Streaks with zero skill. Every shot below comes from the same fixed make-probability β nothing about the shooter's skill ever changes. The only thing changing is how correlated the noise is with its own past.
100-shot illustrative game log, seeded deterministically to resemble publicly-reported shooting patterns (not a live API pull). Meant only as an eyeball comparison against the simulator's low-Ο vs high-Ο regimes β drag the slider until the grid above "feels" like this one.
Purple = 400 simulated games at the current memory setting (Monte Carlo). Gray outline = the exact theoretical distribution for truly independent shots at the same 45% make rate (computed via exact dynamic programming, n=100). At Ο=0 the two overlap almost perfectly. Crank Ο up and the purple mass drags right β longer streaks become common with no change in shooting skill.