My point is, saying definitively that no streaming technological advances have been made, they are all exactly the same and we have apparently reached the limit in streaming technology, based on "the game had XXX total latency" doesn't make sense when a large part of that latency is from the game itself, unless you are comparing the exact same game on the different platforms.
Well, I don't think I said quite all those things - especially about reaching the limit of streaming tech. I don't think we have - or if we have, i'm sure it's only for the moment. If I said 'no advances' or 'exactly the same' - let me qualify: no
big leaps. For whatever contributors to latency make up the end experience, and for their variances across services, in the end they had a similarly laggy, or lag-free (depending on your point of view) experience with cloud gaming 2 years ago as they've measured this year with Stadia and xCloud. Back then they're absolute best experience was 118ms, this year it was 121ms. Back then they played a fighting game off the cloud with more or less the same overall responsiveness as a fighting game they tested this year off the cloud. I know these are different games with different pipelines, but I think it's noteworthy enough, given some of the hype previously, that the overall bar wasn't pushed up so much - at least for DF, here.
That doesn't mean it's the same for everyone, it doesn't mean that there haven't been
some advances in individual components or that some things are to some degree or another 'better' now, or that these services aren't better than others on one front or another. Especially when it comes to - e.g. - game builds that go for optimised input latency from the beginning, or higher framerate, or both, in the case of something like Stadia, where specific builds are possible. But for some of the hype I saw previously (e.g. single-digit additional latency), the evidence in this DF testing at least isn't there for that idea of a breakthrough leap in latency this year. Whether it's 40 or 50 or 60ms of additional latency, these are all quite a bit more in the same ballpark than the e.g. 4ms being touted from some controlled previews.