This would be perfect in theory, but in the real world it hardly works because:
- the most demanding games are often not very popular anymore or have no multi-player component at all which means once you beat them, you probably don't revisit them. For example, Crysis 1/Warhead, Metro 2033, Witcher 2 EE);
- the most popular games are often the least demanding (since developers want them to run on as many systems as possible). For example Diablo 3, WOW, SC2, Left 4 Dead 2, Warcraft 3, TF2, Minecraft, Sims
- At the moment, only BF3 falls into both Popular+GPU demanding+still good enough to test next generation GPUs. So are we going to have a review of just 1 game?
To be able to test future generation of GPUs you have to look at different things: performance with high resolutions/texture mods, or include modern games that use of some new graphical feature (i.e., tessellation or global illumination). Ironically, Sleeping Dogs, Sniper Elite V2 and Dirt Showdown are 3 modern games that use global illumination and why are actually a foreshadowing of a possible new graphical lighting model via direct compute shaders (or they may not be depending on how the industry evolves).
If anything, it makes a lot more sense to test Sleeping Dogs than BL2 because BL2 runs like butter on a $230-300 GTX660/660Ti but Sleeping Dogs chokes on a $500 videocard. The other aspect of testing Sleeping Dogs which is appealing is that it allows reviewers to test very heavy AA performance, which stresses the GPU's ROP/memory bandwidth setup. Sleeping Dogs is thus used to test both the use of a new lighting model technique and AA, while BL2 is good for testing CPU limitations/IPC scaling and PhysX High CPU load.