Take a look at the user review statistics for "Prince of Persia: Sands of Time"
- Playstation 2: 2,064 ratings for an average of 9.0
- Xbox: 986 ratings for an average of 8.9
- Gamecube: 619 ratings for an average of 9.0
- PC: 1,162 ratings for an average of 8.8
- Game Boy Advance 15,102 ratings for an average of 8.9
- Playstation 3: 13 ratings for an average of 8.5
Obviously, these numbers are inaccurate - there's no way the GBA version of this game received more than three times as many user ratings than all the other versions combined. Somehow, in the site relaunch, Gamespot lost the version-specific ratings data, at least when it comes to ratings that don't have written reviews. Now, they apply the ratings by guesswork. So the GBA version might have had an average of 8.2, but a bunch of 9.1 and 9.0 ratings from the other versions got reapplied to the GBA version and brought the average up (and possibly brought the average from the other versions down).
Check multi-platform game at all, and you will see the same obvious evidence of inaccurate data.
My suggestion is that Gamespot simply acknowledge that the ratings-only data for multi-platform games is inaccurate, and that it's useless to present it if it's not correct. Assuming there's no way to recover the accurate data, I think they should delete all the inaccurate numbers, leaving only (a) information for games that are exclusive to one platform, and (b) ratings from written user reviews. Then Gamespot can have a call-to-arms for their users to repopulate the ratings with accurate information.