The data set is paywalled so it's hard to know. If they picked shovelware most people would rather pirate then yeah, they could reach that conclusion easily.
Denuvo could also be just making people forget about the game once the hype dies down so they never end up trying it which ends up never buying it.
Some people also end up buying the game in sale later, or well after they played it. I personally ended up buying a lot of the games I pirating a while back, well after their release.
If I am unable or unwilling to purchase a game (lack of funds, ethical concerns, etc) and I cannot pirate it or get it heavily discounted then I do something very simple and efficient.
I just don't play it. There are heaps of great games out there, new and old, that I can put my time into instead.
Many of the games that pack Denuvo these days are AAA trash that I wouldn't play anyway.
I can't get in to see the methodology, but I just fundamentally have a hard time buying that there's going to be an approach that actually addresses the confounding variables in any way. The sample size is way too small for how massive the variance is between whatever you consider "extremely similar" games.
My opinion is that with or without denuvo a good game or product will always sell itself.
Take Wukong black myth for example. Even without denuvo i believe it will still become a huge commerical success. Perhaps it would even have sold more units because there are people who simply refuse to buy any game no matter how good it is as long as it has denuvo.
Piracy definitely has an impact on mediocre game that is lacking in every aspect. But then again the average person who pirate is probably broke and is probably never going buying it anyway so 🤷♂️