--- title: "The research behind it" description: "The 179 sources behind Professor Oops, on play, simulation, real facts and respect for children, each labelled by what it really did for the game." source_name: "Game Tailors" source_url: "https://professoroops.com/en/inside/research/" site_home: "https://professoroops.com/" source_description: "The drawing app where the paper is real. No ads. No accounts. No personal data: only anonymous counts, which you can switch off." locale: "en" legal: - "Game Tailors · Schieweg 15Y-40, 2627 AN Delft (NL)" - "KvK 92437702 · VAT NL866044474B01" - "info@gametailors.com · +31 6 4227 5919" copyright: "© 2026 Game Tailors" published_at: "2026-11-10T12:03:38.846Z" modified_at: "2026-11-10T12:03:38.846Z" --- # The research behind it ## We read 179 sources, and say what each one really did Professor Oops draws on research about how children play, how to make water, wind and fire behave on a screen, how real animals and kitchens work, and how to treat children fairly. We built the toy first and read much of the learning research afterwards. So every source carries a label: applied directly, informed the design, or found later. Some of it pushes back, and we say where. [For engineers](<#hood-engineers>) ![Painted illustration of the painter stroking his beard as he looks at a light bulb in his hand, a magnifying glass and a crayon on the table.]() ## Research on play, on simulation, on real-world facts and on respect for children each has its own chapter It's tempting to say a children's app is "based on research". For Professor Oops that would be too simple, and partly untrue. Research did four different jobs here, at different times, with different strength. **Play.** The game is a sheet of paper where everything reacts, with no goals and no way to fail. That came from earlier toys, and the research on how children learn was read afterwards. Much of it fits. Some of it argues against what we may claim. **Simulation.** To make wind blow, watercolour dry with a dark rim and drops leave long stains, we read computer graphics and physics papers. A few methods went in by name. Most were cut down to what a tablet can run 60 times a second. **Real facts.** Ants, pigeons, a cat and a dog, and a kitchen full of things that melt, brown and burn. Real findings set which way things go. Play sets how fast. **Respect.** Before the shop was built, we read the law and the store rules on selling to children. That is where research was applied most directly, line by line. We read 179 sources, and say what each one really did For engineers ### We read 179 sources, and say what each one really did 181 reference entries across four chapters, 179 distinct sources \(two appear twice\). 23 applied directly, 23 informed something that was built, 36 considered and not built, 76 found after the design \(15 of those push back or only partly fit\), 23 listed as gaps. "Applied directly" clusters in two places: engineering methods followed by name, such as Stam's stable fluids \(1999\) for the wind and the forensic stain rule W/L = sin α, and 14 laws, regulators' guides, ad codes and store policies on selling to children. No learning paper is applied directly in the released game, and we have run no study of our own. [Under the hood: how a source earns its label](<#hood-labels>) How a source earns its label Under the hood ### How a source earns its label Every entry was looked up again in November 2026: DOIs against Crossref, abstracts from Europe PMC, OpenAlex or the publisher, books by their ISBN record, laws and store rules at their own page. Four simulation papers were read in full. Anything that rested only on a search snippet was dropped. Then each source was traced to the code. The labels mean: - **Applied directly**: a concrete rule in the game traces to it, and the code or a design note says so. - **Informed the design**: read while designing; it shaped an approach, with no one-to-one rule. - **Considered, not built**: read and weighed, then turned down or never built. - **Found later**: read after the thing was built, and consistent with it. - **Pushes back**: found later, and it challenges part of the design or of what we may say. - **Not used yet**: strong work we haven't drawn on, listed so nobody mistakes it for an influence. Each chapter words its labels for its own field; this page maps them onto these six. ![Painted illustration of the painter finger-painting a big sheet covered in hand prints, paint splats, a toothpaste squiggle, glitter and a cracked egg.]