I have just done an analysis of the last 5000 SmackBot edits, maybe I have just wasted another 5 hours of my life, I hope not.
The background is that there are essentially tasks 4 that one should be aware of running in this period:
1. AnomieBot does what SmackBot does- over the course of this sample with any delay between 0 and 30 minutes after a tag was added.
2. SmackBot Perl (p603, p604, p605)
3. Smackbot AWB (Dating tags)
4. SmackBot Ref fixer.
This represents a new environment, since mid December. Nontheless the FAQ for SmackBot (on my user page) is still accurate.
Discounting a 18 edits caused by a new template and 24 consecutive edits by the two varieties Perl/AWB the results are
| Description | Number | percent |
|---|---|---|
| Dated a tag | 4769 | 95.38% |
| Fixed reference error | 51 | 1.02% |
| Other fixes | 67 | 1.34% |
| Logging | 8 | 0.16% |
| Detailed above | 42 | 0.84% |
| Fixing "Citation needed" | 18 | 0.36% |
| Invisible fixes | 31 | 0.62% |
| White space only | 13 | 0.26% |
| Software errors | 1 | 0.02% |
| 5000 | 100.00% |
Of course interpreting this data is the key. If you take the view that there is approximately 1% "noise" edits (much higher than before, and almost exclusively due to AWB being beaten to the edit by AnomieBot) then it might not seem bad, especially as this rate is likely to fall dramatically as the perl version takes on more of the functionality. If on the other hand you you look at the three percent of edits that are neither dating a tag, fixing a reference nor logging, and construe them as heinous crimes against humanity, and ignore the fact they all occurred with the older AWB engine, you might think differently.
Rich Farmbrough, 19:59, 23rd day of January in the year 2011 (UTC).