Can you expound a little on how the GPA is derived?
>[Letter grades for groups sampled]
>For 1 day time periods ending close to 1998/04/01 12:39 GMT
> 6 newsgroups Insufficient data for grading.
> (Less than 5 articles expected.)
> 187 newsgroups "A" (At least 93% of expected)
> 11 newsgroups "B" (85-92% of expected)
> 0 newsgroups "C" (78-84% of expected)
> 0 newsgroups "D" (70-77% of expected)
> 0 newsgroups "F" (Worse than 70% of expected)
> GPA: 3.94
The GPA is a single number rating of the server. It is calculated based on the "A", "B", "C", "D", "F" letter grades as you would expect to see in an academic environment: The "GPA" is the weighted average of A,B,C,D, F letter grades.
Divide by total number of newsgroups graded. (Don't include newsgroups tagged as "insufficient data for grading")
- Each "A" counts as 4.0 points,
- B = 3.0
- C = 2.0
- D = 1.0
- F = 0
Although having a single number to rank a server is convenient, it does not tell the whole story, just as someone who got an "A"s in "basket-weaving" and "volleyball" might have failed the subjects that are more important. (No offense intended to you volleyball-playing basket-weavers. More power to you.)
The calculation of the performance as a GPA does have some nice characteristics though:
But don't base all your decisions on the GPA. (I know, some people will ignore this advice anyway.) Here is what I say to them.
- It limits the amount of "damage" one newsgroup can do to your GPA.
If one newsgroup gets 20,000 postings a day (None do, but if one did) the total of the
posts in all newsgroups could cause this one newsgroup's performance
to drag all the other 10-20 post per day newsgroups with it (either up or down.)
Example: Many sites don't do well in alt.binaries newsgroups. Since there aren't that
many of those in the sample set, you can still get a very high GPA when you do
well in the rest of the newsgroups. (Get an A in 180 non-binaries newsgroups,
fail (get an F) in 3 binaries newsgroups results in a GPA of (180 * 4.0 / 183) = 3.93.)
- It limits the amount of "good" one newsgroup can do.
The highest you can get is an "A" If you have 150% in one newsgroup, that extra 50%
does not offset poor performance in another newsgroup. (And if you had 150% it is
probably because you are catching up from backlog, or have a local spammer.)
- You can compare two servers, even if they don't have the same sample newsgroups.
Groups you don't have are excluded as "insufficient data for grading." This allows
you to compare the GPA's of server A, with 150 newsgroups, to server B, with 199
groups in the sample set, for example.
Very few things in life are well characterized by a single number. Newsserver performance is complex. Knowing that something is performing poorly may not help you know what to change to improve it. There is a reason that the other parts of the report exist. Learn to read and use them also.
When you want details, move from the GPA to the [Newsgroup Heirarchy Reports] This section can show interesting differences in performance. It may show that you did very well on comp.*, but poorly elsewhere. Note that these are totalled values, so one newsgroup can drag the performance of others up or down.
And finally, the [Individual Newsgroup Reports] section is an interesting scan from time to time when you are trying to find out why something is good or bad, or to check on your favorite newsgroup, like "news.software.nntp."
View an example report
Why are results reported by heirarchy?
Newsrate Report Format: What does "n/a" mean?
What exactly is being counted/compared?
How do I signup?
Up to The newsrAte RKT
Up to newsrAte home
Up to Mib Software home
Copyright 1998, Forrest J. Cavalier III, Mib Software
INN customization and consulting