() ### [How children learn through play]() Surprise, self-led play, rewards and guided play. 30 sources, most found after the design, 7 that push back. ![Painted illustration of the painter brushing a rainbow watercolour that bleeds into wet paper, with crayons, markers, a paint box and a child's rainbow drawing.]() ### [The science of simulating]() Fluids, watercolour, drop stains, fire and mould. 56 sources, two methods followed by name. ![Painted illustration of the painter watching a line of ants march to a drop of honey through a magnifying glass, with a snail, sprouting cress and a watering can.]() ### [Real facts, real animals]() Ant trails, fleeing flies, ignition points and cabbage colours. 67 sources: 57 traced to the code line by line, 10 strong ones not used yet. ![Painted illustration of a wooden toolbox tied with a red ribbon next to a single coin on a yellow sheet with a ketchup zigzag, the painter reading a book beside it.]() ### [Designed with respect for children]() No locks or prices in play, a sum for grown-ups, three asks at most. 28 sources, 15 applied directly. ## Only 23 sources were applied directly, and we show the rest as they are If you count every source by what it really did, the picture is modest. A small share was applied directly. More shaped an idea that we then built our own, simpler way. A large share we found only after building, and we list it as consistent with the game, or not, never as its basis. Some we read and chose not to build, and some strong work we haven't used yet. The chapters differ a lot. The respect chapter is mostly "applied", because laws and store rules can be followed line by line. The play chapter is mostly "found later", because the toy came before the reading. That difference is the honest shape of how the game was made. What 181 source entries did for the game"Later, fits" and "Pushes back" were both found after building: the first consistent with the game, the second pushing back on it, only partly fitting or contradicting it. Two sources appear in two chapters, so 179 different sources. 23 applied directly, 23 informed something built, 36 considered and not built, 61 found later and fitting, 15 found later and pushing back, 23 not used yet. Data table What 181 source entries did for the game | | sources | | --- | --- | | Applied | 23 | | Informed | 23 | | Not built | 36 | | Later, fits | 61 | | Pushes back | 15 | | Not used yet | 23 | Source: Counted from the reference lists of the four research chapters, November 2026. The same labels, chapter by chapter Respect is mostly applied directly \(15 of 28\); play mostly found later \(27 of 30\); simulation mostly informed or not built \(28 of 56\); real facts spread across all six. - Applied directly - Informed, built - Considered, not built - Found later, fits - Found later, pushes back - Not used yet Data table The same labels, chapter by chapter | | Applied directly | Informed, built | Considered, not built | Found later, fits | Found later, pushes back | Not used yet | | --- | --- | --- | --- | --- | --- | --- | | Play | 1 | 2 | 0 | 20 | 7 | 0 | | Simulation | 2 | 12 | 16 | 13 | 0 | 13 | | Real facts | 5 | 7 | 20 | 19 | 6 | 10 | | Respect | 15 | 2 | 0 | 9 | 2 | 0 | Source: Counted from the reference lists at the foot of each chapter. Simulation "informed" splits into 12 that shaped something built and 16 whose method was weighed and not built, as that chapter says. Real facts "considered, not built" is 13 sources behind plans never built and 7 in our notes with no use traced in the code. [Under the hood: the counts as a table](<#hood-counts>) The counts as a table Under the hood ### The counts as a table | Chapter | Applied | Informed, built | Not built | Later, fits | Later, pushes back | Not used yet | Total | | --- | --- | --- | --- | --- | --- | --- | --- | | Play | 1 | 2 | 0 | 20 | 7 | 0 | 30 | | Simulation | 2 | 12 | 16 | 13 | 0 | 13 | 56 | | Real facts | 5 | 7 | 20 | 19 | 6 | 10 | 67 | | Respect | 15 | 2 | 0 | 9 | 2 | 0 | 28 | | All | 23 | 23 | 36 | 61 | 15 | 23 | 181 | Two sources sit in two chapters, labelled by what they did there. A study of manipulative design in children's apps [Radesky 2022]() is "applied" in the respect chapter, because the shop's rules follow it, and "informed" in the play chapter, because it shaped those rules, not the play. A count of ads in apps for under-fives [Meyer 2019]() is "informed" in both. So 181 entries, 179 sources. The play chapter's one "applied" entry isn't research: it's the review of earlier creative apps and sandbox games \(Kid Pix, Tux Paint, falling-sand games\) that the first design drew its rules from. 179 Different sources across the four chapters, each checked against its own record. 23 Applied directly. 