Basic Statistics
Robert Meusel
Oliver Lehmberg
Christian Bizer
Sebastiano Vigna

This document provides basic statistics about the topology of the Web Data Commons - Hyperlink Graph extracted from the 2012 Common Crawl Corpus covering 3.5 billion web pages and 128 billion hyperlinks between these pages.
We analyze the page-level graph as well as a pay-level-domain aggregation of the graph.


1. Page Graph

The page graph consists of over 3.5 billion nodes and over 128 billion arcs between those nodes. In the following we will report about the basic findings within the page graph. A more detailed analysis can be found in: Graph Structure in the Web - Revisited by Robert Meusel, Sebastiano Vigna, Oliver Lehmberg, Christian Bizer, an accepted paper at the 23rd International World Wide Web Conference (WWW2014), Web Science Track, Seoul, Korea, April 2014

1.1 Indegree and Outdegree Distribution

The following two figures show frequency plots of indegrees and outdegrees in log-log scale. For each d, we plot a point with an ordinate equal to the number of pages with that have degree d. Note that we included the data for degree zero, which is omitted in most of the literature. We then aggregate the values using Fibonacci binning to show the approximate shape of the distribution. We find the page with the highest indegree to be referenced by of 95 million other pages and the page with the largest number of links to contain almost 56 thousand references to other pages.

Indegree Frequency Plot
Frequency plot of the indegree distribution
Outdegree Frequency Plot
Frequency plot of the outdegree distribution

1.2 Connected Components

The following figure (left) shows the distribution of the sizes of the weakly connected components using a visualization similar to the previous figures. The largest component (rightmost grey point) contains about around 94% of the whole graph (over 3.34 billion pages). The right shows the distribution of the sizes of the strongly connected components. The largest component (rightmost grey point) contains 51.3% (over 1.82 billion pages).

WCCs Frequency Plot
Frequency plot of the distribution of WCCs
SCCs Frequency Plot
Frequency plot of the distribution of SCCs

1.3 Bow-Tie Structure

Having identified the giant strongly connected component, we can determine the so-called bow tie, a depiction of the structure of the web suggested by Broder et al.. The bow tie is made of six different components:

All these components are easily computed by visiting the direct acyclic graph of strongly connected components (SCC DAG): it is a graph having one node for each strongly connected component with an arc from x to y if some node in the component associated with x is connected with a node in the component associated with y. The bow tie of the page graph is shown in the following figure:

Bow-Tie Structure of the Page Graph
Bow-Tie Structure of the Page Graph

2. Pay-Level-Domain Graph

The PLD Graph is created from the Page Graph by aggregating all pages that belong to a pay-level-domain (PLD) into a single node. For example, the pay-level-domain of is All pages under one PLD are aggregated into one node and all internal links are removed. External links, i.e. those leading to other PLDs, are kept, but only once.
The PLD graph consists of over 43 million websites/nodes and over 623 million distinct arcs between those nodes.

2.1 Weakly and Strongly Connected Components

The PLD graph contains one giant weakly connected component that consists of 39.4 million nodes (92%). The following diagrams show the distribution of the weakly connected components (WCC) and the strongly connected components (SCC) in the PLD graph. The x-axis charts the size of the components and the y-axis draws the number of components for a particular size. Both axes are log-scaled. The respective distribution is marked by the blue dots and the calculated power law distributions are shown as black lines. The distribution of the sizes of WCCs follows a power law with an exponent of 1.411. The power law is calculated excluding the giant component.

wcc distribution
Distribution of the Weakly Connected Components (PLD-Level)
wcc distribution
Distribution of the Strongly Connected Components (PLD-Level)

2.2 In- and Out-Degree Distribution

The following diagrams show the in- and out-degree distributions of the PLD graph. The respective distribution is marked by the blue dots and the calculated power law distributions are shown as black lines. Both distributions follow a power law, where the exponent for in-degree is 1.56 and the exponent for the out-degree distribution is 1.49.

indegree distribution
In-Degree Distribution (PLD-Level)
outdegree distribution
Out-Degree Distribution (PLD-Level)

