This is my favorite image from the 111×109 set, and at this moment any set.
The image is a result of a lot of work – both by my computer, and myself. Although I changed the color of many lines within my program – it is still as it came out of my program with no post-editing (though, my program does post-process each image to boost contrast).
The level went through multiple stages of image generation and line editing to reach it’s current form.
Two Other Versions – Before and After Edits
Version 1:
The edited version might have been the only image of this level in the gallery – there were 3 days between when I “finished” it, and I started editing another original.
I think this is a perfect example of my approach to line-editing when I started working on the gallery: treat each sector as its own piece and try to make each one as good as possible.
That being said – there are some nice things about it. I plan on revisiting it later to see if I can make it a bit more harmonious. I almost included it in the gallery anyway (as it was my plan , until I didn’t.
Version 2:
The left image was my first choice to edit when I decided to make an alternate version of the level for the gallery. I really like the look of the center – maybe even more than what made it into my gallery. The big rainbow also stands out very nicely.
The last of the 43 edited revisions is nearly the same as the original – the only difference is in the top right corner. It took me an hour and a half to get there – one hour to make changes everywhere, and 30 minutes to undo almost all of those changes because I messed it up.
Even though I “failed” at editing it, I think this version is just fine as it is. Still, I couldn’t think of anyway to significantly improve it that I hadn’t tried already, and it didn’t have the look that I wanted for the gallery, so I moved on to another original.
The Current Version, In Four Stages
Top Left to Top Right
The biggest difference is between the original (top left), and my first major edit (top right). I always envisioned this level with rainbow colors, given the rainbow pattern – and I was really happy that it worked with this version. I spent just 30 minutes on the edit, which I started almost immediately after I finished the previous version (which made me almost completely forget that I edited a 2nd version of this level).
After this revision, the image not only was clearly better than my edits of previous images – it was my favorite image in the set.
Top Right to Bottom Left
While first edits focused entirely on major lines in the center (with 1 minor exception), the edits done from the top right version to the bottom left version were entirely focused on the edges.
At first I was not at all sure about my changes to the top right. While scrolling through the images in the sets later, I found that my changes had the desired effect: put the focus on the rainbow and the “pot of gold”, at least on first look.
This revision also took about 30 minutes – the resulting image was not only my favorite in the set, but a clear contender for my favorite image of any set.
Bottom Left to Bottom Right
This revision was actually multiple revisions – with 4 separate sessions totaling maybe 3 hours, resulting in multiple intermediate versions that I won’t show unless asked. This doesn’t include the hours I spent obsessing over them, questioning my revisions, and trying to find and think of fixes to flaws in them.
All but one of the color changes shifted no more than 1 shade or tone from the previous version. The exception is the blue line between the magenta lines bordering the rainbow – this went a full 2 shades lighter (from #4 blue to #2 blue).
This change, along with 2 other minor shifts, makes the rainbow feel much fuller, in my opinion.
The other changes were mostly in the center area and directed at making it slightly more modest yet at the same time a bit more colorful.
One little area that I spent probably far too much time obsessing over:
This is the bottom right area of the board beneath the rainbow’s end.
The top is the version from the bottom left revision and the bottom is from the bottom right revision.
The big almost black gap from the end of the purple line to the green line bothered me, and the middle version was my first attempt to fix it.
The problem is, it didn’t feel right to me – and I (think) I figured out why: the dip that the purple line makes just doesn’t fit with the perfect straight series of lines going up and down from there.
My fix was to make the short straight line blue – filling part of the gap. The blue is as bright as I could make it without writing some code; as the shorter lines are restricted to a 6 color subset of dark colors.
Still, I think it is just enough and strangely fitting with the rest of the central pattern.
Admittedly, it is highly unlikely if I didn’t point it out anyone would notice a real difference had I left this area exactly as it was.
Similarity Shading Is A Hero (for this image, at least)
Here’s the image in the gallery next to a version of that image without any shading:
In my opinion – the unshaded version is much too busy and distracting from the main pattern. The most noticeable difference is around the edges and corners.
The image in the gallery has green/blue shading (usually, I just call it cyan, though the B/G values are calculated separately).
Most of the images in my galleries have red/blue shading (which I call magenta). Here is the image next to a magenta shaded version:
The edges and corners are fine in the magenta version – I slightly prefer the cooler edges, but that’s a matter of personal preference.
The magenta shading has made the center worse by highlighting the hidden dark red line (you can see this line if you want in the first version that I edited above).
An even bigger problem I think is the red line bordering the left side: it is way too bright in the magenta version (and the unshaded version). This is where the cyan shaded really worked, in my opinion – the shading points the focus inwards towards the center and rainbow.