14 of them are laws, regulators' guides, ad codes and store rules on selling to children. 15 Found later that push back or only partly fit. Each chapter names them. 3 of 107 Facts in our own fact sheet rated "measured", its strongest grade. All three are electricity. 0 Studies of our own. We say what the design draws on, never what it proves. ## The game began as a toy, and most of the research on learning was read afterwards Professor Oops started as a toy, not a lesson. Its first design says it in one line: no goals, scores or failure states. That came from looking at what children already love, such as Kid Pix with its dynamite eraser and falling-sand games, where a few simple rules make endless stories [Prior-art 2026](). The learning papers came twelve days later, when we wrote pages for parents. Much of what we then read fits. Even babies learn more about an object that surprised them, and test it in ways that fit what went wrong [Stahl 2015](). Preschoolers given a muddled toy choose to explore it and sort the causes out in free play [Schulz 2007](). Children shown one thing a toy can do mostly stick to that one thing [Bonawitz 2011](). And promised rewards can dull the wish to draw [Lepper 1973](), [Deci 1999](). The game has no points, stars or streaks, a rule set before we read any of this. Some of it pushes back hard. Across 164 studies, explicit teaching beat discovery without help [Alfieri 2011](), and reviews argue that minimal guidance doesn't teach content [Kirschner 2006](). A sandbox on its own is discovery without help. So we never say the game teaches science by itself. The help the research favours comes from a grown-up: the website pairs experiments with a version to do for real in the kitchen. What the game is engineered for is the part a surprise needs: it holds up when you poke it again. Paint a line of water and light the paper beside it, and the fire is held at the line until its heat has boiled the water off, because fire reads how wet each spot is, whatever made it wet. Try again and it's held again. 16 Oct 2026 ### Earlier toys reviewed No rules, no failure, destruction as play, an undo a child can trust, no ads. 17 Oct 2026 ### The first design, and the realism rule "No goals, scores or failure states." The same day: no monsters, giants or black holes, only real things a child could do to paper. Oct 2026 ### Two research notes on simulating Graphics and physics papers read by one rule: believable beats correct. 20 Oct 2026 ### The shop's rules Written after consumer law, regulators' guidance and a study of manipulative design in children's apps. 29 Oct 2026 ### The first learning papers Cited on the website, on a page for parents asking whether children learn anything. Nov 2026 ### Every source checked again Looked up at its own record and traced to the code it shaped, or didn't. ## The simulation takes a few methods by name and cuts the rest down to fit a tablet Our research notes on simulating had one rule: believable beats correct. A dark rim on a dried wash is a signature people recognise. The exact speed of a ripple on three millimetres of water is not, at least not to a child holding a tablet. Two methods went in by name. The wind is "stable fluids" from computer animation [Stam 1999](), changed for a page seen from above, where a fan blowing down is a source of air. And every drop stain follows a forensic rule: its width over its length is the sine of the angle it landed at [Neitzel & Smith](). Far more shaped an idea we then built more simply. The watercolour took its picture of water, fibres and pigment from one 1997 paper [Curtis 1997](), and none of its machinery. And some famous work was turned down on purpose. Mould could grow as random walkers [Witten 1981](), but then "come back in three days" would mean simulating three days. Instead each colony's size is a formula of its age, one sum on any device. One finding was our own. The first wind leaned: draughts blew about a fifth harder one way than the other, because the classic solver updates cells in reading order. The fix was an old method used carefully, multigrid, built so that a mirrored gust gets a mirrored answer to the last bit. [Under the hood](<#hood-simulation>) The simulation takes a few methods by name and cuts the rest down to fit a tablet Under the hood ### The simulation takes a few methods by name and cuts the rest down to fit a tablet **Wind**: semi-Lagrangian advection on 64 × 48 air cells on desktop \(36 × 27 on a phone\), each 8 paper cells across. Only part of a screened pressure correction is applied each step, so a fan reads as a source of air seen from above. It is solved by multigrid, and