My quest to see all of Tetris | Antithesis
Summary
Antithesis uses a graph-aware fuzzing strategy to discover user inputs that reach Tetris level 255, triggering a 'rebirth' where the level counter rolls back to zero.
Antithesis explores via random trials.
The system runs small trials of random inputs at known states and remembers interesting game situations relative to a specific goal.
run smallish trials of random inputs at known starting game states
Graph-aware strategies find high-reward paths.
Josh Reagan and Dave Scherer treated game state buckets as nodes in a graph, using edges to carry rewards for traversing between them.
each bucket becomes a node, each edge carries the reward
Mathematical errors stall automated progress.
The team's progress hit a wall at level 160 because the objective function contained a bug regarding the high-order byte format.
The high order byte was not BCD.
Rebirth requires reaching level 255.
Advancing past level 255 rolls the counter to 0, resetting the speed and color scheme to level 0 parameters.
running through all the levels of Tetris until the level counter rolled over from 255 back to 0.
Fuzzing may unintentionally bias RNG.
The survival-of-the-fittest strategy might select inputs that happen to nudge the random number generator toward more favorable piece sequences.
Read the original ↗Save to your Sponge
Summarized with Sponge. Quotes are the author’s own words.