Skip to content

Instantly share code, notes, and snippets.

@JoshCheek

JoshCheek/notes.txt

Last active July 25, 2016 09:15
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save JoshCheek/a3608ec2b7d849a7413df7e25f643b03 to your computer and use it in GitHub Desktop.
Save JoshCheek/a3608ec2b7d849a7413df7e25f643b03 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
notes.txt
* Working through http://ungineer.herokuapp.com
* Widgets: confirmed that I can click it to change its colour. Tried a few other ways of clicking (double clicking, dragging) but none did anything
* Constraints: Took a while, but eventually identified the pattern is red/blue/green/yellow, I'm remembering them by RGB, where RG=Yellow,
if you put them into a circle, it would be:
R Y
P G
B C
if se omit Pink and Cyan, it rotates counter-clockwise when you click it.
I then went back and checked that Widgets behave the same.
I didn't initially realize I could click the bottom one.
After some experimentation, I determined that when the top one changes to Yellow, it changes the bottom one to Green.
(Just tried it for all colours to confirm)
Note that I later went back and realized that the bottom one can be changed, too, and when it changes to B, the top changes to R
* Constraints Test: This one, I wound up wanting to back and forth a lot, eg I'd get it stuck and want to reset, or want to go back and try it on the smaller one.
I eventually realized I can click "Go Experiment" to do this, but it took quite a while.
Clicking Blue does nothing, I didn't anticipate this, was sort of expecting each one to change the one beneath it.
I note there are now two dimensions of relationships, which I'm not sure how to deal with.
Eventually I try clicking other ones, and they change all crazily.
I eventually start refiguring out the rules for this system since I don't see the relationship with the previous one.
Hypothesis: Red always goes to Blue.
Experiment 1: click topleft
Setup: RB | Predict: BB | Actual: RR
RR | RR | BG
Experiment 2: click botleft
Setup: RB | Predict: RB | Actual: RR
RR | BR | BR
Experiment 3: click right
Setup: RB | Predict: RB | Actual: RB
RR | RB | RR
Conclusion: no, experiment 1 violated this
Observation:
The two which didn't follow the expectation were horizontally/vertically adjacent to the Blue.
The previous setup was, as well, maybe I didn't test it appropriately.
Lets try the RHS of the square on the previous one
Hypothesis: These two stacked widgets will behave as the RHS of the square
Experiment: click bottom
Setup: B | Predict: B | Actual: R
R | R | B
Conclusion: This is not the case
I have no real clue, and it could be anything
* number of a given colour on the screen
* number of a given colour in relation to another colour on screen
* horizontal colour relationships
* vertical colour relationships
* diagonal colour relationships
* number of a given adjacent colour
* number of a given adjacent colour, but only in a certain direction
* the colour of some specific other spot on the board
* there could be an invisible dimension of colours that they are rotating through, thus different rules for the same visible board
* colours could map to something like number values and be modded by something (probably 3)
* any of these, but the rules change depending on which colour
* any of these, but the rules change depending on how many clicks in you are
* any of these, but the rules change depending on the starting positions
* any of these, but the rules change depending on which square you're clicking
* any of these, but the history of previous boards is relevant, too
Lets just try to map out all the behaviour we can identify:
number the squares like this:
12
34
RB+--1-> RR+--1-> BR+--1-> noop
RR| BG| BG+--2-> noop
| | +--3-> noop
| | \--4-> noop
| +--2-> noop
| +--3-> BR+--1-> GR+--1-> noop
| | RB| RB+--2-> noop
| | | +--3-> noop
| | | \--4-> noop
| | +--2-> noop
| | +--3-> GY+--1-> GR+--1-> YR+--1-> RR+--1-> BR+--1-> noop
| | | RG| GB| GB| GB| GB+--2-> noop
| | | | | | | +--3-> noop
| | | | | | | \--4-> noop
| | | | | | +--2-> noop
| | | | | | +--3-> RR+--1-> ??
| | | | | | | YB+--2-> ??
| | | | | | | +--3-> ??
| | | | | | | \--4-> ??
| | | | | | \--4-> noop
| | | | | +--2-> noop
| | | | | +--3-> YR+--1-> RR+--1-> ??
| | | | | | YB| YB+--2-> ??
| | | | | | | +--3-> ??
| | | | | | | \--4-> ??
| | | | | | +--2-> noop
| | | | | | +--3-> YR+--1-> ??
| | | | | | | RB+--2-> ??
| | | | | | | +--3-> ??
| | | | | | | \--4-> ??
| | | | | | \--4-> noop
| | | | | \--4-> noop
| | | | +--2-> noop
| | | | +--3-> GR+--1-> YR+--1-> YR+--1-> ??
| | | | | YB| GB| YB+--2-> ??
| | | | | | | +--3-> ??
| | | | | | | \--4-> ??
| | | | | | +--2-> noop
| | | | | | +--3-> YR+--1-> ??
| | | | | | | YB+--2-> ??
| | | | | | | +--3-> ??
| | | | | | | \--4-> ??
| | | | | | \--4-> noop
| | | | | +--2-> noop
| | | | | +--3-> GR+--1-> YR+--1-> ??
| | | | | | | GB+--2-> ??
| | | | | | | +--3-> ??
| | | | | | | \--4-> ??
| | | | | | RB+--2-> noop
| | | | | | +--3-> RR+--1-> ?? <-- this is apparently the winning state
| | | | | | | BB+--2-> ??
| | | | | | | +--3-> ??
| | | | | | | \--4-> ??
| | | | | | \--4-> noop
| | | | | \--4-> noop
| | | | \--4-> noop
| | | +--2-> noop
| | | +--3-> RY+--1-> noop
| | | | BG+--2-> noop
| | | | +--3-> noop
| | | | \--4-> noop
| | | \--4-> noop
| | \--4-> noop
| \--4-> noop
+--2-> noop
+--3-> RB+--1-> noop
| BR+--2-> noop
| +--3-> noop
| \--4-> noop
\--4-> noop
Things I looked at:
