A 3D memory game built on real coordinates asks you to study one marked object inside a literal X, Y, and Z scene, then locate that same world point after the camera yaws and pitches around it. Unlike a flat board where shapes slide or cards flip, this format treats depth as a first-class citizen, so two objects at the same screen spot can sit at different depths and project to different pixels once the view rotates. Each round begins with a covered scene, opens with an explicit Reveal phase that exposes three objects along the X, Y, and Z rails, then closes with a Hide phase that removes every object before a yaw and a pitch rotation are applied. The four candidate cards that follow each show the rotated coordinate frame with a crosshair at a different projected screen position, and your job is to pick the card whose crosshair lands where the remembered world point actually ends up. Five authored rounds, exactly one match per round, and a fixed 1,000-point ceiling make the structure completely deterministic, so progress is governed by keeping a world point in mind, selecting the matching card, and avoiding two distinct incorrect options.
What Sets This 3D Memory Game Apart From 2D Memory Puzzles
Most memory games published online lean on flat recall. A Number Memory Test asks you to type back a digit string. A Visual Memory Test asks you to recreate highlighted cells on a fixed two-dimensional grid. A Position Memory Game asks you to return named objects to their original cells. Those formats share one assumption: the answer is the same point on the same flat surface. A 3D memory game rejects that assumption, because the camera itself moves.
In Spatial Memory Game, the marked object's screen position before the rotation is never the answer. You remember a world point defined by horizontal, vertical, and depth coordinates, and then you figure out where that point projects after the view changes. That difference is not cosmetic. A point that sat on the left side of the original scene can move to the right side of the recall view, slide upward, or sink lower, depending on how the yaw and pitch combine with its depth. Two objects at the same X and Y but different Z values routinely land at different screen positions after rotation, because the perspective denominator shrinks as the rotated Z approaches the camera.
| Memory Game | What You Remember | What You Must Reproduce |
|---|---|---|
| Spatial Memory Game (this 3D memory game) | One object's X, Y, and Z world position | Its projected screen position after a yaw and a pitch rotation |
| Number Memory Test | A digit string | The same digit string in order |
| Visual Memory Test | Marked cells on a 2D grid | The same highlighted cells on the flat grid |
| Position Memory Game | Named objects on a board | The original cells for each named object |
Side by side, the four formats share the word memory but solve entirely different cognitive problems. The 3D variant asks you to keep a world point in mind while the camera literally turns around it.
How to Play the 3D Memory Game
The round flow is small, explicit, and the same across all five rounds.
- Press Reveal scene and study the dashed marker around the named object. Note where it sits along the X rail (left and right), the Y rail (height), and the Z rail (depth). You can take as long as you want.
- Press Hide only after the position is clear. The three objects vanish, the scene becomes covered, and the round's published yaw and pitch are applied to the camera. The view now looks rotated.
- Inspect the four option cards. Each card shows the same rotated coordinate frame, but the crosshair sits at a different projected screen position.
- Pick the card where the remembered world point lands after the perspective projection of the new view.
- Repeat for five rounds. Every correct projection awards exactly 200 points. Five correct rounds produce the fixed maximum of 1,000 points.
The interface refuses to shortcut that order. A covered scene does not accept an answer, and Hide cannot run before Reveal. After a correct non-final answer, focus returns to Reveal, ready for the next scene. Restart always returns to the first covered scene with zero score and no stored failures.
The Geometry Inside Each Round
The recall view is not a guess. Spatial Memory Game stores three literal world points per scene, applies a yaw matrix around the vertical axis, applies a pitch matrix around the horizontal axis, and then runs a perspective step that uses the rotated Z coordinate as depth. The convention is right-handed and active, so the object itself rotates with the camera change rather than the camera spinning around a fixed point.
The matrix form traces to Carnegie Mellon lecture notes on 3D axis rotations. Microsoft Learn documents the projection transform that scales the rotated X and Y by a depth denominator, and the MDN reference for the perspective CSS function independently cross-checks positive-versus-negative depth behavior, which matters because a point closer to the camera should appear larger than a point at the same X and Y but further back.
The exact projection formulas are screenX = 160 + rotatedX × 240 ÷ (320 − rotatedZ) and screenY = 110 − rotatedY × 240 ÷ (320 − rotatedZ), with the focal length 240, camera distance 320, and a 320 by 220 viewport serving as authored display parameters. To see how they behave, take a point at world coordinates (40, 20, −40) and apply a 90° yaw with no pitch. The yaw rotates the point around the vertical axis, sending X from 40 to −40, holding Y at 20, and keeping Z at −40. With rotatedZ = −40, the denominator becomes 320 − (−40) = 360, so screenX = 160 + (−40 × 240) ÷ 360 = 160 − 26.67 = 133.33, and screenY = 110 − (20 × 240) ÷ 360 = 110 − 13.33 = 96.67. The point that originally sat on the right side of the scene now lands on the left side of the rotated view, above the center of the frame.
Two consequences follow. First, every rotation preserves the distance from the origin, so the recalled point's geometry never drifts. Second, because Z enters the denominator, points closer to the camera spread apart on screen while points further from the camera compress. The four candidate cards per round are picked from authored screen positions, and exactly one of them sits within 0.001 screen unit of the projected target. Production stores no correct index and exposes no semantic answer label before selection, so the matching card is the one whose crosshair sits within 0.001 screen unit of the projected target.
Scoring, Errors, and the 1,000-Point Route
Each correct projection adds exactly 200 points to your score. Five correct rounds produce the fixed maximum of 1,000 points. If you pick a wrong option, the run is not over on the first miss. One distinct incorrect option in a round can be repaired in the same round, because the failure signature combines the round number and the display position rather than counting every wrong click. Repeating that unchanged wrong option does not add another error. A second distinct wrong option closes the run and freezes every control.
Once a run is deadlocked, Reveal, Hide, cursor movement, and answer submission all stop responding. A readable terminal status takes focus so a keyboard user can read what happened, and Restart returns the game to the first covered scene with zero score and no stored failures. If your browser allows local storage, the shared game shell remembers only the best completed score on that device. Blocking storage removes that convenience without changing any geometry or any answer.
Controls, Privacy, and What the Tool Is Not
Pointer and touch users can press Reveal, Hide, or any option card directly. Keyboard users Tab into the option group and use the Arrow keys across its two-by-two layout, pressing Enter or Space to select. Roving focus keeps the actual focused button, the visible focus ring, and the logical cursor aligned, so the next move is always predictable. Initial page load never steals focus, which matters for screen reader users who arrive on the page mid-task. The marked reveal object exposes its exact X, Y, and Z world coordinates, and each visible candidate exposes its exact screen x out of 320 and screen y out of 220. Those descriptions give a precise semantic equivalent without marking correctness.
Spatial Memory Game is entertainment, not a measurement. The five fixed authored scenes are not normed, calibrated, adaptive, or comparable between people. Display size, input method, familiarity with perspective drawings, and ordinary distraction can affect a single play session without supporting any conclusion about memory, intelligence, spatial ability, attention, school readiness, or occupational fitness.
Everything runs locally. The game is built with React and SVG, never uses WebGL, never inspects device orientation, never accesses a camera, never reads motion hardware, never calls an API, and never uploads coordinates. There is no account to create, no package to install, and no random scene generator. The geometry is deterministic.
For a deeper look, see Brick Builder Studio Rules: 5 Milestones to 1,000 Points.