FAF Statistics Megathread 2 Statistics Boogaloo

[IndexLibrorum]

@Sainse There's an error here and there in the labels, soz

[Sainse]

Ranked (rated) FAF stats

Goals and methods

I would like to contribute by providing more graphs focused on competitive gameplay. So all images you are going to see are for rated replays. I've used the same API interface IndexLibrorum had described in the initial post. Thanks to him for helping out with the initial setup.

AI, coop, sandboxes as well as unranked games are excluded from my analysis. Data scope is between 1st January 2017 until 31st July 2025. It's about 23 GB of .json data later converted to .parquet for processing.

Total number of games

The total number of rated games on a yearly basis ranges between around 250k to 400k. FAF peaked at the middle of pandemic with 395k games played in 2021. In 2025 there has been 138k games played by the end of 31st July. On average around 60-65% rated games are played by this date. That allows to make an approximate projection that by the end of 2025 the amount of rated games should reach about 235k.

Total total numberof all FAF games in 2018-2019 was around 1.25 million. It allows to conclude that around 20% of FAF games are rated. However it does not mean that 80% of FAF is coop/ai/survival, as there are many more reasons for game to not become ranked, such as desync, con issues or being a sandbox game.

Rated games

The peak year was in 2021, but what about peak month? November 2020 with 40 099 games.

New players

The best way to "zoom in" new players dynamics without downloading pre-2017 data is to consider first 3 months players to be "old" players. This is still not accurate for first 3-6 months, but after that point it better shows ups and downs. You can see that amount of new players somewhat increased between 2020 and 2021, but not drastically if compared to 2018-2019 levels.

New and unique players

Note that the data points for new players aren't accurate for the first couple of months. The unique players are accurate for the entire image.

Players and ratings

Note that there are two Y-axis. The left one is number of games (tens of thousands and no less than 16k), the right one is amount of new/unique users (thousands).

Leagues distribution

For simplicity’s sake I call game type/game mod/rating type a ‘league’. In the past the only game types were global and 1v1 ladder. In 2021-2023 to the matchmaker were added 2v2, 4v4 and last but not least 3v3. 2v2 never got any significant share of games. 4v4, on other hand, was pretty significant - until it got eaten by 3v3. Ultimately 3v3 TMM grabbed some share from global and 1v1.

Number of games

Percentage share

Global

Note than median and average are not the same. Median means that 50% of players are below and another 50% of games are above median. Top players drive the average significantly higher than median.

Average/median rating and new/unique players

Arrival of new players drives both average and median games rating down. The lack of new players does the opposite, since skilled players are more likely to remain. There is no other observable factor to rating "inflation" or "deflation".

Number of avg rated games

In global almost every game with an average rating higher than 500 is also higher than 1000. In 2017-2020 they were almost perfectly matched. Today they are a bit more spread apart, but still surprisingly close.

Average games rating, percentage

Ladder

Players and ratings

Number of avg rated games

Ladder is the only league with a significant amount of games with average rating less than 500. The share of games with average rating between 500 and 1000 is ladder is also much higher than in global, where 500-1000s players are much more likely to get teamed up with 1500+ in some big lobby. As a consequence, in ladder both median and average are around 300-400 points smaller compared to global.

Average games rating, percentage

Top ladder

Amount of ladder games with an average rating above 1800 has peaked at over 1000 games per month at the end of 2018-beginning 2019 and decreased ever since. It did not benefit from the increased player activity during the pandemic.

Number of avg rated games

Share of games

Table

Here only Januaries of every year are compared, not every month or every year. With the overall decrease in games the ratio of >=1800 to >=2000 games increases, as two 2k players are becoming more likely to be matched against sub-2k player during the overall shortage of activity.

Year, month	>= 1800	>= 2000	Ratio
2017-01	634	265	2.39
2018-01	917	356	2.58
2019-01	1257	730	1.72
2020-01	447	149	3.00
2021-01	470	232	2.03
2022-01	476	174	2.74
2023-01	311	86	3.62
2024-01	306	86	3.56
2025-01	187	27	6.93

2v2 TMM

Players and ratings

Number of avg rated games

In 2v2 surprisingly share of 500-1000s games is even less than in global. There is also virtually no <500 games. Average and median are only slightly below global levels.

Average games rating, percentage

3v3 TMM

Players and ratings

Number of avg rated games

Average games rating, percentage

4v4 TMM

Players and ratings

Number of avg rated games

With introduction of 3v3 the amount of 4v4 games has significantly decreased. Most people queue both for 3v3 and 4v4, and 3v3 inevitably fills faster than 4v4.

Average games rating, percentage

[IndexLibrorum]

Nice, thanks for doing more stats. Glad the code I sent helped.

I think you posted the same graph twice under Leagues distribution.

Arrival of new players drives both average and median rating down. The lack of new players does the opposite, since skilled players are more likely to remain. There is no other observable factor to rating "inflation"/"deflation".

Without a proper regression analysis you cannot really draw this conclusion. Also, how did you calculate rating? From memory your numbers are different than mine. I suspect you might have calculated rating wrong.

The graph for 'Average games rating, percentage' would make more sense as a stacked bar chart of a stacked area chart. Current the line for >=1000 is this high because it includes ALL games above average 1000 rating. That's not very informative.

[Sainse]

@IndexLibrorum said in FAF Statistics Megathread 2 Statistics Boogaloo:

I think you posted the same graph twice under Leagues distribution.

Fixed, thank you.

