1.) This system is designed to work over a very long stretch of time. The TGSAP is not an immediate gratification machine. It is designed to endure years and thousands of adjudications without recalibration. 100 adjudications is nothing and can give no real measurable statistic. Its been only around 15 days, and already the system has moved through 130 adjudications generating more than 2500 posts with an ever increasing rate of submission. Someone starting off as a new adjudicating user, is not going to be an expert adjudicator only after 15 days and a hundred or so votes... After a year of submissions and adjudications passes, only then will people begin to fully realize how bad a move it was for them to just randomly guess. Their votes of course by this time will not be near the weight of users who adjudicated properly. Only then will you clearly be able to see the experts emerge - all of it hard earned.
2.) Credibility is relative. If everyone is always correct and everyone is always building, then everyone will remain the same. It does not matter how high the number goes, if everyone has the same number, there is always a consistent balance. It is only the differential between users that matters. So that "guessing" tactic will prove fatal over long time periods because you will be wrong far more often than someone who is actually looking at everything. Again, by design this is expressed over time. The TGSAP is new, so the credibility spread between people is not significant - but rest assured it will be. The goal is to try to never be wrong, which is pretty easy if you are actually looking at submissions.
3.) New users. New users arrive with 1000 CR. Not a big deal now, but 6 months from now when participating users are sitting at 2 or 3000 CR, that new user differential will create a large spread and start shaping the bell curve. This will be meaningful.
Barthax makes a good point that most gamers who are submitting WANT to submit good submissions and so the average submission is likely to be quite acceptable, which means that it actually becomes more and more important that users properly adjudicate to be certain that they are right because being wrong just a few times can put them at the bottom of the average relative to everyone else.
Right now, everyone is essentially at the beginner level within the the system in terms of how it handles certain things like point awards and deductions. As people grow and as adjudicators and become more "powerful," awards and deductions will become more dynamic.