The afd process could not possibly be responsible for that change in the graph. Running flat out it could only process about 200 articles per day. More usually it runs between 100 and 150 per day. Even if every single listed article were deleted it wouldn't dent the growth rate. A more realistic estimation of the deletion rate at AfD would be about 75%, or perhaps 85% if you regard a redirect, merge or userfication as a deletion for the purpose of counting the number of distinct articles in mainspace.

To estimate our article creation rate I used special:newpages to count the number of articles created two weeks ago (16 June). This gave me 1866. These 1866 articles are what remains after two weeks during which the worst are speedy deleted. So I think it's a pretty reliable predictor for the growth of article count. The encyclopedia is growing by approximately 1900 articles per day.

We've deleted 1726 items from article space in the past 24 hours by speedy deletion, proposed deletion or articles for deletion.

Notice that the number of articles deleted daily and the growth rate (after deletions) are very close. Put it another way, we're deleting nearly half of all new article starts.

Now going back to Lih's pretty graphs, I see firstly that my figure of 1900 is congruent with his graph for "increase per day". He shows the figure hovering between 1500 and 2000 per day during the past two quarters. This is a pretty reasonable growth rate and very much the kind of thing I'd expect. Li is missing the kind of detail on deletions that I've provided here.

Notice that this growth rate of 1900 per day is still rather higher than it was (according to Li's graph) when we turned off anonymous article creations in late 2005. We haven't turned off the faucet, though we may have avoided a flood.

One thing I don't really understand about Li's graph is that he chose to scale it so that the increase per day graph is overlaid exactly on the "Number of articles" graph. This would only make sense if he were positing exponential growth (the first differential of an exponential is also an exponential). But exponential growth is not possible in an environment with limited resources. Matching the two graphs like that misleads the viewer, who looks at the growth graph and thinks something catastrophic must have happened because it isn't rising exponentially.

The rest is just a curve-fitting exercise. Garbage in, garbage out. The truth is (as always) we're getting more article starts than we know what to do with so we're junking half of them, those that we've decided aren't promising.