I think any statements predicting what specific compression ratios we'll see is foolish. That goes for you, me, and
chris 1515 . Because in almost all cases, we're talking about lossy compression. So predicting ratios isn't purely a technical discussion, it's also inherently a prediction about what amount of quality loss art directors and tech leads will live with. For example, I feel like a stickler for render quality, but I look at the examples RAD is providing and even at lambda=40 I'm like, "Well, that's not
too awful." Whereas some employed artist with real expertise could be thinking "What have they done with my texels?!"
So projecting lossy compression makes the mix of confounding factors include not just varied game styles, not just teams with different skill levels, but specific individual psychologies. If anyone thinks they can Nostradamus through that haze they're goofy. It's like arguing over conflicting estimates of what bitrate IGN will use in their next Youtube video.