It does this pretty consistently all around the board, which is why I think it works so well for this level.
Constellation Map and the Shading on Its Own
Here’s the constellation map of the level with cyan shading. For a constellation map I remove all the lines, make the dots a star-like color with slightly varied sizes, and set the similarity shading to full.
There are various settings that I use that change the similarity shading – the shading that gets used for the image is much less than what you see here.
Here’s what I get if I use the same settings used to generate the gallery image, but get rid of all the lines and dots:
Quite a bit darker – but also not (quite) an accurate representation of how the shading affects the image. The issue is that the settings I used increase the contrast of the image after initial generation. This is what is actually overlaid onto the image before the contrast adjustment:
Finally, I’m going to show what you get if you subtract the version without shading to the gallery image and vice versa.
The left shows what is “added” to the gallery image, and the right shows what is “subtracted” from the shadeless version.
One major point – the “rainbow” is almost completely unaffected by the shading. This makes it stand out even more, and is a perfect example of what similarity shading was designed to do: highlight the more interesting parts of the level.
This Comparison Makes Me a Bit Sad..
But I have to make it. Here it is against it’s main competition: level 90 from 100×100, shown with updated colors here: two new images.
The sad part to me is that the two patterns are extremely unusual – and probably closer to each other than any other image in any of my galleries.
The levels are actually quite a bit different – but it’s hard not to see many similarity in the smaller patterns.
Color-wise – they have a similar rainbow-like overall scheme and both feature a bright yellow line surrounding a red line as the main focal point.
The differences:
- cyan shading (left) vs. magenta shading (right) makes the left side significantly cooler (as in cold). The red isn’t nearly as bright, and there is a lot less yellow.
- the right image tries to make every part of the board relatively “equal”; the bright magenta lines are the most obvious example. The left image has a much more clear focal point with several subdued areas.
- the darkest 6 colors of the 30 color theme used for the left are a lot darker than the darkest 6 colors of the 24 color theme used for the right. This wasn’t intentional – I didn’t take into account the effect of the contrast boost post-processing. Still, I liked and took advantage of the nearly-black colors of the new theme often.
- both levels have one fairly large contained area filled with… mostly short “junk” lines. For the right image, my fix was to show 1 crazy bright line in the center of this, and make everything else dark. For the right image, I used a series of shorter lines to create an interesting effect with the colors. None of the lines are anywhere near as interesting as the magenta line on the right, but there is considerably more depth to the left image in this area in my opinion. Admittedly, this would not have worked nearly as well if the dark colors weren’t darker than the 24 color theme.
My approach to editing the right image was to emphasize every impressive line – and there are quite a lot of them. My approach to editing the left image was to make the big pattern stand out and make the rest of the board fit.
The right image is about the lines, and the left image is about the overall look.
I’m Going to Eat My Own Words, But
The final version of this level is my favorite image of all of my galleries, easily surpassing my previous two favorites. My explanation for why those images were my favorites was fairly technical – complex big patterns, crazy interior lines, etc.
My reasoning for liking this one the most is much more basic. It looks the best. The test I perform when trying to decide on favorites is to switch back and forth between them and gauge my reaction. This image always wins, so it’s my favorite.
I Don’t Know Why I Track This Stuff, And I Don’t Know Why I Write About It
I wrote this before, maybe someone cares.
Stages of Image Generation and Line Editing
My program generated a total of 272 images with randomized line colors in four batches:
| Date | Number | Batch |
| 6/15/2018 | 12 | The initial batch of images generated for all 60 levels. |
| 6/15/2018 | 40 | 17 of my favorite levels. |
| 6/16/2018 | 100 | My top 2 levels before line editing. |
| 6/20/2018 | 120 | Prior to work on an alternate version. |
Of the 272 images – I made manual line color changes to 3 of them (though I considered many others). While I am editing line colors, my program generates many versions of each image with a small set of changes so that I can look back at them and compare them.
Although I try to finish each level in a single session (typically around 30 minutes, but sometimes longer), for this level I edited lines in 8 sessions and 7 different days:
| Version | Revisions | |
| 6/16/2018 | 1 | 47 |
| 6/17/2018 | 1 | 15 |
| 6/20/2018 | 2 | 43 |
| 6/20/2018 | 3 | 20 |
| 6/22/2018 | 3 | 16 |
| 6/23/2018 | 3 | 29 |
| 6/24/2018 | 3 | 15 |
| 6/25/2018 | 3 | 35 |
I estimate I spend about 1.5 – 2 minutes of manual time per revision – my program takes about 1 minute to generate an image, and besides change the color of the lines, I look at each image compared to the previous image and decide which I think is better.