A test you can actually read
Write the hypothesis in a sentence: "A face with the setting on screen will beat the gameplay-only frame." Swap only that. If you also change the title, the length, and the topic, you learned nothing.
On YouTube, use the built-in test when you have it, or swap after a fair number of impressions and compare CTR in the same traffic source. For hooks, publish two similar shorts with different first lines and compare the three-second hold. Write the result down. The point is the next video, not a trophy.