Hochevar was coming off two starts in which the Royals won easily, scoring 16 and 13 runs, but he didn’t last the necessary five innings to secure the victory. The Royals urged Hochevar to vary his speeds more and make his pitch selection more unpredictable. It seemed to work.

I think Dutton released a video a couple days later talking about adjustments for the RHP’s.

That used 30 inning samples tho
A 10 inning sample is even less meaningful than a 30 inning sample

No, you used the average from his previous 4 starts or “an average sample size of just over 21 innings.” Here is what you wrote:

Plotting each start on its own will show mostly random noise, so I chose to plot a moving average with each data point reflecting Duffy’s previous 4 starts up to that point. (The first data point shows Duffy’s BB/9 over his first 4 starts, the second data point shows starts 2 through 5, etc) This gives an average sample size of just over 21 innings per data point which isn’t a rock-solid indicator of future performance, but should be enough to show real changes in performance over time

Since then, Duffy has started 4 games and pitched 20.2 innings. His most recent 4 game average is 4.8 BB/9. That “should be enough to show real changes in performances over time,” right?

Kidding aside, 4 or 5 game samples sizes are just not going to provide much meaning. Duffy has thrown 72 innings in the majors, and his BB/9 is 4.46. Based on his minor league track record, it is reasonable to expect that number to come down, at least somewhat, but using 4 or 5 game samples to suggest a permanent change against a larger sample size is silly.