Cheery Chimp

Where it started

This one began with another game. I had been playing a small color puzzle on a grid. Colored squares sit in pairs, you drag a square along its row or column to its partner, and the pair disappears. Two different colors dragged together make a third, so yellow and blue give you a green to match with another green. Squares cannot move diagonally or turn a corner, and blocked cells get in the way.

The part I liked was the mixing. The part that worried me was building a copy of someone else's game, especially one that seemed to be the only one of its kind. So the plan was to keep the mixing and change nearly everything around it.

Five changes

Nothing disappears when two matching pieces meet. Each level has rings that want a color, and the job is to make that color and get it there. That one change turns a matching game into a delivery game.

A mix turns the corner. When one ball is pushed into another, the new ball appears where the second one was sitting. Push yellow into blue and the green is in blue's spot. Push blue into yellow and it is somewhere else. Which way you push became the main decision in the game.

Three colors make mud. Mud mixes with nothing, so a careless push leaves you with something in the way. It can still be moved, which makes it a more interesting mistake than one that simply ends the level.

Filters take a mix apart. A filter is a block with a color. Push a ball holding that color into it and the color comes out the far side while the rest stays behind. Even mud can be rescued, one color at a time.

Ice carries a ball until something stops it.

A small push

I did not want squares. I pictured soft balls that squash when they touch and only merge if you keep pushing. That turned out to matter for more than looks. Because touching and mixing are separate, one ball can rest against another and use it as a stopper, and that is what makes the ice levels work.

I also decided that balls sharing a color would bump instead of mixing. Red will not go into orange, because orange already has red in it. A ball you cannot mix with is a good thing to lean on.

Colors that add up

Under the paint, each color is a number. Red is 1, yellow is 2 and blue is 4. Mixing adds them, so orange is 3, purple is 5, green is 6 and mud is 7. Two balls can mix only when they have no color in common. The whole color system is a few lines of code, and a filter is the same sum run backwards.

The marks follow the same idea. Red carries a triangle, yellow a circle and blue a square, and a mixed ball shows the marks of whatever is in it. They are there for anyone who has trouble telling the colors apart, and they also spell out the mixing rule. Orange is a triangle and a circle.

The rule that had to be invented

"Balls move along a row or a column and cannot turn corners" sounds complete until you build it. If a ball can stop anywhere, two drags turn any corner, and the direction of a mix stops mattering.

The answer was to give every ball a home. A ball can only change direction at home. Once it leaves, it is on a rail and can only run back and forth along the line to its home dot. A mix makes a new ball with a new home, and so does a trip through a filter. Every level in the game leans on that one rule.

Thirty levels

My question after the first dozen levels was whether the game could carry thirty without new pieces. It could. Twenty of the thirty were drawn by hand, each around one idea: a filled ring used as a backstop on ice, or a ball that hitches a ride inside a mix and is filtered back out further along.

The other ten came from a generator, and it took two tries to get one that worked. Scattering pieces at random was useless. Of 600 random boards, not one was worth keeping. What worked was playing the game forward. The generator scatters pieces, makes random moves, and then puts rings under wherever the balls ended up, so there is always an answer. A solver finds the shortest answer, nudges pieces around to make it longer, and removes any wall, ice or filter that turned out not to matter. The ten were picked from what came out and tidied by hand.

It found something I had not seen. One filter can be used over and over, and that makes long puzzles out of almost nothing. Level 29 has three balls and one filter, and takes eleven moves.

Every level is checked by the same solver. Par is the fewest moves there is, not my best guess.

Dead ends

The hard levels have a cost. Because a ball cannot turn away from home, one wrong push can make a level impossible, and the board does not look any different afterwards. On several of the harder levels, more than half of the positions you can reach are already lost. On the last level it is 78 percent.

So the game works it out for you. When a level opens, it quietly maps every position that can be reached, about 43,000 of them on the biggest level, and marks which ones can still be won. Make a move that cannot be recovered and it says so and points at Undo. The puzzle should be choosing the right push, not finding out ten moves later that you chose the wrong one.

Where it is

There are thirty levels. The first twelve each teach one thing, which is why the how-to-play page is short. Three unsolved levels are open at a time, so one hard level cannot stop the game. There is no timer and nothing to pause.

The rules run without a screen, which lets sixty tests check them, and every level, without opening a browser. The art is drawn in code and the sounds are made on the spot.

If the game needs more later, white and black can join as colors to mix with.