Key Takeaways
The simple idea
- Water Sort Puzzle is a mobile game style where colored liquid is poured between containers until each container holds only one color.
Why it shows up everywhere
- Many apps look the same because the core idea is easy to copy, fast to build, and fits ad-driven mobile trends.
What it can train
- It can practice planning and focus, but research on big “brain power” gains from games is mixed.
Why “reward” matters in AI
- In Reinforcement Learning, “reward” is just a number that guides learning, not a feeling.
Story & Details
A game that explains itself in seconds
Water Sort Puzzle is easy to read at a glance. A few tubes. A few bright layers. One goal: make each tube a single solid color. The rules are strict but simple. Pour only when the top color matches the top color in the next tube, and only when there is space. Restart is always there when a move turns out to be a trap.
This is why the game has so many close cousins. The core action is one tap, one pour, one small win. It is calm, quick, and clear. That mix is a strong fit for mobile play, where many people want a short puzzle that does not demand a long setup.
Two levels, two clean paths
Some stages look like they have already been cleared, with a “next” screen after Level Seven and Level Eight. Two other stages invite real work: one with five tubes (two empty), and one with three tubes (one empty).
For the five-tube stage, a clean route can be expressed with tube numbers from left to right. The pour chain runs like this, in one smooth flow: 1→4, then 3→4, then 2→5, then 3→5, then 2→3, then 1→5, then 2→5 to finish the red tube, then 1→3, then 2→3 to finish the purple tube, and finally 1→4 to finish the yellow tube. The point is not speed. The point is keeping one empty space alive so the board never locks.
For the three-tube stage, the same idea holds: protect the empty tube like it is a spare hand. The sequence can be carried through as 1→3, then 2→1, then 2→3, then 1→2 to stack purple, then 1→3, then 2→1 to complete purple, and then 2→3 to place the last yellow. Each move is small, but the order is the whole puzzle.
Why there are so many “the same” games
This genre sits inside a wider mobile pattern often called hyper-casual. These games are built to start fast, explain themselves fast, and keep sessions short. Many rely on ads, and many ship huge numbers of levels because the level “content” is mostly the arrangement of colors.
Tools like Unity make this faster. A small team can build a clean pour-and-sort game, change the art, tune difficulty, and publish a new version quickly. That speed is part of the reason the stores fill up with near-twins.
Law also plays a role. In broad terms, copyright does not protect the raw idea of a game or the basic method of play. It can protect the creative expression around it, but not the basic concept of “pour matching colors until sorted.” That leaves plenty of room for look-alike products.
What the puzzle may do for the mind
These puzzles are not just time-fillers. They push a few skills again and again: hold a plan, remember what is buried under the top layer, resist a tempting move, and recover when the board tightens. That is real practice of attention and planning.
At the same time, the best research caution is simple: getting better at a specific puzzle does not always mean broad gains in general intelligence. Evidence for wide “far transfer” from video game training is hard to prove. Yet many people still find clear value: a calmer mood, a sense of order, and a small daily challenge that feels good to finish.
A tiny Dutch lesson, built around the game
Dutch often uses short, direct imperatives for actions, which fits a puzzle like this.
Giet het in het glas.
Word-by-word: Giet = pour; het = it; in = in; het = the; glas = glass.
Tone and use: neutral and direct, common for a simple instruction.
Welke kleur bovenop?
Word-by-word: Welke = which; kleur = color; bovenop = on top.
Tone and use: casual, useful when checking the top layer before a pour.
Geen plek meer.
Word-by-word: Geen = no; plek = place; meer = any more.
Tone and use: short and natural, used when a tube has no room left.
When a machine “learns” this puzzle
The same puzzle can be described for Reinforcement Learning. The board is the “state.” A pour is an “action.” After each action, the system returns a number called a reward. The agent tries to choose actions that lead to higher total reward over time.
A reward can be designed in plain terms: give points when a tube becomes a single color, subtract a little for each move to discourage endless loops, and give a large bonus when the level is solved. Nothing in this setup requires feelings. The numbers are enough to guide learning.
Conclusions
Water Sort Puzzle works because it is strict, bright, and readable. The joy comes from turning mess into order, one small pour at a time.
Its many clones are not a mystery. A clear mechanic, fast tools, and a mobile market that rewards quick, simple play can multiply one good idea into a whole crowd of look-alikes.
And behind the scenes, the same puzzle can even explain a core idea in modern AI: “reward” is not praise. It is a number that shapes behavior.
Selected References
[1] Apple App Store listing for Water Sort Puzzle (rules and basic description): https://apps.apple.com/us/app/water-sort-puzzle/id1514542157
[2] Sensor Tower overview for Water Sort Puzzle (release timing and update history): https://app.sensortower.com/overview/1514542157?country=US
[3] Unity on hyper-casual games (definition and monetization background): https://unity.com/blog/what-are-hyper-casual-games-and-how-do-you-monetize-them
[4] Unity Engine product page (what Unity is): https://unity.com/products/unity-engine
[5] U.S. Copyright Office on games (ideas and methods of play): https://www.copyright.gov/register/tx-games.html
[6] OpenAI Spinning Up introduction to Reinforcement Learning (reward signal basics): https://spinningup.openai.com/en/latest/spinningup/rl_intro.html
[7] Sala and Gobet on limits of broad cognitive gains from video game training (abstract page): https://pubmed.ncbi.nlm.nih.gov/29239631/
[8] MIT OpenCourseWare video on Reinforcement Learning (single video reference): https://www.youtube.com/watch?v=to-lHJfK4pw
Appendix
Agent
An agent is the decision-maker in Reinforcement Learning: it picks actions, sees what happens, and learns which choices lead to better outcomes.
Hyper-casual
Hyper-casual is a style of mobile game with very simple rules, very fast start, short play sessions, and often ad-based business models.
Reinforcement Learning
Reinforcement Learning is a machine learning approach where an agent learns by acting in an environment and using reward numbers to improve future choices.
Reward signal
A reward signal is a number returned after an action that says how good or bad the result was for the goal the system is trying to reach.
Tube
A tube is a container that holds stacked color layers; in this puzzle, it is both the storage space and the main constraint.
Unity
Unity is a widely used development platform for making interactive games and apps, especially common in mobile game production.
User interface
User interface means the screens and controls a person sees and touches: buttons, labels, menus, and the way actions are performed.
Water Sort Puzzle
Water Sort Puzzle is a puzzle game pattern where colored liquid is poured between containers under matching-and-space rules until each container holds one single color.
My Diary - Leonardo Cardillo 