Press to dig. Hold deepens; drag ploughs.
Full motion uses phone movement directly. Enable tilt and calibrate first. Touch and keyboard provide tilt only. You can switch back with one tap during play.
Phone motion is not enabled.
Saddle and Basin's route is tested at 8°. Lower budgets may make it impassable.
The open board starts flat, with only the border, a start and a goal ring. Sculpt adds everything else. Each board keeps its own edits on this device.
Solid steel · long coast, firm rebounds. Diameter 10 mm, mass 4.1 g.
Changes the ball, not the board: holes, walls and the goal ring keep their size, so a small ball drops into holes and hollows more easily and a large one rolls over small bumps. Every size stays narrower than the holes. Solid balls keep their density; the table-tennis shell keeps its wall thickness. Level acceptance uses the nominal size. Changing the size starts a new run.
Smooth sealed hardwood. Counter-tilt to brake.
Rigid wooden walls; rebound depends on the ball.
30 × 63 cm boardLevel acceptance uses hardwood and 8° tilt. Other floors are exploratory and may need stronger motion. The maze grows with larger balls to keep passages and holes in proportion. Hole rims remain hardwood. Changing a material starts a new run.
Follow the rises and dips. Hold the whole ball inside the green ring for three seconds to finish.
Sculpt ends the current run and lets you reshape the ground; Done starts a new run from the start. Pressing digs a hollow and pushes up a low rim around it, so a press moves earth without adding or removing it. On the open board you can reshape everything. On Saddle and Basin the authored start shelf, goal summit and hole rims stay as they are, and edits fade out next to them. With Move you can drag any object: walls turn and stretch, and holes, patches and the goal ring resize, by a round handle. ＋ Add places new walls, holes, sand or resin; ✕ removes the selected one. The border stays. Walls, holes and patches keep clear of the start, holes stay wider than every ball, and a way from the start to the goal must remain for every ball size (low walls count as hoppable in full motion). Slopes stay below 14.6°. Edits are kept on this device; an edited board has not been checked by the route bot.
Remove the outer walls: rolling off the board drops the ball and restarts the run. Inner walls stay in place. Changing this option starts a new run.
Sound starts after a tap. Device silent-mode behaviour is respected.
Haptics are available only on supported devices.
iPhone: open in Safari, tap Share → Add to Home Screen, then open Marble Lab from its icon. Keep “Open as Web App” enabled if offered.
Screen axes are remapped automatically.
For phone tilt, open this page directly in Safari or Chrome. Use “Enable tilt”, hold a comfortable viewing angle, then calibrate. If access was declined, check the browser’s motion settings; Safari may require clearing this site’s data before asking again.
Event age measures browser delivery to a rendered frame, not physical sensor-to-photon latency. Timestamps cannot expose the sensor’s internal delay. *Orientation events may be change-driven: their delivered rate is not the hardware sampling rate. Hold at a comfortable pitch, calibrate, then make a quick tilt step to inspect the response.
Clear the recording, perform the selected motion, then export it. A stationary tilt trial checks gravity leakage; acceleration peaks do not establish sensor range.
Level acceptance record · Draft model notes
For the largest view on iPhone, use Share → Add to Home Screen. Sound and haptics can be changed in Materials. Grooves and skids are visual traces; they do not change friction.
Translation, spin and height are simulated together. Contact friction decides whether the ball grips or slides. Hills and dips change the local slope and support load. Crests can launch the ball; the goal dome requires a three-second hold. Relief shading exaggerates height; geometry and light do not follow tilt. Wall and rim impacts can produce hops; free fall and rim contact decide whether a hole captures the ball.
Material profiles are fixed. Ice grip, wood losses and ball bounce use provisional contact estimates. Billiards coefficients are literature reference values, not a calibrated replica. Air drag acts on every ball. Sand uses a load- and density-dependent sinkage and resistance approximation, not individual grains. The model is literature-informed and awaits physical calibration.
Read the equations, parameter values and sources
Test a pitched calibration, rotation, a quick tilt step and app switching on iPhone Safari and Android Chrome. Browser timestamps do not measure full physical motion-to-pixel latency.