is there even a demand for it?
One thing I could think of is the demand to create a web NFT-rating based on machine learning may be plausible, not specifically about the games but in a broader sense, some people may want to know whether some specific NFT is a copycat/legitimate or not. You can also build a system that detects whether some NFTs have been washed-trade or not. Though, neither I know much about machine learning, so I don't whether those things I mentioned are achievable or maybe overkill if you are using machine learning for it.
~Is this a worthwhile research project?
On the other hand, I discovered a previous study[2] that utilizes the similar process but instead of NFT, they identify scam ICOs, which provides me an idea for this study's topic. With the NFT games at peak, will this topic be helpful in many ways?
Looking at the referenced study, it has been cited over
31 times, so if your group are passionate enough, it may help other researchers specifically about NFT instead of ICO.