No "evidence" needed. The actual announcement may not be in December, but there will be something to "get".
By the end of the year, we will all be certain of "something".
I'm not even close to pretending to be one of those whack-job nut-cases that "claims" so have insider info. What I did though was compile data over the last 30 years from developers with regards to both development and announcement. It was a beeyotch to the understatement of the year award x infinity. It has been a year long project of mine because there is so much misinformation on the internet that sorting through actual dates has been crazy... not to mention the absolute nightmare reconciling PC and Console release dates with crappy handheld counterparts. All in all, when I made the post of "get it got it good" I was exasperated. The truth of the matter is that by a factor of 2.376 this December should give indication of what Bethesda is working on. The books throw that off by quite a bit... and I tried several permutations only to get similar results. The marketing of the next novel can reduce that to 0.31 and therefore would dis-allow December from a probable month, but I decided to not include that since the number of novel tie-ins was 0.013% of the overall product release line.
For those who care, I also factored in the recent trademark data but compiled only the trademark data from Halo, Diablo, Fable, Super Mario Bros, Sonic the Hedgehog, Final Fantasy, and Jurassic Park (Sega CD inculded). I needed a broad data pool to make accurate predictions. I used the obscure with a multiplier based upon the overall marketshare of similar type games.
Basically within each subset of data I also looked at fiscal drivers and with regard to the publicly traded release cycles there were ebb and flows that allowed for a statistic and probability map to be drawn. I tooled that mapping along the private release models to gain cohesion and then added in the last part.
The last part has been the staff size of BGS, engine tech, release history, investors, technological perimeters (chips/consoles/DirectX). I created a fluid model of competing drivers that flexed over time and also flexed forward with conservative thresholds.
2.376 may not mean much by itself, but when combined with all the other data, October is sitting currently at 0.000001 and November is at 0.4333 that is a lot. Other than that January and February are also in the "2" range and March is at both a "3" and a "0.x" model since there are reducing factors that will occur at the end of the year.
All in all, for the masses, by what I meant with "get it, got it, good" I meant, "Probably, finally, whew".
EDIT: By the way, numbers crunching is a game in and of itself to me, and after I re-read this post I thought to myself, "Self... you sound like an obsessed lunatic... perhaps you should tell the rest of the folks out there that this "speculation" business is a game just like the game itself is..." I am just a geek with excel folks... and yeah I like math.