They definitely need a good editor and a few more passes through the article, but your complaints about the graph are based on a labeling error on the Y axis and either a reading error on your part on the X axis or a missed label in the version of the graph you were looking at.
The Y axis should either be labeled "Porportion" instead of "%" or should have the value labels multiplied by 100. If you look at one of the earlier tables, you'll see that the actual numbers are in the 66-74% range, not the 0.66-0.74% range.
The X axis (in the graph in the PDF, at least) is clearly labeled as the log edits, which means 1.5 is a perfectly reasonable value. The 0 value corresponds to 1 edit, so there should be a retention rate there as well.
To be honest, even if it was raw number of edits (which would be a mistake given the nature of the edit distribution in Wikipedia), there wouldn't be a fundamental problem with the smooth curve - it would simply be a best-fit line connecting categorical data in a scatterplot to make it easier for the reader to interpret the relationship (although including the scatterplot data is more common practice there, and the X axis should start at 1 in that case).
Referencing a figure which doesn't exist is pretty inexcusable in a final copy - *somebody* should catch that kind of thing - but this is clearly not a final copy and it's something which happens fairly often during the draft revision process. It'll get fixed, either before review or before publication. What you're seeing in the link above is clearly a pre-review draft, and it may not even be their final version for journal/conference submission.
They definitely need to do some more revision, but a cursory glance at the paper suggests that it's reasonably sound. I'll have to read it in the near future, at which point I'll determine whether or not it's really any good, but I don't particularly feel like doing so now.
Slashdot Mirror
They definitely need a good editor and a few more passes through the article, but your complaints about the graph are based on a labeling error on the Y axis and either a reading error on your part on the X axis or a missed label in the version of the graph you were looking at.
The Y axis should either be labeled "Porportion" instead of "%" or should have the value labels multiplied by 100. If you look at one of the earlier tables, you'll see that the actual numbers are in the 66-74% range, not the 0.66-0.74% range.
The X axis (in the graph in the PDF, at least) is clearly labeled as the log edits, which means 1.5 is a perfectly reasonable value. The 0 value corresponds to 1 edit, so there should be a retention rate there as well.
To be honest, even if it was raw number of edits (which would be a mistake given the nature of the edit distribution in Wikipedia), there wouldn't be a fundamental problem with the smooth curve - it would simply be a best-fit line connecting categorical data in a scatterplot to make it easier for the reader to interpret the relationship (although including the scatterplot data is more common practice there, and the X axis should start at 1 in that case).
Referencing a figure which doesn't exist is pretty inexcusable in a final copy - *somebody* should catch that kind of thing - but this is clearly not a final copy and it's something which happens fairly often during the draft revision process. It'll get fixed, either before review or before publication. What you're seeing in the link above is clearly a pre-review draft, and it may not even be their final version for journal/conference submission.
They definitely need to do some more revision, but a cursory glance at the paper suggests that it's reasonably sound. I'll have to read it in the near future, at which point I'll determine whether or not it's really any good, but I don't particularly feel like doing so now.