Another year, another 30 seconds. All I had to do is use orange.
Game objectives
- Color all 16 dinosaurs such that no two adjacent regions are the same color
- Do so using the smallest number of colors (which is 4)
Introduction
See
last year's submission for lots of juicy details on the theorem and the algorithm. There was no great breakthrough or optimization; in fact, I didn't change my code one bit! I just had to change the color definitions.
So, remember how I said that the two-color gradient patterns (which are treated as regular colors for theorem purposes) require a layer of primer, thereby taking double the fill time? Well... what if that primer were already there? Yeah, that's what happens when you use orange. It turns out the primer layer is not always pink, but the third color not in the pattern, so any pink/yellow pattern will take only one pass to fill. I'll be honest with you, I find this a little disappointing. The gray regions were the heart and soul of the colorings. With three identical colors, it barely matters which ones you pick anymore. That said, there is a lot more opportunity for creative routing in terms of dual colorings and trivial differences, so it's more interesting than I initially thought.
To quantify this "barely matters" gripe, by the way: The spinosaurus (#2) went from 4 optimal colorings with gray, to 3072 optimal colorings with orange.
As a refresher, there are two coloring modes: point-to-point (p2p) and freehand, which you access by pressing A and B respectively. Freehand lets you move the pen during the fill, while the p2p pen gets stuck in place; this means freehand is faster, ceteris paribus. However, the freehand pen moves 2 pixels at a time, so certain tiny regions may be inaccessible. See last year's submission for more details.
I started each dino out with that I call a color slap: taking last year's gray coloring and just switching it to orange. Then a quick check of my algorithm to verify that the same regions should indeed stay purple, which was the case for all 16 dinos.
Freehand Optimization
Naturally, a color slap rarely worked on its own, because I had spaced out my gray fills to take advantage of all that waiting. But since all colors are created equal now, it was fairly easy to create little orange pit stops to make up for that without wasting any time.
p2p Optimization
It was much harder to find any improvements over basic color slaps on these ones, but I did manage to find a few. The goal is to take certain regions and save them for later in order to minimize backtracking now... or vice versa. For example, if you need to use pink to go 5 -> 10 -> 2, maybe you could save that 10 for the orange pass and go straight from 5 -> 2 with pink.
You also need to make sure to interleave your B mashing with your right mashing when switching colors, especially since it takes 4 mashes to get to orange. Any movement of at least 3 can reap the full benefits of this (B, right, B, right, B, right, B, A)
Definitions
- Color Slap
- A baseline coloring where I took the old movement and just swapped gray to orange.
- Perfect Slap
- A color slap that gains 0 meander frames and needs no further refinement.
- Pit Stop
- A region that provides a brief reprieve along a pen journey that would otherwise be too long. Typically a region that wouldn't have made sense as gray before but is now free to be orange.
- Meander Frames
- How many frames were wasted meandering with the pen after I could have pressed A. The absolute theoretical minumum is 0 for freehand dinos, and V-1 (V=number of regions) for p2p dinos.
- Miracle Frames
- Unpredictable timesave that pops up based on where in the region you started the fill.
- Miracle Pixels
- Pixels that grant you miracle frames. Seems to favor weird corners.
- Dual colorings
- Since yellow, pink, and orange are identical (mostly), any coloring has many twins where all these colors are swapped. I chose the mathy word "dual" to refer to this.
Due to the way I organized my files, I'm able to tell you the exact number of rerecords I took on each dino. Isn't that neat?
#1 - Stegosaurus
Frames saved: 122
Meander frames: 110
Rerecords: 25
| Old | New |
 |  |
I ran out of time for this one, so I just went with the basic color slap. Hey, if Romero saves the first level for last, so can I.
#2 - Spinosaurus
Frames saved: 88
Meander frames: 18
Rerecords: 67
| Old | New |
 |  |
A perfect slap! Yes, there are 18 meander frames, but those were there from last time. I lost no time from switching to orange.
#3 - Brontosaurus
Frames saved: 55
Meander frames: 9
Rerecords: 528
| Old | New |
 |  |
Technically an imperfect slap, but nothing else I tried was any faster.
#4 - Viktor
Frames saved: 154
Meander frames: 13
Rerecords: 1110
| Old | New |
 |  |
Finally, a unique coloring! This is actually the first one the algorithm spat out rather than a salvaged slap. It saves 4 frames over the basic slap.
the main bottlenecks in the basic slap are 15 -> 8, 19 -> 16; all the alternatives I tried aimed to eliminate these particular meanders.
#5 - Triceratops
Frames saved: 31
Meander frames: Perfect!
Rerecords: 416
| Old | New |
 |  |
The main idea here is to use region 5 as a pit stop between the eye and the ground/feet regions.
It actually takes 1 frame to reach region 4 from the starting position, and that 1 frame was the only thing between me and perfection. So I decided to rework the coloring a bit to make region 7 (where the cursor starts) pink, and scrounged up an actual perfect coloring.
#6 - Vinnie
Frames saved: 23
Meander frames: Perfect!
Rerecords: 155
| Old | New |
 |  |
Pretty underwhelming timesave, but that's just because there were so few gray regions before. The movement was perfect before, and with a pit stop on the arm region, this one is, too.
I actually got 2 miracle frames from the arm region alone using this switch. Never seen that before!
By the way, when I run the algorithm on Vinnie, I run out of heap space. Not even 16 GB is enough (yeah, I got that much RAM lying around)! It's those 3N possibilities on the spots (8-17). Freezing those damn spots makes it run almost instantly. It would've been 2N last year (just pink and yellow), so why didn't I run into this problem then? Well, it's 1,024 vs. 59,049, that's why.
