Even if the anomaly is ignored, it's a downward trend overall.

Sure, the individual scores are good, but the fact that they are trending downward over time should be of concern.

Is it because consumer expectaions are changing? Is it technical?, Is it staffing levels (workload)? Skill set distribution?, etc.

I'm not familar with the software industry, but these are the kind of things I would consider where I work.

I've got some experience with statistics (I had to take it in college and did mediocre in it).  This isn't a good graph to really show a "trend".  The scale should go from 0 to 100.  Also, there are too few scores.  Also, reality check.  Getting a game review in the 90+ is exceptionally rare, and you simply can't consider that the standard to which you should compare all other released games.  First game is a 95, all other games should be 90+.  Prepare for disappointment.  Should be proud of anything in the green, 76+.     

I wonder about the demographics of the responses (age, sex, etc.) and how each demographic rated the individual games (granted similar sampling sizes).

Two things:
The 'real' score of a game on metacritic can basically be measured between the aggregate scores of the reviewers and the aggregate scores of the user base. Most of Stardock's games are in line with one another while having reasonably good scores. Take a look at some other "Triple A" titles and tell me if you can say the same. Stardock doesn't have hundreds of millions to sink into "advertising heh-heh campaigns'.

Secondly, find some of the most well regarded RTSes [or any other game] that has ever been released and ask yourself how they would fare in reviewing circles if they were released today.

All this graph really suggests to me is people like Stardock's sci-fi games more than their fantasy games. Another interesting point of data you might consider adding to that graph is the trend of metacritic scores overall during the same time period.

I also think reviewers are being more critical now than in past years.  For a long stretch certain magazines were giving AAA titles super high scores that were entirely undeserving of those scores.  At the time I could only guess the scores were inflated because the magazine is under pressure to keep advertising dollars.

Personally I think it is because of PlayStations, WII's, xBox'es etc.  No one plays games on the PC much any more.  My son even got rid of all of his PC games years ago (when he was 15) and has never even thought of buying a PC game since he got his PlayStation.

Expectation versus reputation can account for the 'trend' which cannot really be considered of statistical value as it's too small a sample to be truly viable....

A trend has nothing to do with statistics..

A pattern of gradual change in a condition, output, or process, or an average or general tendency of a series of data points to move in a certain direction over time, represented by a line or curve on a graph.

You don't need a "sufficient sample size" to assess the general direction something is going.

For example. If you saw a graph showing 6 months (6 data points) of steadily declining (seasonally adjusted) sales, would you argue that there wasn't enough data points for it to be statistically valid and take no action, or would you try to figure out why the sales were declining?

If you waited for 30 months (30 data points) to assess the sales trend, it could be too late...

Heck, in every company I've worked for, 3 months of declining drivers was sufficient to raise a red flag.

Sorry nice try yourself.

First, I specified "seasonally adjusted" because I new your were going to try to make that arguement. That is, seasonally adjusted numbers are normalized.

In regard to your 100/50 comment....In finanacial analysis, if you encounter 2 extreme numbers in a trend, the first thing you do is look at the accounting records. Did the staff accountant transpose numbers? Was there an adjusting entry from a prior period that was posted in the current perid, etc.

If that doesn't yeild results, then you look at the drivers. (the volume of something that driver sales - e.g number of units sold) -  Did a mistake occur in the recording of those drivers? If not, you go to the ops folks and ask? Was there a supply disruption? An external event out of the company's control? did a customer terminate a contract? is this a one time event? Do you expect it to continue, etc?

Typically, at the beginning of each year, a budget is built based on expectations for the next year that is based on economic conditions and company events known at the time. To standardize the process, dollar amounts are built on a rate/volume basis (e.g.units x sales per unit, or job hours worked x hourly rate, etc.). If the actual results deviate from the budget, closer examination is warranted. This analysis is typically done on a monthly basis.

Only a fool allows downward trends to continue for more than three months ("until there is enough data") without examining the underlying cause.

I seriously hope you're not a member of SD's Finance team....

I don't think you are reading my full post... I already took normalization/seasonailty into account. If the decrease is know ahead of time (as built into the budget), then yes it can be ignored, but if it's outside the norm (expected results), then it should be examined.

I never said statistics was "God's gift". As a matter of fact, I was arguing, statistcs are not needed to assess a trend. Either you confused my post with someone else, not reading carefully, or are intentionally trying to be difficult.