Serverlist Statistic

[Lejo]

I wanted to have some nice graphs to show how minetest developed.
These Statistics are based on the fetches to the official minetest serverlist.
The statistics are generated from 10 day logs.
So these statistics are NOT exact and can have huge inaccuracy.

Total Fetches in this 10 days.

Ipv6 getting more and more used...

Comparing the Ingame serverlist and the server.minetest.net website

Linux vs. Windows vs. Apple-Stuff


Android vs. iOS
There is no official minetest client for iOS currently.
All iOS clients are unofficial forks mostly containing advertising.


Desktop Players vs. Mobile players
Pros vs. Noobs
All Devices with ARM Architecture are counted as Mobiles


Some 0.4.x and 5.x comparison

Versions used 2020
You can see one minetest or fork version at 0.4.13 which seem to lose updates.

In the last 3 years...
Shows how players switch to the new versions

I wonder why so much players are stuck at 5.0.
This is the Protocol Version spread which shows that a lot of players are version 36
Version 36 were never released! So this seems to be either a bug or a hyper popular fork.


There are a lot of factors which have a impact on these statistics and makes it less exact.
For example Multicraft got it's own serverlist so they disappear in the statistic

If you see any mistakes just let me know so I can fix them :)

Thanks to sfan5 for the serverlist log extraction
Source

serverlist stats.zip

(66.76 KiB) Downloaded 5 times

What else do you want to see?

[rubenwardy]

Nice!

Multiple times, you show two graphs - absolute numbers and percentages - in the same image. What's the y axis value? Didn't your science teacher tell you to label your axes, and give descriptive titles :D

[Lejo]

Multiple times, you show two graphs - absolute numbers and percentages - in the same image.

Yep, to show where the new clients come from. You can see for example that Linux clients increased and Windows clients stagnate. When only seeing the percentage you could think the clients decrease.

What's the y axis value? Didn't your science teacher tell you to label your axes, and give descriptive titles :D

Just the count of fetches to the Serverlist

[PolySaken]

Interesting. I never would have thought so many were using mobile over desktop. The android vs IOS comparison makes sense as there's no proper version for IOS. I'd have thought more people used windows, too.

[rubenwardy]

Yep, to show where the new clients come from. You can see for example that Linux clients increased and Windows clients stagnate. When only seeing the percentage you could think the clients decrease.

What's the y axis value? Didn't your science teacher tell you to label your axes, and give descriptive titles :D

Just the count of fetches to the Serverlist

Oh right, the percentage is the percentage each year, rather than the same value

[sorcerykid]

Most of the players on my server use the Multicraft client or some other iOS client. And overall, Multicraft severs tend to dominate the server list. So I suspect these statistics are heavily skewed by not properly accounting for iOS.

Also could you please document your methodology, as that is standard practice.

[Lejo]

Most of the players on my server use the Multicraft client or some other iOS client. And overall, Multicraft severs tend to dominate the server list. So I suspect these statistics are heavily skewed by not properly accounting for iOS.

Yep that’s sadly true! You could also have a look at the source files, there are more interesting things and information I will add soon.

Also could you please document your methodology, as that is standard practice.

The download zip file contains a Calc document. You find all calculations in it. The source of the numbers are the source txt files.

[firefox]

isn't the large number of mobile users because of the many multicraft servers?
i often see multicraft servers listed with 99+ active players...

[Lejo]

isn't the large number of mobile users because of the many multicraft servers?
i often see multicraft servers listed with 99+ active players...

Partly and other Forks on Android and iOS.
As new Multicraft versions have it’s own server list they aren’t rated so much.

[Lejo]

Added 0.4.x 5.x comparison and other version comparisons.

[firefox]

I wonder why so much players are stuck at 5.0

they were forced to upgrade to 5.0 because 0.4 and 5.0 versions are not compatible on multiplayer.
many mods also had to be updated to new/different features of the 5.x series.
after that, newer versions were "only" fixes and updates, so there was no urgent need to install a newer version.

i'm getting automatic updates on my linux mint from PPAs, but not all systems work that way.
since there is no in-game update, it's always the full installation process. i remember that in the older days, manual installation required some terminal commands and a lot of time.
i still did it anyway, but i guess many people wouldn't if there was no significant upgrade. or perhaps some just don't know or don't care.

[sorcerykid]

I'm curious why the graph for client versions shows that 5.0 clients vastly outnumber those of 0.4.x?

Everytime that I scroll through the server list, I notice that 40-50% of the servers with 4 or more players are Minetest 0.4.x, which means the clients must also be legacy. So how is it possible that the graph doesn't seem to reflect what is clearly evident in the server list itself?

[Lejo]

I'm curious why the graph for client versions shows that 5.0 clients vastly outnumber those of 0.4.x?

Everytime that I scroll through the server list, I notice that 40-50% of the servers with 4 or more players are Minetest 0.4.x, which means the clients must also be legacy. So how is it possible that the graph doesn't seem to reflect what is clearly evident in the server list itself?

I thought the same.
But it’s just the visualisation of the Numbers sfan5 logged.
Maybe It’s again cause some old forks have their own server list.

[rubenwardy]

There's a large number of clients with an invalid protocol number - a number which was never made into a production release, it existed only during the development of 5.0 and isn't compatible with 0.4 or 5.0. You might want to filter that out, and see how it compares

[Krock]

There's a large number of clients with an invalid protocol number

List of in-development protocol versions that you can filter:

32: latest 0.4.x stable
33 and 34: 0.4.16-compatible
35 and 36: both are breaking changes
37: last breaking change for 5.0.0 stable

5.0.0: https://github.com/minetest/minetest/bl ... protocol.h
0.4.16: https://github.com/minetest/minetest/bl ... protocol.h

[Lejo]

Yep you are right!
I added the Protocol Versions to the top post.
And this is the Cleaned Effective Version (protocol 35 and 36 subtracted)

Do you know what are all these version 36 players?
I mean it's an unreleased version, why should someone use this?

[Lejo]

This also has a huge impact on the 0.4.x vs. 5.x series
dev are protocol version 35 and 36. Only 36 is significant.

Now 0.4.x is even more popular than 5.x

But what are these "lost 36th"?
They are a very huge group.