2.3 Bow-Tie

We further determine the Bow-Tie structure of the PLD Graph. The Bow-Tie structure consists of the largest strongly connected component (LSCC) that is flanked by two other components, IN and OUT. The IN component contains all nodes that are not part of the LSCC, but are connected to it via a directed path. Similarly, the OUT component contains nodes that are, via a directed path, reachable from the LSSC. Connected to the IN and OUT component, there are TENDRILS that lead away from the IN or towards the OUT component respectively. If two such TENDRILS are connected, they form TUBES. The TUBES component contains nodes that directly connect the IN and OUT components, without going through the LSCC. All remaining nodes, i.e. those that are not part of the largest weakly connected component, belong to the DISCONNECTED component.

Bow-Tie of the PLD Graph

2.4 Frequently Interlinked PLDs

We also identified weakly connected components (WCC) of frequently interlinked websites. The picture below shows one WCC in which all websites are interlinked by at least 500,000 links connecting individual pages of both websites.

WCC of Strongly Interlinked PLDs

2.5 PLDs ranked by In- and Out-Degree

The following table shows the pay-level-domains with the highest in- and out-degree.

Rank PLD out-degree
1 3,898,561
2 2,249,553
3 1,078,938
4 862,705
5 699,609
6 668,271
7 650,884
8 551,360
9 496,645
10 496,045
11 476,890
12 424,646
13 423,918
14 392,543
15 371,087
16 370,918
17 362,310
18 352,668
19 312,924
20 276,578
PLD in-degree 1,822,440 1,319,548 1,243,291 1,156,727 1,034,450 782,660 710,590 646,239 554,251 339,231 314,051 312,396 289,605 269,003 226,596 201,863 196,690 196,507 193,907 189,663

2.6 Top-Level-Domain Distribution

The following table shows the distribution of top-level-domains in our PLD Graph. The third column of the table compares the ranking of the PLDs with the ranking published in the Verisign Domain Name Industry Brief from the same time when the crawl was gathered. According to the Verisign statistics, our PLD Graph contains 18% of all registered pay-level-domains.

TLD PLDs Percent Rank Verisign Rank
com 21,205,742 49.44% 1 1
de 2,994,885 6.98% 2 2
net 2,321,230 5,41% 3 3
org 2,194,525 5.12% 4 6 1,547,132 3.61% 5 5
nl 1,017,387 2.37% 6 8
ru 724,943 1.69% 7 9
info 697,814 1.63% 8 7
it 646,139 1.51% 9 - 524,610 1.22% 10 -
others 9,015,393 21.02% - -

2.7 Linkage of Websites grouped by TLD

Beside the general connectivity of the different pay-level-domains we analysed the connectivity of pay-level-domains between the different top-level-domains they belong to. The figure below shows the hyperlinks of the pay-level-domains aggregated by their top-level-domain of the 10 largest TLDs. All remaining domains are collected in the group named "others". In this diagram, the outermost circle labelled with percentages represents the total number of links for each group. Directly adjacent to this circle are two smaller bars for each group. The outer bar represents the number of in-links while the inner bar represents the number of out-links. Further to the centre there is another circle labelled with absolute numbers. This circle, again, represents the total number of links. From this circle ribbons spanning through the middle represent interconnections between the different groups. Incoming links have a white gap between the ribbon spanning in the middle and the circle part labelled with absolute numbers. Also, the ribbon has the same colour as the group that it originates from.
Linkage of Websites grouped by TLD
For the two largest groups of websites, ".com" and ".de" we find that the percentage of intra-TLD links is outstandingly high. These domains tend to strongly link within their top-level-domain. The opposite we find for the ".org"- and ".net"-TLDs. Both only have a small fraction of intra-TLD-links compared to inter-TLD-links.

3. Code used for the Analysis

The source code that we used to generate the statistics presented above can also be downloaded from our Subversion repository.

4. Credits

Lots of thanks to

The creation of the WDC Hyperlink Graph was supported by the EU research project PlanetData and by Amazon Web Services through a Machine Learning Research Grant. We thank your sponsors a lot for supporting Web Data Commons.

PlanetData Logo    AWS Logo