I figured out a pattern! Well actually I'm halfway there but its consistantly working... I also have figured out why when people try all combinations that it doesn't work. I've been taking note of the positions of the water and the valves/levers (which I will refer to with the generic term "switches") and have found a lot of repeats when they're moved in the same place. Unfortunately, now I'm short of time but I'll explain my methology and conclusions so perhaps you guys can continue in my path that has so far proved to be promising.
My method:
1) Take a screenshot of all rooms.
2) Enter a room.
3) I would change the switches systematically in one room, trying every combinations until either something happened or I've exhausted all of the combinations.
4) If something happened I would immediately take a screenshot of all rooms again, compare changes, record them and repeat in a new room. If nothing happen I would move all switches to exactly how they were before I tampered with that room and would repeat this in the next room.
Raw Data (up to this point):
http://esnips.com/doc/52b56475-99d2-47a ... puzzle.zip
How to read my data:
The file name is a description of the room number (that's in the url) as designated by "r#" r1 = room 1 for example.
It will then be followed by 2 or 3 letters which describe the postion of the corresponding valvues/levers. Valves are noted by p for "plus sign" and x for "x position" while levers are noted by u (for up) and d (for down).
The insides some files contain more information. Intially, when I began, I only took screenies of that room and that was all but eventually I began circling changes in each room with red to note changes. Towards the end, I began to take screenies of the other rooms and put a small versions of them in green boxes and noted changes in them too.
Speculations/conclusions thus far:
Some files will have an extra number in the end of the file name. This would be because some switch combos had different water positions; however, even so there were overlaps. When I discovered this I began to put the other rooms in green boxes because it seems as though the variation could be accounted for by influence from the other rooms.
Overall, I consistently found that certain positions of the switches would reroute water a simular way. It seems that the positions designate where water is allowed and if all three rooms allow water in the same spots then water would appear there. That would also explain why sometimes we see no change... because two positions in the same room both allow water to be there so the effect is null or the places of another room block the places in the current room. So the way to solve, after no changes are found, is to change it to exactly how it was before as to eliminate confusion by changing only one room at a time. I so want to solve this but I got to go very soon...
What you can do with my data:
See if the positions would do simular things as the examples I've provided. There are likely many other position combinations that I haven't had time to find yet so keep following atleast for several stepsand say if you observe simular changes.