#7 - Pterodactyl
Frames saved: 57
Meander frames: 11
Rerecords: 48
| Old | New |
 |  |
Another perfect slap.
#8 - Crested
Frames saved: 85
Meander frames: 4
Rerecords: 346
| Old | New |
 |  |
Region 13 is used as a pit stop here, plus I reworked the titular crest so I could go straight to region 15 and perform the titular action on this member of the titular species. Titularly.
While it seems like an extra pit stop in the eyes would be helpful, it really isn't. Pink also kind of needs that pit stop. I may have to revisit this in the future, because I don't actually have a branch that does it.
#9 - Duckbill
Frames saved: 241
Meander frames:
Rerecords: 248
| Old | New |
 |  |
A basic slap that loses 2 frames between regions 8 and 3. I had one other coloring that makes region 0 orange in an attempt to provide some waiting time, but it's 23 frames slower.
#10 - Vern (Virgin's own!)
Frames saved: 199
Meander frames: 58
Rerecords: 378
| Old | New |
 |  |
I don't care what the rerecords say, this is the one I put the most effort into. The main idea is to avoid switching from orange back to pink, which I accomplished by coloring 6 orange, and swapping regions 5, 8, and 31 around (and painting the toenails to match 💅) so that I would hit them along the way with their respective colors.
#11 - Crested Duckbill
Frames saved: 173
Meander frames: 40
Rerecords: 211
| Old | New |
 |  |
I guess you could say this is an adapted dual of a basic slap. I hope you're keeping up with your vocabulary. The main idea is to avoid an awkward backtrack between 6 and 22/25.
#12 - Vera
Frames saved: 130
Meander frames: 59
Rerecords: 22
| Old | New |
 |  |
Basic slap. I didn't have time to investigate this one, either.
#13 - Mini Stegosaurus
Frames saved: 235
Meander frames: 4
Rerecords: 678
| Old | New |
 |  |
What a beautiful sunset :')
The idea here is to make the ground orange so I can have a nice buffer after one of the tiny regions 27 or 28. Then I make an alternating pattern on the spines so I can have lots of pit stops for yellow and orange along the back.
#14 - The Mighty T-Rex
Frames saved: 131
Meander frames: 82
Rerecords: 286
| Old | New |
 |  |
This is a basic slap. Technically all p2p slaps are perfect, because you have to wait for the full fill before you can move the pen. Again, I really hope you're keeping up with your vocabulary, here. A few other solutions came within a few frames, but nothing faster.
#15 - Torosaur
Frames saved: 132
Meander frames: 3
Rerecords: 587
| Old | New |
 |  |
The main idea here was doing an alternating pattern on regions 8-16 so I could have lots of nice pit stops for pink and yellow. Then the orange part was easy.
I actually got a coloring with an exact frame tie. It's a similar idea, but I just like this one better. I think this one has a miracle frame, too.
And hey, don't forget the forgotten pixel! It's region 29.
#16 - Protoceratops
Frames saved: 129
Meander frames: 2
Rerecords: 181
| Old | New |
 |  |
The main idea here was to make the background orange so I could have some breathing room between the tiny regions.
Totals
00:33.06 saved in all. This is the actual real life frame count. I occasionally had to move the p2p pen a couple of regions aside to reveal the delicate features it was hiding, which I specifically factored out of the dino-by-dino notes, so this number may be a few frames off from what you expect.
The rerecord count for this one is substantially lower than last time. That's because I already had a strong basis of colorings and code, so I could experiment quickly and confidently.
FAQ
Full Anticipated Questions, AKA, FDHMEMT (framing device to help me express miscellaneous thoughts).
- How did you miss this orange thing?
You know what happens when you assume... I only discovered it by accident when I was adapting a coloring (Vern, I think?) to use orange instead of gray since I was wrapping back to pink.
- Should this count as a different category or something?
I definitely see what you mean. It does feel very different than before. But what would the categories be? Graymode/orangemode? First Four Colors? 16 Star? That doesn't make much sense to me. The only thing I could think would be to do an "ocean" mode (see below FAQ) that leaves the backgrounds purple and requires you to use pink, yellow, gray, and orange. But even that would be weird because the first palette actually has a second color 1/color 2 pattern, unlike all other palettes. So suddenly you're locked to the red/white/blue palette. So... oceanic, no funky patterns?
(copied from previous submission)
- This algorithm of yours... may I see it?
- Why do you switch palettes? Isn't that slow?
Yep; it's a time/entertainment tradeoff. I just think it's easier on the eyes. This run would sync on any palette;
even though I can move the pen during the palette swaps, I chose not to so that it could sync on the first America palette.
I might make a version that uses that palette.
- Can't you do anything about all that waiting?
Uh, yes? I just saved 30 seconds of it. Besides, you're not playing, you're just watching. It's ALL waiting to YOU.
- I mean, can't you do anything funny with the pencil during the flood fills?
There's not a whole lot I can do. There's no music to jam to, and I'm unable to create sound effects of my own.
In the p2p dinos, I can't even move my pencil, so I am TRULY stuck!
In a couple of freehand dinos, I had to modify a coloring halfway through; in those cases, I implemented the change as a last
second juke. I also tried to put the pen somewhere cute at the end of each freehand coloring. I could always do more, but
would it really make a difference?
- Does it matter what order you do them in?
Nope! Since you have to end each coloring by pressing start, you can't end input early on a long region.
- It's kind of bogus that the background regions count. Aren't those like the oceans in a map?
Fair point. Apparently the RTA community calls that Any%, whereas my interpretation is 100%.
I think the 100% version is more interesting.
- Can't you just reset instead of listening to that stupid jingle?
The title screen is slow. This would lose exactly 200 frames each time.
Final Thoughts
That's it. Bye bye.