runs as a WebAssembly twin that matches the JavaScript to the last bit; an automated test checks both. **Drop stains**: the landing angle comes from how fast the drop falls against how fast it travels, and the stain's width over its length is the sine of that angle, as in the forensic rule. Very shallow drops are capped, so a stain never streaks across the page. **Watercolour**: no velocity in the paper water. Neighbouring cells trade water and dye by how much wetter one is than the other, and the paper's fibres make some pairs trade more easily than others. The rim comes out 1.3 to 2.5 times as dark as the wash, measured by an automated test. **Mould**: each colony's radius is a closed formula of its age that slows as it grows, up to a cap, so a skip of any length costs one sum per colony. No test compares the air, water, paint, fire or soap with data from any paper. The one comparison with real arithmetic is electricity: Ohm's law to three figures. ## Real animals and real kitchens set the direction, and play sets the pace Before we build an animal or a kitchen substance, we read. Findings like these decide the mechanism: what the animal senses, and which way it moves. The exact numbers are then set by play, on a sheet about 31 by 23 cm, where a cat is the size of a thumb. Our ants turn toward the side of a trail that smells stronger, by the difference between left and right divided by the total. Researchers measured turns of that same form in Argentine ants [Perna 2012](), but our code doesn't name the paper, so we call it a match, not an implementation. Real garden-ant scent lasts about 47 minutes [Beckers 1993](). Ours fades many times faster, so a child can watch a trail go within one sitting. Every temperature in the game hangs on one real number. Paper catches fire at a point on the game's own scale set to mean 230 °C, inside the published 218 to 246 °C [Autoignition, Wikipedia](). Checking works both ways. Baking soda in water sits at about pH 8.3 [Science Insights](), and the released game let it read pH 10, turning cabbage juice teal-green. In the next update next update it stops at 8.3, and the cabbage goes blue, as it does in a kitchen. Flies filmed at take-off jump directly away from a looming threat [Card 2008](). That paper informed one thing in the game: every startled animal flees straight away from what alarmed it. The other half of the demo below is ours, and no paper behind it: slow movement counts for far less than fast movement. That's what lets a cat creep up on a pigeon. [Under the hood](<#hood-real>) Real animals and real kitchens set the direction, and play sets the pace Under the hood ### Real animals and real kitchens set the direction, and play sets the pace **Ant steering**: three feelers a little ahead of the head. The ant turns by the difference between left and right divided by the total it smells, the same form measured in real ants. Scent fades exponentially, many times faster than in a garden. **Temperature**: one unitless number per cell on a linear scale, from room temperature up, with paper catching fire at the point that means 230 °C. **Movement seen**: anything that moves writes into a shared motion layer. Bigger and faster things show more, slow creeping shows almost nothing, there is a cap, and the signal falls off with distance and fades within a fraction of a second. **Getting used to it**: each animal compares what it senses with what it has got used to, and only the sudden part alarms it. Each species has its own sensitivity, its own startle threshold and its own pace of getting used to things, so a fly bolts sooner than a pigeon. Automated tests check how the animals move and startle. Try it: sneak up on a fly The game's own ruleA hand moves in from the left at the speed you set. Drag on the paper to move it yourself, or press **Start the hand again**. **What to look for:** at a brisk 120 cells a second the fly darts off about 3 cm before the hand arrives, directly away from it: that is the half the research informed. At 45 or less the hand shows no movement at all and stops right beside the fly. Between about 50 and 65 it shows a little, but the fly gets used to it as it grows. A pigeon ignores a hand below about 120 cells a second and lets it come much closer. This is the game's motion rule, falloff, fading and getting-used-to, for one animal on a 360 × 200 cell page \(one cell is about half a millimetre\); the field is read at one point instead of across the game's grid, and the flight is simplified. ![A hand moving toward a fly on a sheet of paper; the fly darts away if the hand moves fast enough]() - noticed here - alarm at the