Also, how did you calculate rating? From memory your numbers are different than mine. I suspect you might have calculated rating wrong.

Sum of players' ratings at the beginning of the game divided by the amount of players in the game. This is games' median/average, not players' median/average. If Player1 has 10X more games than Player2, it will have 10X more impact on games' median/average. So it pretty much should give a different number.

The graph for 'Average games rating, percentage' would make more sense as a stacked bar chart of a stacked area chart.

I've reused the template I already had as it's still readable. But ngl that's indeed a better idea, I will take a look at it.

[Skrat]

@Sainse Awesome! Thanks!

[TexasDeath]

For brand-new players or “noobie noobs,” FAF can be brutal. Maybe keep the kids in a sandbox until they’ve put in some playtime or shown basic skills, before releasing them to the blood-thirsty high-rated players who like to pick off noobs or punish them for taking the “wrong pile of rocks.” (Referring to the DG map with its unwritten house rules.) The learning curve is just too steep for new players to jump straight onto the main server—most don’t have any fun and leave.

The “foe” system also needs work. Foeing should be temporary—maybe 1 day, 5 days, whatever—because players forget who they foed. Suddenly they can’t host a game because others can’t see their lobby. Permanent foes just shrink the active player base. A timed expiration would let wounds heal without permanently cutting connections.

International play is another huge issue. Right now, Russian, AU, and North American players can rarely get a smooth game together. Two out of three games restart within a few minutes of launch. That kills momentum and frustrates everyone. Think about it—FAF is supposed to be a global platform, but cross-region games barely function. That cuts the player pool in half (or thirds), making it harder to get games started at all. Back in 1995, Gamespy with C&C hosted thousands of players and every game launched. FAF should be able to do the same. If that means scrapping the current setup and starting over, then do it.

So:

Give new players a sandbox before throwing them into the general population.

Put a timer on “foe” lists so the community doesn’t shrink.

Fix international game play, even if that means rebuilding the infrastructure.

Finally, maybe turn on a light-touch abusive language filter. Just enough to remind players not to cross the line during games—but not so strict that “bad bunny” gets flagged as a racial slur. lol

[IndexLibrorum]

@TexasDeath Not sure why this is in this thread about statistics tbf. Might be better to put this in its own thread? I can move it for you, if you'd like.

[TexasDeath]

This post is deleted!

[IndexLibrorum]

As requested in the discord, an overview of the highest and lowest ratings recorded. Ratings shown are calculated display rating (mean - 3x deviation).

Some rating manipulation has been used to reach the highest ratings (specifically for John_Doe, though lilSidlil did something similar as far as I know). The lowest ratings might be caused due to some bugs with the rating code.

For these reasons, the data is only somewhat useful, but it's kind of interesting so I've put it here for safekeeping.

Highest:

Username	Calculated Rating	Score Time (UTC)	Game ID
John_Doe	3685.206	2024-04-05 16:43:21	22215921
lilSidlil	3405.606	2022-11-27 20:46:14	18616894
Suzuji	3056.911	2019-06-24 19:39:30	9833239
derMarc117	3035.329	2012-04-08 01:40:24	54992
TheWeakie	2898.142	2025-09-21 22:00:12	25636874
Tagada	2775.317	2024-12-25 21:18:21	23926136
galacticfear	2770.694	2015-03-23 20:02:56	3240060
Nomander	2714.813	2025-10-25 23:24:32	25819059
Blodir	2653.693	2023-07-07 20:20:12	20327385
Terarii	2652.204	2025-10-24 19:08:43	25810813

Lowest:

Username	Calculated Rating	Score Time (UTC)	Game ID
RedDevil	-1936.952	2016-02-13 10:58:05	4428632
Valdez	-1630.885	2016-02-16 20:02:13	4443231
QualityDamage	-1582.118	2016-02-21 02:43:40	4459290
ViolentNoob	-1580.698	2016-02-22 16:52:56	4465062
MaxKofman	-1526.452	2012-11-23 18:23:24	444994
HodenGips	-1410.853	2016-02-17 13:00:10	4445214
Mikepara	-1388.161	2016-02-20 03:17:30	4454587
Vunt	-1356.464	2016-02-22 17:25:16	4465378
El_Ombligo	-1352.501	2012-04-03 03:09:10	47479
Sweaty	-1338.874	2016-02-29 21:54:50	4492828

[Defiant]

Interesting. Is there some context around this?

[JaggedAppliance]

There were major server changes in February 2016 which led to rating bugs so that's why you see almost all the lowest ratings being recorded then.

[derMarc117]

@IndexLibrorum Hey man, that's pretty interesting. I've just come across this post when looking up my old gamer tag.

Not sure if I can help you with that in any way or offer some input that might help. It has been about 13 years since I've low-key religiously played FA.
I recall that back in the day I was very fond of playing those smaller 5x5 km maps a lot, because I really enjoyed that T1 tank spam and ACU overcharge micromanagement.
Perhaps I was doing pretty good with that.

Are those ratings ranked or custom, does that even differ?

Have a good one guys. Maybe I'll have to turn on FA once again.

[derMarc117]

Okay I have been digging in my old drop box account. And found some rather dated screenshots.

They are the real deal though, Windows XP, ICQ in the system tray and even a date on two of them: 9th and 10th of April 2012.

FA Forever in version 0.7.511.

Check these ancient relics:

[Amygdala]

@derMarc117 Nice one! A bunch of those players are still hanging around from time to time. @zlo