How Can you argue it either way? Is the point not, that you cant do that?
A complete answer to this would get us pretty far pretty quickly, so I'll try to give a general outline.
Classical statistical models only measure association, not causation. However, a statistical model is only one tool. In practice, some knowledge of the problem and the underlying (in this case economical) processes could help in arguing that a causal link that goes one way or the other, based on common knowledge and general proofs of concept.
However, this is where most researchers stop, while this step can introduce strong researcher's bias towards a conclusion one wants to obtain. Correct statistical analysis at this point would call for adapted causal inference models which do exist (the simplest being a structural equation model) and that do a decent job under some assumptions.
There is still a risk of misspecification of these models as well and even if events happen in a certain order, there is never a (statistical) guarantee of causality (e.g.: in economics a later event could be a lagging indicator of an earlier event, although one might assume that an earlier event is the leading indicator of a later event). My personal rule of thumb is that the process needs to be known and well-documented (or in exploratory research: objective findings that point towards causality beyond the scientist's imagination) before you can claim it. And, even in that case, the total effect could be a combination of a known causal process and an unknown confounder.
So yeah, that prejudice about statisticians never being certain: we actually try to tell other researchers that they shouldn't be too certain of themselves. ;-)
Disclaimer: I'm personally not a specialist in causal inference, but half of my research group is doing research in causal inference so I do catch a few things once in a while. ;-)
For this specific example, I'm following Gus. I'm not familiar enough with the material, but viable arguments for both causal processes (given in my previous post) exist and could be theoretically tested. I have the feeling that the researchers themselves tried to circumvent the tricky causality question by using "gaming innovations" rather than gaming itself. But it would be dangerous to claim that changes in gaming innovations can certainly not be (indirectly) caused by changes in market hours.
Of course, the best and only (almost) certain way to solve the causal problem would be using an experimental setup including randomization (and a few other lesser requirements). This is often not possible, not practical or extremely expensive and time-consuming. It would be fun, but kinda unethical in this particular example I think. ;-)