fly - alarm at the pigeon Hand speed120 cells/s AnimalFlyPigeon hand— that is— movement shown— alarm— startles above— noticed the hand— ## Respect for children was built line by line from the rules on selling to them This is where research was applied most directly, and it was read before the shop existed. European law lists a direct appeal to children to buy, or to get an adult to buy for them, among practices that are always unfair [EU 2005](). Researchers who coded 133 apps played by 160 children aged 3 to 5 found about four in five used at least one manipulative design, such as a character asking for money [Radesky 2022](). So a child at play meets no locks, no prices and no pictures of what's for sale. Our first warm idea, a locked present the child could see, was dropped once the research named it: pester power by design. Anything a grown-up must decide sits behind a short sum. The game asks a grown-up about the full version at most three times, a week apart, and "don't ask again" means never [FTC 2022](). There are no ads and no analytics code from another company. Where we chose softer than the strictest reading, the chapter says so: a quiet "\+N" tile for grown-ups at the end of a tab, a sum that won't stop many eight-year-olds. And research found later can still bite. Our pan-fire buttons offer "Put a lid on", as fire guidance says [NFPA 2026](), but also "Damp tea towel", which one UK fire service now warns against [Essex 2024](). We'd rather show that than leave it out. [Under the hood](<#hood-respect>) Respect for children was built line by line from the rules on selling to them Under the hood ### Respect for children was built line by line from the rules on selling to them **Asks**: the game remembers only how often it has asked, when, and whether a grown-up said never. Three asks at most, a week apart, a "never" that is permanent, and no ask at all once the full version is owned. An automated test pins each of these. **The sum**: a two-digit number plus a round ten, a few tries, and a new sum after each wrong answer. **The split** between free and full is plain data, and every tab keeps at least one free tool. The "\+N" tile has no tool pictures, words or price, and a screen reader hears it as the grown-ups' button; a test checks that. **Data**: anonymous daily totals with no identifier and nothing finer than a day, off by default in 32 countries and wherever the country is unknown. Android's advertising-ID permission is stripped, and a test holds the list. ## Every feature can be traced from an idea, through the research, to what you can play Ideas came first. On the day of the first design, several hundred ideas for messy things a child might do to paper were scored for fun, ease, novelty and "you are never allowed to do that" thrill. Then a rule overrode the scores: only real things a child could really do to paper. That cut the giants and black holes, even the ones that scored highest. The table below has the real scores. Research came next, where an idea needed it: how the thing behaves, how to compute it cheaply, and how to offer it fairly. The map after the table follows thirteen ideas from start to finish, each with the research it met and the label that research earned. The real scores: the top 12 ideas under the team's formula, 17 October 2026 | \# | Idea | Fun | Ease | Unique | Extreme | Score | Outcome | | --- | --- | --- | --- | --- | --- | --- | --- | | 1 | Hammer \+ fruit smash | 10 | 8 | 6 | 10 | 8.80 | in the game | | 2 | Giant's footstep | 9 | 8 | 8 | 10 | 8.70 | cut: not real | | 3 | Muddy dog shake-off | 9 | 8 | 8 | 8 | 8.40 | in the game | | 4 | Magnifying-glass sunbeam | 8 | 9 | 7 | 9 | 8.30 | in the game | | 5 | Pigeon: footprints, droppings | 9 | 8 | 8 | 7 | 8.25 | in the game | | 6 | Giant sneeze | 8 | 9 | 7 | 8 | 8.15 | cut: not real | | 7 | Ice-cream scoop stack | 9 | 8 | 7 | 7 | 8.10 | in the game | | 8 | Egg drop, paper fried egg | 9 | 7 | 8 | 8 | 8.10 | in the game | | 9 | Hungry caterpillar | 8 | 8 | 8 | 8 | 8.00 | not in the game | | 10 | Volcano lava ooze | 8 | 8 | 7 | 9 | 8.00 | not in the game | | 11 | Washing-machine tie-dye | 8 | 7 | 9 | 9 | 8.00 | not in the game | | 12 | Black hole \+ white hole | 9 | 5 | 9 | 10 | 7.95 | cut: not real | Each idea was given four scores from 1 to 10 on 17 October 2026; the team's formula weights fun 0.4, ease 0.3, novelty and thrill 0.15 each. It puts a giant's footstep second and a black hole twelfth. With the realism rule they're gone, and Sticky little sibling, Mould garden, Ketchup & mustard move up into the twelve. Several ideas near the top are not in the game, and plenty further down are. ### Play [How children learn through play]() ### A sheet where a child can't fail *Applied directly.