* A blue with a red clockwise after it (sometimes works, sometimes doesn't)
* A blue with a red counter-clockwise after it (sometimes works, sometimes doesn't)
* RB gives all noops, but half of BR yielded new boards
BR RB
* Frequencies seem interesting, I counted frequencies after exploring quite a few boards and got:
* position 1 yields a board 6 of 9 times
* position 2 yields a board 1 of 8 times (later explored more and found an error in recording, it is always 0)
* position 3 yields a board 5 of 8 times
* position 4 yields a board 0 of 8 times
This maybe implies there is a dependency of either a thing can change if the one to its right / below it is a certain way.
Since position 4 has nothing below it, or to the right of it, it can never initiate change.
Based on that, I went back and hit some unexplored ones, trying 4 first and then 2.
The frequency prediction held up.
Some of my numbers were wrong (the board did not change, I had recorded it wrong)
Now, we see that 2 and 4 have never yielded a new board.
Once I believed this, it was easier to convince myself to go down and check through the leaves.
The prediction held up and I discovered that there is apparently a winning state of RR (I didn't realize that)
BB
Now that I know there is a winning state, and the path to it (1331333), I'm noticing the number of "Tests" and "Clicks" change
I went back and checked if there was anything I missed that explained this was my goal, but I didn't see anything
I figure I should also look through the above map to check for a relationship between the "Constraints" section and the "Test" section
It appears that every time 3 changes to B, 1 changes to R, which is consistent, but there's not enough evidence to be highly confident
(best evidence is just that you intentionally put it there as the experimental blocks)
Same thing is true for 1 changing to Y implies 3 changes to G
So I guess the next thing to do is check if clicking 1 and 3 follow the order of R->B->G->Y->R
Immediately, we see they don't
In trying to check whether 1 or 3 is more constrained, I realize that my updated map actually shows they both always yield a new board
While trying to infer the horizontal pattern, I notice that R seems to be in 2 an awful lot
it's probably worth looking for a pattern there, but I'm going to try to figure out what "Tests" and "Clicks" mean instead
I notice:
Every time I get a wall of 4 noops, I have 0 "Clicks"
When I'm on the correct path, each time I click a colour, it reduces "Tests" by 1, except one of them also bumps "Clicks"
Writing out the path to success:
Position Clicked | Board | Tests | Clicks
-----------------+-------+-------+-------
| RB | 5 | 1 \
| RR | | |
1 +-------+-------+--- |
| RR | 4 | 1 |
| BG | | |
3 +-------+-------+--- |
| BB | 3 | 1 |-- Here, "Tests" drop by 1 each time
| RR | | | and the RHS changes each time
3 +-------+-------+--- |
| GY | 2 | 1 |
| RG | | |
1 +-------+-------+--- |
| GR | 1 | 3 / <-- here "Clicks" jumps
| GB | | \
3 +-------+-------+--- |
| GR | 1 | 2 |
| YB | | |-- Here, "Clicks" drops by 1 each time
3 +-------+-------+--- | and the RHS stays stable
| GR | 1 | 1 |
| RB | | |
3 +-------+-------+--- |
| RR | 0 | 0 /
| BB | | <-- here, both "Tests" and "Clicks" drop by 1
and the RHS changes
I don't see anything worth hypothesizing about,
so I'm going to try the next section using a strategy based on clicking and watching how the numbers change
and then going back and adjusting clicks to follow the most successful ones
like a timing attack on a password crack or something
* Quaterny: Well, they're going to be randomly generated, so I assume my strategy won't work
Playing in the sandbox, yields nothing obvious. I can't tell if it's related to the previous rules or whether these rules are new.
I think they are new rules. I think I'd be more compelled to try if I could set them into a given state and perform my experiments.
As is, I can get them into a couple of stable states, but still just don't know what the patterns look like.
* Quaterny Test:
Hypothesis: the RHS ones don't change when you click them
Conclusion: After a fair number of attempts, without seeing the RHS change, this seems to be true
(but I have no way of confirming my checks were holistic)
The pattern seems to match the previous one:
same number of LHS ones, each of which change when clicked, each LHS one has a single RHS one which doesn't change when clicked
I don't know what the goal state is
Looking back at the previous one, the experiment phase started with R and the test solution was RR
B BB
With this one, the experiment phase starts with R so maybe the test solution is RR
B BB
G GG
Y YY
The number of "Clicks" changes depending on the randomly generated configuration,
so I'm choosing one with a low number under the assumption that it's closer to the solution
the ones that are closer seem to have a lot of blue in them, IDK if that's significant
* Yeah, I don't know.
If I had the ability to set up my own experiments, I'd be a lot more inclined to continue
But I've looked at it for a couple of hours now and am not seeing any pattern in how clicking one on the LHS affects the RHS,
and still don't know what "Tests" vs "Clicks" means.
As of right now, the path that looks most fruitful is to go back and finish out the "Constraints Tests" mapping,
but Erik and Sarina just got in, so I'm going to chat with them for a bit.
Hope this was insightful ^_^
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment