are many other differences that are irrelevant to this TAS.
In a TAS environment, the startup screen is
00:02.29 slower, and many object actions that took a single tap in the original release now force a multi-option menu. This most prominently applies to exiting the tank (as well as mounting/dismounting other vehicles, attempting to path to an object that is far away, and unequipping gear that Maxwell is wearing), and loses ~
00:00.08 each time it comes up.
I made a TAS of Collection and put it
side by side with the most recent revision of this run, and unfortunately it ended up being
00:01.40 slower. When I was watching it back afterwards, I noticed that some loads may be different in Collection, as it sort of pushes ahead and pulls back in the first 3 levels of world 10, on a scale greater than just the difference from not being able to exit the tank immediately. I don't know the full potential impact of load differences between versions, and I didn't even consider that this was a possibility at first.
As far as this submission is concerned, it does seem like the original release is the way to go, but I'm still unsure if that's also true for a main branch (i.e. All Levels) TAS, since I think it will ultimately depend on whether indefinite jetpack time is more important than being able to use the container glitch.
DrD2k9: Updating with
553 frame improvement after numerous re-works by the authors. Note: The current encode is not representative of this improvement.
