9 rolls isn't nearly enough to determine if the game's rng is weighted. Humans are bad at detecting randomness. Especially because we expect hits and are frustrated by misses, so we're so much more likely to remember multiple misses. As @Droata says, you are one of ~a million players playing BG3; having 9 failures in a row on this check statistically has to happen to one of them.
With a d20, you need to record at least 300 (preferably 500+) rolls. Only then can you perform a statistical test to see if your rolls are significantly different from expectation. Larian's non-loaded rng is known to streaky, but it streaks both with low- and high-valued rolls and over the course of 300+ rolls it approaches something approximating true randomness.
