Introduction
Approximately 5 months ago, I conducted a statistical analysis on
node and noder data. The analysis answered several questions with
respect to average writeup reputations, distribution of
noders per level, noder retention, etc. Over the past few months, I have
received many requests to prepare an update on these statistics.
However, there was a major obstacle holding me back in
doing this...
Until now!
JayBonci sent me a new E2 datafile that was collected on January
20, 2002. In this writeup, I shall make an attempt to compare the new
statistical results to the ones that were gathered on September 29,
2001. For further reference, please read the following:
Note that the previous statistical analysis was primarily conducted
to review changes in the Level Advancement
system, a.k.a. the Honor Roll. The current statistical
analysis is mainly for general interest.
The statistics were collected in similar manner to those collected on
September 29, 2001. Data for noders with zero writeups was discarded.
Also, all Webster 1913 data was discarded. As a result, the statistics
are representative for the active noders on E2.
Average writeup reputation per noder
The
merit (
Interquartile Mean writeup-reputations) of each
noder were calculated. The distribution for the average writeup
reputations is given in the following table:
Rep. 092901 012002 092901 012002
% <rep % <rep % %
... ... ... ... ...
-5 0.8 0.9 0.4 0.4
-4 1.5 1.5 0.7 0.6
-3 2.9 2.6 1.4 1.2
-2 5.7 5.5 2.8 2.9
-1 11.6 11.0 5.9 5.4
0 25.5 22.3 13.9 11.4
1 42.6 36.1 17.0 13.8
2 58.2 49.5 15.6 13.4
3 71.2 62.0 13.0 12.4
4 80.3 72.2 9.1 10.3
5 86.5 79.9 6.2 7.7
6 90.4 85.3 3.9 5.4
7 92.9 89.0 2.5 3.7
8 94.6 91.6 1.7 2.7
9 96.0 93.6 1.4 1.9
10 96.9 94.9 0.9 1.4
... ... ... ... ...
The data reveals that the average writeup reputation for each noder
has increased over the past months. For instance, on 09/29/01, 13.5%
of the noders (100-86.5) had an average node reputation of 5 or higher.
On 01/20/02, this amount was increased to 20.1%. The last two
columns in the table show that the largest percentage of noders have an
average writeup reputation between 0--1. This may seem a bit surprising,
but this is probably due to a large number of noders who "try" E2 for
a few writeups, and quickly give up. The average noder writes at a
merit (Interquartile Mean reputation) of 3.06; this
average was 2.31 on September 29, 2001. Apparently, the writeup
quality has improved over the last few months. Note that this average is
not directly apparent from the data in the table, due to the skew in
the distribution. The lower tail of the distribution is relatively short
(no noders with really bad merit), while the upper
tail of the distribution is relatively long.
Per-level data
The following table shows the statistics on a per-level basis. The data
for Level 1 excludes noders with zero writeups. It is interesting to
note that the average merit for each level has increased over the past
few months.
Another interesting statistic is the "noder retention", which can be
expressed by the number of noders in Level 2 and up, divided by the
total number of noders with one or more writeups. On September 29,
2001, the noder retention was 17.4%. On January 20, 2002 the
noder retention was 17.9%. Although there is a slight increase in
noder retention, approximately 82% of the community will never node themselves past level one.
noders | average merit | average devotion
Level 092901 012002 | 092901 012002 | 092901 012002
1 10251 10299 | 1.7 2.7 | N/A 14
2 1061 1126 | 3.4 4.3 | 138 179
3 514 548 | 3.6 4.4 | 353 440
4 261 256 | 3.9 5.0 | 739 943
5 140 142 | 3.9 5.2 | 1169 1547
6 63 66 | 3.8 4.5 | 1716 2001
7 37 34 | 4.2 5.4 | 2493 3160
8 30 30 | 4.7 5.4 | 3871 4434
9 25 24 | 3.3 3.6 | 3578 3783
10 25 23 | 3.3 4.0 | 4778 5619
11 4 3 | 3.1 4.2 | 6233 8793
12 0 0 | - - | - -
13 1 1 | N/A N/A | N/A N/A
Writeup data
The following table lists the distribution of all writeups, as a
function of writeup reputation. This data shows that there are
relatively few writeups with
very low reputations in the
database. Only 0.4% of all writeups on E2 have a reputation of -5 or
lower. At the upper end, there is a lot more spread. Approximately 90%
of the writeups have a reputation of 16 or lower. However, a writeup
with a reputation of 100 or more ranks among the top 0.1%. There are
approximately 280 writeups with a reputation over a 100.
092901 012002
Reputation #WU % < rep #WU % < rep
-5 747 0.4 664 0.4
-4 1513 0.8 1184 0.7
-3 3105 1.6 2639 1.4
-2 7436 3.5 6207 3.0
-1 18988 8.3 16027 7.1
0 57866 22.8 43799 18.3
1 54741 36.6 47885 30.6
2 44137 47.7 41314 41.2
3 35475 56.6 34459 50.1
4 27985 63.7 28510 57.4
5 22306 69.3 23274 63.4
6 17846 73.8 19217 68.3
7 14439 77.4 16137 72.4
8 11593 80.4 13057 75.8
10 7654 84.6 8822 80.8
12 5992 87.8 7102 84.6
14 4546 90.3 5504 87.6
16 3433 92.2 4343 89.9
20 2157 94.8 2647 93.1
25 1231 96.8 1614 95.7
30 716 98.0 952 97.1
40 259 99.1 385 98.7
In the period September 29, 2001 - January 20, 2002,
a net total of 7393 writeups were deleted. Most deletions occurred on
writeups with a reputation ranging from -1 to 1.
Honor Roll data
If we consider
only the noders with 25 writeups or more (
Level
2 noders, assuming they satisfy the XP conditions), 488 out of 2303
noders have a
merit of 6 or more. Thus,
21.2% of all
Level 2 noders and up will benefit from the
Honor Roll.
On September 29, 2001, 267 out of 2205 noders had a merit of 6 or
more, which amounts to 12%. The initial proposal called for a
merit of 5 or more to enter the Honor Roll, which would benefit
approximately 20% at that time.
In the future, the target entry level for the Honor Roll will be
adjusted, so that approximately 20% of the Level 2 noders and up will
benefit from it. The entry level will most likely be adjusted upwards in
the near future. This should also be beneficial for noders with
merits above the upper cut-off for the Honor Roll (2 ×
the entry value for merit).
Conclusions
The average writeup reputation for all noders is steadily rising. This
may be attributed to deletion of superseded writeups, better
noding and more voting noders present.
Only a small amount (less than 0.4%) of the writeups has a writeup
reputation less than -5. Apparently the life-time of these type of
writeups is not very long. Poor writeups are removed before they gather
significant downvotes.
From September 29, 2001 to January 20, 2002 there was a negative
growth in the number of writeups. However, this is most likely a
beneficial effect on the database. E1 writeups are consolidated,
and often one detailed writeup replaces several older writeups. The
statistics suggest that the overall quality of the site is continuously on the rise.