WebTug.Spider science · Physics · ArcadeChoose a game

GENERATOR TEST REPORT

The forest is random, but it is not unconstrained.

WebTug chooses changing gap centers, then limits vertical jumps that would contradict the movement model. This report explains the rule and publishes one seeded stress test instead of asking players to trust a vague fairness claim.

WebTug Classic guide. The controls, scores, and shared character rules on this page describe Classic. The six science arcade games have their own controls and simplified physics.

Published August 10, 2026Updated October 20, 20265 min readBy Rafael, WebTug creator

WHAT THE GENERATOR CONTROLS

Each new tree opening begins with a candidate vertical center. The generator keeps the full opening inside the arena, compares it with the previous center, and clamps the change to a reachable range that scales with arena height. On portrait screens it also blocks a second severe move in the same vertical direction immediately after a severe move.

The rule does not flatten every sequence. Reversals and large single changes remain, because reading those changes is the central skill. It removes a narrow class of hostile sequences without converting the forest into a repeating lane.

TEST SETUP

The October 20 reproduction reran the same reachability function used by the game with deterministic seed 20260726. The simulated arena was 700 pixels high in portrait orientation, with the current gap and world-scale values. The script generated 20,000 consecutive transitions and counted moves larger than 65% of the allowed maximum. Separately, after a move above 78% of the maximum, it verified that a same-direction continuation did not exceed 55% of that maximum.

Sample20,000 consecutive transitions

Arena700 px, portrait

Seed20260726

PRNG32-bit linear congruential generator

Allowed center range138.971–542.029 px

Maximum step146.876 px

REPRODUCTION FORMULAstateₙ₊₁ = (1664525 × stateₙ + 1013904223) mod 2³²

Divide each unsigned 32-bit state by 2³² to obtain the next value in [0, 1). Each transition consumes two values. The first previous center is 336 px; the first transition has no previous-previous center. The downloadable generator is an exact copy of the production chooseReachableGapY source, identified by its SHA-256 fingerprint in the results.

RESULTS

20,000transitions tested
7,335large moves over 65%
36.675%large-move share
0forbidden continuations

The largest measured change was 146.876 pixels, matching the configured cap within the 0.000000001 px comparison tolerance. More than a third of transitions still counted as large under the test threshold, so the constraint is not silently reducing every run to small steps. The protected continuation rule applied 2,029 times. There were 0 out-of-bounds centers and 0 steps above the cap. The zero forbidden-continuation count applies only to the specific pattern defined above.

DOWNLOAD AND REPRODUCE THE EVIDENCE

These files contain the actual inputs and all generated rows, rather than just the rounded numbers in this article. They are a versioned snapshot; the results and source hashes stay together.

RUN IT LOCALLY

Save the script and generator source beside each other. With Node.js 24 or newer, run:

node reproduce.mjs --output ./results

This writes results/summary.json and results/transitions.csv. If you also save the published JSON and CSV beside the script, this command verifies both byte for byte and exits with an error on a mismatch:

node reproduce.mjs --check

The October 20 rerun reproduced the previously reported 7,335 large moves and zero forbidden continuations. Publishing these files adds an independently checkable record; it does not change the game generator or its difficulty.

WHAT THIS TEST DOES NOT PROVE

A generator constraint cannot guarantee that every player can recover from every prior mistake. A technically reachable opening may still be difficult if the spider enters the previous opening at a bad height or at maximum downward speed. This test does not simulate those player states, measure perceived difficulty, reaction time, browser frame loss, or every possible screen shape.

This release adds automated evidence, not a human playtest or a measured learning study. A suspicious sequence should still be reported with screen orientation and, if possible, a recording so the player's entry height and velocity can be examined separately.

The exact movement values behind “reachable” are documented in the physics guide. Device-specific space changes are explained in the mobile guide.