ForumArcade ► Unnatural Selection
Unnatural Selection

Try to make every genome in the population match the target genome. Press the "Breed" button to create the next generation with parents chosen from the current one. Click and drag to toggle genomes - selected genomes have double chances to be chosen as a parent. Use the "radiation" slider to adjust the rate of mutation per nucleotide per generation.

Shortcuts:
  • Space: breed
  • A/S: Select None / Select ALL
  • D: Show/hide guide
  • Q/W/E/R: Min / less / more / max radiation
  
Fun!

Took me a minute to realize that selecting and unselecting were actually inverted from expectation.
  
Feature reqeusts
  • "Deselect all" button
  • Use a different border color (or some other indication) for each individual gene that matches the target. In the beginning it'd be great if the matches show up clearly and in the end game it'd be great if the subtle differences showed up more clearly. Oh geeze, I never thought to click on "Enable Guide". D'oh.
  • Some sort of overall match indication for each genome. e.g. "25%" next to each row
  • Quick shortcut keys to toggle 0% and 5% mutation rates
  
Just making sure there's no mechanic in there for linked traits, or else I've found a bug-

I've whittled a population down to just two genotypes (they're all identical except for the last two genes)
About half of them end in green-green and half end in blue-yellow. I'd expect after some breeding to see a few green-yellow and blue-green genotypes, but I'm seven generations in and those are still only the two.

EDIT: I tried another iteration and got the exact same problem, just ending in blue-red and green-blue, with no blue-blue or green-red showing up over several generations. I went to 20 generations this time just to be sure, but all my little color bars are refusing to intermingle.

I've got screenshots if that will help, once imgur fixes itself.
  
I noticed this problem as well. It's possible to get the right combination after enough tries, but there definitely seems to be a biased correlation.
  
Good catch. It was an off-by-one error in the code that chooses the recombination index. It was never possible for the genomes to split just before the last nucleotide.

The fix plus the keyboard shortcuts are live.
  
Make it touch friendly and release it for iOS and Android. There should be build targets for both of these in the default TwoCans project build file.
  
I'll probably do that at some point. I haven't deployed an iOS app since college and an Android app since ever.

I improved the UI and shortcuts to be more ergonomic. I also added the number of matching nucleotides next to each genome when the guide is turned on, but I think it might make it too easy and thus not as much fun... Not sure.

Edit: You can now click and drag to toggle multiple genomes at a time without clicking again.
  
After breeding, start out all the new samples UNSELECTED. I'm ragequitting now after the fourth time of getting almost done and then accidentally breeding out the best candidates by accidentally unselecting them.
  
I agree with GWS that would be best at the beginning, although the current system is good near the end. A toggle control would be ideal, but if it has to be one or the other, then please go for deselect all as the the default.

The guide is fine. If you blindly follow the numbers, then you risk losing alleles where many others are right but the mismatches are on the same site.

A generation count might be good as a way to track how effective you have been.
  
I can't update the embedded version because Blake's Crayon compiler server isn't running right now, but I've uploaded a new version here.

Changes:
  • Replaced Select All/None buttons with Dflt. Select toggle, which is off by default. Hotkey: S.
  • Added a generation counter.
  • Changed Hide/Show Guide hotkey to G.
  • Added a target genome directly below each column of the current population.
Also, interesting target genome I found while debugging:

  
36 generations.
  
Wow, nice! I think that's at least 10 generations better than my personal best. 😅
  
My previous bests were mid-40s, so I think I got lucky.
EDIT: Or not. I just did it again. :D

Losing a colour from a position is by far the biggest slowdown, so I'm very defensive about that.
  
Could this now be updated in the Games section?
  
༼ つ ◕_◕ ༽つ GIVE ONLINE CRAYON COMPILER ༼ つ ◕_◕ ༽つ
  
Port the Crayon compiler for JS, to JS, so we can get client-side compilation.

Partially a joke but maybe webasm can do this.

Edit: Yes, C# to WebASM has been done.
  
34generations. :D
  
That is super impressive! 👏
  


I think any further progress is down to luck at this point.
  
That's a big jump! I don't think I'll ever beat you at my own game. 😄

I may add a new mode to this later that's more puzzle-y, with more control over the mutations and/or breeding to more consistently reward good strategy and understanding of the processes.
  
If you do find time to do more development, please make the guide default to on. Also see if you can poke Blake for whatever is necessary to get it updated on the games page.

Some suggestions for new stuff...
  • the ability to click on a column and highlight all the correct instances.
  • a limited ability to exclude rows from the breeding round.

I'm a little concerned at the idea of rewarding "correct" strategy. Surely a lot of the point of the game is to discover what works? On the mutations front, how about an option to add a set of all one colour to the breeding pool? That would help with the scenario where you are wasting generations trying to regain a colour in one position.
  
Yes, reducing the luck needed to get a required letter when it's not present anywhere in the population would be my goal. Not sure what form that will take.
  
The guide now defaults to on.
  
Forum > Arcade > Unnatural Selection