FAF Statistics Megathread

win ratios on Seton's Clutch for different factions:
All ratings:
1606829188100-download19.png
1500+
1606829192125-download18.png
Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=10#p139093

Following on from my last graph, I decided to investigate how many "active" players we have in each rating group.

I decided that "active" for globally rated games was 50 games or more:
1606829499626-download-3.png

From this we can see that the growth in active players is mainly driven from people with global MEANS in the 1000 - 1500 range (aka, ~800 to 1300 rating)

The larger numbers of these players masked trends for higher rated groups in this graph, so I decided to plot them separately:

While there were never very many players in the 2000+ mean bracket (1800+ rating) there has been a clear downward trend, with there only being half the number active in that group as this time two years ago.
1606829526557-download1.png
I also did the same for ladder, with "active" meaning 10 games or more in a month (since fewer ladder games are played:

1606829674619-download2-3.png
Here its quite clearly the low rating groups which are the odd ones out, with the number of active people with 1000 mean or less (~800 rating or less) on ladder increasing drastically in recent months.

All other rating groups for ladder seem to be falling significantly though.

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=20#p139889

Summed to periods of a month

factions, assuming 1=uef, 2=aeon, 3=cybran, 4=sera. no idea what 5 and 6 are

all ratings
DfXCTZm.png

1000+ mean
KRdMxco.png

1500+ mean
iYBxMRo.png

1800+ mean
cI5ushU.png

2000+ mean
ofPijxh.png

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=30#p149466

mount of ladder games / week.
There are some differences with this one viewtopic.php?f=2&t=13449&start=10#p139077
My graph has some drops in activity at newyears that are not in soflty's graph.

ladder_popularity.png

Here is a version with data only from 2016-2017
ladder_popularity_16-17.png

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=30#p149475

Old Forum stats :

Users with the most topic creations in general discussions :
1606830570473-download7.png
And a version for those with a very wide screen:
loong.png

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=40#p151990

Highest mean views per topic created (at least 5 topics created in General Discussions):
1606830581545-download10.png
Number of topics created each month in General Discussions:
1606830584799-download9.png

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=40#p152003

The following graphs were created from a join the the game_stats and game_player_stats table using the game id.

Methodology:

  • Games where no players could be found were discarded
  • Even teams => At least 2 teams, and all teams have the same number of players
  • Game is marked as valid => FAF thought it was a ranked game

1606830644299-download11.png
1606830649853-download12.png
1606830653035-download13.png

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=40#p152860

Global rating over total games (global + ladder)
https://plot.ly/~bsse/27/
Screenshot from 2020-12-01 15-08-09.png
Ladder rating over total games
https://plot.ly/~bsse/25/
Screenshot from 2020-12-01 15-08-25.png
Global rating over ladder rating
https://plot.ly/~bsse/23/
Screenshot from 2020-12-01 15-08-41.png

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=80#p164738

Ladder win percentages by faction and rating:
https://plot.ly/~bsse/45
Screenshot from 2020-12-01 15-09-28.png

Games played by each faction and rating:
https://plot.ly/~bsse/50/
Screenshot from 2020-12-01 15-09-42.png

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=80#p167079

Lower rated players gain rating from draws, and the data for "0 rating advantage" actually refers to the range 0-10 rating advantage.
1606831053764-download17.png
Second view, focused on new players.
1606831057164-download18.png
Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=90#p168635

1606831352997-download22.png
1606831357294-download23.png
Mean success rate: 324034/521269 ==> 62.2% (Trueskill)
Mean success rate: 286803/521269 ==> 55.0% (Simple Ladder)
Mean success rate: 247667/521269 ==> 47.5% (Random Guessing)

The 'Simple Ladder' scheme awards players -1, 0, and 1 points for losses, draws, and wins respectively.
Note that the data used for these experiments is biased by having had the trueskill rating scheme used as the matchmaker.

Both plots are for 1v1 results only. Other rating schemes may appear in future plots, if you have ideas or questions let me know.

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=90#p169175

A player is considered "active" if their player details have been updated in the past month, e.g. by logging in.
1606831388832-download24.png
A player is considered "retained" if they are active.
1606831397912-download25.png

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=90#p173021

Old stats from the dark world of FAF moderation :

M9HyCzV.jpg

Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=100#p175559

we had some 2012 supremacy discussions recently, i've got bored and made some plots.
all stats are from the current leaderboards, so the plots don't account for smurfs, inactive players etc.

rating on the x axis, number of players on the y axis, different sub plots by years

top 5000 global rated players
E3wM1Gd.png

top 5000 ladder rated players
of5j9GP.png
Source : https://forums.faforever.com/viewtopic.php?f=2&t=13449&start=120#p184762

LOTS GROUP STAGE 2020 STATISTICS – ALL GROUPS
There were 54 total matches. None of the matches went to a draw.

How many times was each faction played?
UEF – 38 (35%)
Seraphim – 30 (28%)
Cybran – 24 (22%)
Aeon – 16 (15%)

How many mirror matchups were there?
Total – 11 (20%)
UEF – 5 (9.3%)
Seraphim – 4 (7.4%)
Cybran – 2 (3.7%)
Aeon – 0 (0.0%)

Particular Non-Mirror Matchups
UEF-Seraphim - 11 (20.4%)
UEF-Cybran - 9 (16.7%)
UEF-Aeon - 8 (14.8%)
Seraphim-Cybran - 7 (13.0%)
Seraphim-Aeon - 4 (7.4%)
Cybran-Aeon - 4 (7.4%)

Favorite Faction
archsimkat – Aeon (50%)
Blackheart – Seraphim (43%)
Blast_Chilled – Aeon/UEF (33%)
BlInChIk – Cybran/UEF (50%)
Blodir – Cybran/UEF (50%)
Bullydozer – No favorite
Espiranto – Seraphim (50%)
JaggedAppliance – Aeon/UEF (33%)
Nexus – Cybran (50%)
Paralon – No favorite
Petric – UEF (38%)
Swkoll – Seraphim (43%)
Tagada – UEF (67%)
TURBO2 – UEF (86%)
TurinTurambar – Cybran/Seraphim (33%)
ZLO – Seraphim (71%)

LOTS GROUP STAGE 2020 STATISTICS – GROUP B

Morax provided me with the faction choices and vetoes for each of the 12 matches played by Group B.

