Well, I think you're kind of misrepresenting his point as "competitive programming doesn't improve your programming abilities".
It's not that competitive programming correlates poorly with job performance; it's that, given you've been hired by Google, being a competitive programmer correlates poorly with job performance.
Hypothetically, let's say there's 2 dimensions for a programmer's ability, competitive programming and job experience. Each of these is distributed independently from 0-10. You get hired by Google if competitive programming + job experience > 10. However, let's say that for their actual job performance, it's 0.5 x competitive programming + 1.5 x job experience.
You now have a case where competitive programming could correlate poorly with job performance at Google, despite having a positive correlation in general.
It's not that competitive programming correlates poorly with job performance; it's that, given you've been hired by Google, being a competitive programmer correlates poorly with job performance.
Hypothetically, let's say there's 2 dimensions for a programmer's ability, competitive programming and job experience. Each of these is distributed independently from 0-10. You get hired by Google if competitive programming + job experience > 10. However, let's say that for their actual job performance, it's 0.5 x competitive programming + 1.5 x job experience.
You now have a case where competitive programming could correlate poorly with job performance at Google, despite having a positive correlation in general.