Discarding all the moderate users seems like a design flaw. A possible effect on cognition is a fairly natural question for regression.
Maybe you saw it in the text but, "This threshold was selected as it exceeds the American Academy of Pediatrics screen time guidelines, which recommend that videogaming time be limited to one to two hours per day for older children."
The thought process I can sort of imagine is that the tests are fairly costly and the hypothesis they're testing is that more play than the official recommended limit should decidedly give detectable impairments to cognition. (Seems they could have rejected that, if they'd had comparable groups in the first place. Now I'm not sure what it says)