* Earlier creative apps and sandbox games: product lessons, not research [Prior-art 2026]() **Built:** No goals, scores or failure states, and an undo a child can trust, kept within a fixed memory budget. [How we keep it right]() ### No stars, points or streaks *Found later.* Promised rewards dull the wish to play [Lepper 1973](), [Deci 1999](). *Pushes back:* the harm is narrower [Cameron 2001]() **Built:** No rewards of any kind, a rule from before the reading and stricter than the evidence. We call it a choice. [How children learn through play]() ### Surprises a child can poke again *Found later.* Children learn from surprises and explore muddled causes [Stahl 2015](), [Schulz 2007]() **Built:** Materials share per-cell layers, so a fire is held at a line of water every time until it has boiled it dry, and finds the thin part first. A test pins it. [Built for surprises]() ### Simulation [The science of simulating]() ### A hair dryer that blows flour about *Applied directly.* Stable fluids [Stam 1999]() **Built:** The wind's advection and projection, with a softened pressure correction, solved by mirror-exact multigrid. [Air that really moves]() ### Watercolour that dries with a dark rim *Informed the design.* Computer-generated watercolour, read in full [Curtis 1997](), and MoXi's fibre noise [Chu 2005]() **Built:** A neighbour exchange of water and dye gated by fibres; the rim comes out 1.3 to 2.5 times as dark as the wash, measured by a test. [Water in paper]() ### Mould that grows while you're away *Considered, not built.* Diffusion-limited aggregation, too slow for a time skip [Witten 1981]() **Built:** Each colony's radius is a formula of its age, so three days later is one sum per colony. [Dough, mould and time]() ### Vinegar fizzes on baking soda, lemon turns cabbage pink *Informed the design.* A talk on sending interactions through shared elements [Dohta 2017]() **Built:** Dissolved elements and a data-driven reaction table, so rules grow as elements × materials, not materials × materials. [Chemistry and pH]() ### Real facts [Real facts, real animals]() ### Ants that follow a trail *In our notes, same form.* Turning by difference over sum [Perna 2012](). *Informed the design:* trail scent lasts about 47 minutes [Beckers 1993]() **Built:** Three feelers that turn the ant by difference over total; scent fades many times faster than real, on purpose. A test pins it. [Animals with their own minds]() ### Paper that catches fire *Found later.* Paper catches by itself at 218 to 246 °C [Autoignition, Wikipedia]() **Built:** Paper catching fire at 230 °C anchors the game's whole temperature scale. [Heat and fire]() ### Cabbage juice on baking soda *Found later, didn't fit.* Baking soda in water is about pH 8.3 [Science Insights]() **Built:** Released: it read pH 10, teal-green. Fixed to stop at 8.3, so cabbage goes blue next update. [Chemistry and pH]() ### Respect [Designed with respect for children]() ### Selling the full version *Applied directly.* No appeal to children to buy or to nag [EU 2005](); "ask your parents" coded as manipulative [Radesky 2022]() **Built:** No locks, prices or pictures of paid tools in play; the locked present was dropped; one purchase behind a sum. [Designed with respect for children]() ### Reminding a grown-up *Applied directly.* A request with no permanent no is nagging [FTC 2022](); a fine for countdown pressure [ACM 2024]() **Built:** Three asks in all, a week apart, each with "No thanks, don't ask again"; no timer and no "only today". [Designed with respect for children]() ### A pan fire on the page *Found later, fits in part.* Lid on, heat off, never water [NFPA 2026](). *Cuts against:* no damp towel [Essex 2024]() **Built:** "Put a lid on" and no water button; water flares the oil and the painter says what a grown-up does instead; a test pins it. The towel button is in question. [Heat and fire]() ## One fact, such as how wet paper resists fire, shows up wherever the game reads the paper's wetness None of these was written as a pairing of two things. A finding goes into one shared layer, such as how wet the paper is or what's lying on it, and everything that reads that layer is affected. That's why a fact checked in one chapter shows up in another. Shared: wet \+ temp ### Paint a thick line of water, then light the paper beside it The fire is held at the line, slips through where it's thin, and in the end boils the rest dry. [Heat and fire]() · pinned by a test Shared: wetness \+ dye per cell ### Paint a watercolour wash and let it dry The edge dries 