Exercise 3: What Goes Wrong Beyond "the Angle Is Slightly Off" — Possible Solution ==================================================================== WHAT THE VERIFIED DRIFT ACTUALLY MEANS ------------------------------ This chapter verified that after 200,000 repeated matrix multiplications composing a small rotation, the resulting matrix's determinant drifted to 1.0000000000188298, rather than the exact value of 1 that every true rotation matrix must have. A determinant that isn't exactly 1 means the matrix, strictly speaking, is no longer a PURE rotation at all - it has picked up a small amount of SCALING mixed in with the rotation, since a matrix with determinant greater than 1 stretches areas (and therefore lengths) slightly larger with every application, while one with determinant less than 1 would shrink them. WHAT WOULD START TO LOOK WRONG ------------------------------ For an object that spins continuously over a very long play session, using this same matrix-multiplication-every-frame approach, the accumulated drift would cause the object to gradually and very subtly change SIZE over time, not just have a slightly wrong facing angle. Beyond just the size changing, the matrix's own rows can also drift away from being perfectly perpendicular to each other (losing strict orthogonality, not just determinant-1 scaling), which for a visual object can manifest as a subtle, gradually increasing SHEARING distortion - the object's shape appearing to skew rather than staying rigid as it rotates. Over a long enough session, an object that was originally, say, a perfect square could visibly stretch into a slightly distorted parallelogram shape purely from this accumulated numerical drift, with no change to the game's actual logic or intended behavior. WHY THIS IS WORSE THAN JUST AN ANGLE ERROR ------------------------------ A simple wrong-angle bug (the object facing a slightly incorrect direction) is a rotation error and nothing else - the object's shape and size stay perfectly intact. This drift is a genuinely different and more insidious category of bug, because it corrupts the object's own geometry over time rather than just its orientation, and the corruption compounds continuously the longer the session runs - unlike an angle error, which doesn't get progressively worse on its own once it's introduced. WHY THIS WORKS AS AN ANSWER ------------------------------ The explanation connects the verified determinant-drift number directly to its geometric meaning (unwanted scaling and shearing, not just an angle error), describes a concrete, observable visual symptom (an object's shape distorting over a long session), and explains why this failure mode is qualitatively different from - and worse than - simply having a slightly wrong rotation angle.