Veto Frequency:
Seraphim – 10 (42%)
UEF – 9 (38%)
Cybran – 3 (13%)
Aeon – 2 (8.3%)

Most-Vetoed Faction
archsimkat – none
Nexus – none
Paralon – UEF (66%)
TurinTurambar – Seraphim (66%)

How often was a player’s first choice vetoed?
Total – 18 out of 24 (75%)
Seraphim – 9 (37.5%)
UEF – 5 (20.8%)
Cybran – 3 (12.5%)
Aeon – 1 (4.2%)

by player - how often this player's first-choice faction was vetoed:
archsimkat – 5 (83%)
Nexus – 5 (83%)
Paralon – 4 (67%)
Turin – 4 (67%)

How often did a player veto their own first choice?
Only 4 out of 24 times (16.7%). Each of the players did this only 1 time.

How often did a player veto the same faction for the same opponent?
Each player in Group B played 2 games against each other player.
Paralon vetoed UEF twice against Nexus
Turin vetoed Seraphim twice against Paralon and against Nexus
Turin and archsimkat both vetoed UEF twice against each other

How often did both players make the same first choice?
8 out of 12 times (67%).

How often did both players make the same first- and second choices?
Only 1 out of 12 times. On desert arena, Nexus and Turin both chose Seraphim and UEF.

Impressions:

The veto system seems to be accomplishing its goals (reducing mirror matchups and increasing faction diversity). Two mirror matchups were very poorly-represented in the group stage (just two Cybran-vs-Cybran games, and no Aeon-vs-Aeon). Otherwise, every single possible faction matchup was represented at least 4 times.

Seraphim at the moment tends to be favored, at least in ladder matches, but it was not the most common faction in this tournament. There were a very small number of Seraphim-Seraphim mirror matchups, just 4. They took place on crazyrush (2 games), desert arena, and Loki. Based on information from Group B, this is apparently because of the veto system.

It is surprising to me how infrequently players vetoed their own first choice. Each of the four players in Group B played 6 matches and only did this 1 time. Also, it was less than half the time that players vetoed the same faction for the same opponent. Each player in Group B played 2 matches against each other player in Group B. Only 5 times out of 12 did the players veto the same faction in both matches.

These figures indicate that players look at both the map, and the other player’s skills with a particular faction, when deciding what to veto. It seems that players did not just decide “Swkoll likes UEF, I will ban that” or “UEF is best on crazyrush, I will ban that.” The exception is that everyone who played on Crossfire Canal (archsimkat, Nexus, and Paralon) chose Seraphim AND banned Seraphim.

Even though players rarely vetoed the own first choice, they frequently vetoed their opponent's first choice, 75% of the time. This suggests that the players have a good idea of which factions their opponents prefer to use on any particular map. If you know that your first choice will probably be vetoed (a 75% chance) and that your opponent's first choice will probably be vetoed (a 75% chance) this means there is a high chance you and your opponent will both get the second-choice faction. Of the 12 games, in 7 of them, both players got their second choice.

While players usually made the same first choice of faction (8 of the 12 games), they rarely also made the same second choice (only 1 out of the 12). While players may have some idea that a particular faction is best on a particular map, it seems players have more diverse opinions about which factions may be second-best. And they might be playing a meta-game along the lines of "when I ban Seraphim, my opponent's second choice will be Cybran, and if they're going to be Cybran, I would rather have UEF."

If players had the same idea about what the second-best faction is on a map, we would expect to see a lot of mirror matchups as a result. For example, if both chose sera-cybran then both would probably veto seraphim (because 75% of the time, players picked their opponent's first choice to veto) which would result in a mirror matchup. But we didn't see that very often, at least with the limited data we have (Group B only).

Comparison to LotS 2019

In 2019, out of 56 matches, 41 were mirror matchups (73%). Cybran was the most common (38%), followed by UEF (20%), Aeon (9%), and Seraphim (7%). The primary reason given for adopting the veto system was the prevalence of mirror matchups during LotS.

in 2019, few people wanted to play Seraphim. In 2019, only 10 out of 56 matches (18%) had any Seraphim. This time, people were largely prohibited from playing as Seraphim, and still 26 out of 54 matches (48%) had Seraphim.

In 2019, Aeon-Cybran and Seraphim-UEF only occurred once each. Aeon-Seraphim and Aeon-UEF did not happen at all. So those four matchups occurred only a total of 2 out of 56 times in 2019 (3.6%). But in 2020, those four matchups occurred a total of 27 times out of 54 (50%).

Clearly, the veto system has accomplished what it was intended to do, namely to promote faction diversity at the tournament and to discourage mirror matchups. Whether it is a good thing, or a bad thing, is a separate question--this analysis doesn't see any of the drawbacks to the veto system. Maybe players think it is clunky, or some players hate not getting to play their chosen faction. I can't say whether it is good or bad, but no one can say that it isn't working.