1.3 to 2.5 times darker than the middle. [Water in paper]() · pinned by a test Shared: fine dust on the cell ### Sift baking soda across an ant trail No ant crosses, and the scent under the line is gone. [Animals with their own minds]() · pinned by a test Shared: carbonate in the paper → pH ### Burn a patch, then paint cabbage juice on the ash The juice turns blue in a thin ring round the burn. There's no cabbage code in the fire. [Chemistry and pH]() · pinned by a test Shared: latent heat, 2,260 J/g ### Drip water onto burning oil Steam throws the oil up, and the painter says what a grown-up does instead. [Heat and fire]() · pinned by a test Shared: noise \+ motion ### Bang the hammer near two pigeons The near one flies, and its wings send up the other. [Sound and shaking]() · pinned by a test ![The game: a line of sifted baking soda across an ant trail, with ants stopped on both sides of it and none on it.]() **A fact meets a layer.** Fine powder lying on the paper wipes the ants' scent, whoever put it there. No source says baking soda keeps real ants away, and we don't claim it. ![The game: a dried blue watercolour patch whose edge is darker than its middle.]() **An idea from a paper, built our own way.** The dark rim of a dried wash, from a 1997 paper's picture of how watercolour works, computed with none of its machinery. What we don't claim - **We don't say the game teaches.** No study, no control group, no learning measure. The research says discovery without help teaches content weakly. - **We don't say the simulation is accurate.** No test compares it with data from any paper here. Tests pin direction, conservation and self-consistency. - **We don't say "based on" for anything found later.** A source we found after building is listed as consistent with the game, or not, never as its basis. - **We don't say the game is fire-safety advice.** The safety lines say what happens and what a grown-up does. It's a game, not guidance. ## The game takes each study's direction and claims nothing it hasn't tested Using research well means knowing where to stop. Each row starts from what a source says or does and ends at what a tablet runs for a child, with the cost stated. **In the research or in real life.** Learning studies tested infants and preschoolers in labs, older pupils in classrooms **In the game.** We apply their direction to children of about 4 to 10 on a tablet **Why that's enough.** Nobody has tested a tablet sandbox at these ages, so we claim fit, never effect. **In the research or in real life.** Guided play has a learning goal and an adult who scaffolds [Weisberg 2016]() **In the game.** No goal and no adult built in; kitchen versions and teacher plans live on the website **Why that's enough.** The child leads. Lost: the guidance research says helps most, so we claim no teaching effect. **In the research or in real life.** Positive feedback can raise motivation, d = \+0.33 [Deci 1999]() **In the game.** No rewards, scores, stars or streaks at all **Why that's enough.** Stricter than the evidence, and we say it's a choice. **In the research or in real life.** Stable fluids keeps the flow incompressible [Stam 1999]() **In the game.** A softened, partial pressure correction on a 64 × 48 air grid, simple damping, no viscosity step **Why that's enough.** Seen from above, a fan is a source and a fire a sink. No real speed or pressure is claimed. **In the research or in real life.** Watercolour as a fluid layer with a velocity field and Kubelka-Munk glazing [Curtis 1997]() **In the game.** One linear exchange gated by fibre noise; colour as absorbance per channel **Why that's enough.** Rim, bloom and bleed are there and tested. Lost: currents inside a wash, and light paint covering dark. **In the research or in real life.** Garden-ant trail scent lasts about 47 minutes [Beckers 1993]() **In the game.** Scent fades many times faster than real **Why that's enough.** A trail must clear within one sitting. Lost: real persistence. **In the research or in real life.** Heat is energy, with real heat capacities for paper, water and oil **In the game.** One unitless number per cell on a linear scale, anchored on paper at 230 °C **Why that's enough.** One scale serves everything. Lost: a cell of water warms about ten times too easily. **In the research or in real life.** The strictest reading: no sign of paid tools in the child's space [EU 2005]() **In the game.** A static "\+N" tile with reading glasses at a tab's end, no pictures, words or price **Why that's enough.** A grown-up can see where there's more without hunting. It sits between two rungs of the chapter's ladder, and we say so.