Sorry I'm dumb. Fixed

I like your point regarding subsets. I've worked up top-10 peak ranks by region and class.

Top 10 Peak Rank by Region:

https://raw.githubusercontent.com/alex-p-6/TF2_Skill_Engine/refs/heads/main/Images/Region%20Top%2010%20Peak.png

Top 10 Peak Rank by Class:

https://raw.githubusercontent.com/alex-p-6/TF2_Skill_Engine/refs/heads/main/Images/Class%20Top%2010%20Peak.png

To look at when scout got an edge likely requires more match results or maybe some fresh eyes. I struggled to pull out a decent way to represent role power over time with the match data available.

You are very right, ranking systems such as this can only go so far but still brings out some interesting findings.

I built a TF2 6s Premier/Invite skill model based on the matches hosted on Liquidpedia. I thought some of the results were interesting so I'm sharing it in this post. You can access the CSVs and python code on github here.

Data: I extracted 4,373 games from Liquidpedia from 2008 to 2024 (not including the current in-progress seasons) at Premier/Invite level played between international LANs, regional LANS, ETF2L, RGL, ESEA and Ozfortress seasons. (basically every S and A tier event from here: https://liquipedia.net/teamfortress/Portal:Tournaments, not counting AsiaFortress and SA TF2).

There may be some cases where mercs/subs were used but given the amount of match data I haven't had time to check each one. There were a lot of seasons where regular season data was missing, but what is there is a good start. Here is a representation of the spread of games extracted for the model, by year:

https://raw.githubusercontent.com/alex-p-6/TF2_Skill_Engine/refs/heads/main/Images/Distribution%20of%20Matches%20by%20Year.png

Ranking Skill: I applied two methods of ranking player skill, an Elo method (info here) and an OpenSkill method (info here). The elo method had an 80.55% accuracy in predicting match results whilst the OpenSkill method had an 80.86% match prediction accuracy. It is pretty close, but given the OpenSkill method was slightly better, i'll use the results from that model for the remaining post.

I ranked players based on their player id/class combo. So there are many players who are listed multiple times on different classes.

A note about OpenSkill - the model assigns each player an average rating and a confidence interval. As more games are played, the player's average rating is adjusted whilst the confidence interval gets smaller. These two figures that represents a player's rating can be combined for an Ordinal Skill, used for ranking.

The full rankings are on Github, but here is the overall top 20 current ordinal skill rankings from the model:

https://raw.githubusercontent.com/alex-p-6/TF2_Skill_Engine/refs/heads/main/Images/Current%20Skill%20-%20Top%2020.png

And here is the top 20 peak ordinal skill rankings from all time:

https://raw.githubusercontent.com/alex-p-6/TF2_Skill_Engine/refs/heads/main/Images/Peak%20Skill%20-%20Top%2020.png

I definitely think some of these ratings are due to the match data available, but insightful to see what made the list. The full list is available here. I included, as best as i could, region codes for each player so you can filter on that.

Role Power: I found this discussion from 4yrs ago interesting around role influence in TF2 6s outcomes and I have used this model to have a go at a more data-informed answer. For each match in the dataset and using the current skill ratings at the time the match was played, i used the model to predict match outcome based only on a single role vs role match-up (i.e. combo scout vs combo scout). Here are the results, in priority order:

• combo scout match-up predicted 72.4088% of matches
  
• pocket soldier match-up predicted 71.7081% of matches
  
• demo match-up predicted 71.6115% of matches
  
• medic match-up predicted 71.4182% of matches
  
• flank scout match-up predicted 70.8142% of matches
  
• roamer soldier match-up predicted 70.8142% of matches

I don't think huge surprises that combo scout and demo are up there, but i was surprised by the high pocket influence. This could perhaps be skewing from games in earlier metas that were more pocket-centric. I did try to graph role-influence over time (year-by-year) but it didn't come out well so i haven't included in this post.

Best Teams of All Time: Since i found this video interesting ranking all-time TF2 teams, I also used the model to combine individual player ratings to define an overall team rating. Here are the top-20 of all time:

https://raw.githubusercontent.com/alex-p-6/TF2_Skill_Engine/refs/heads/main/Images/Team%20Skill%20-%20Top%2020.png

End: That is all i have to share today. If anyone has access to more match results than what is available on Liquidpedia from ETF2L, ESEA, RGL and Ozfortress I'd be interested to include them in for a potential version 2 ranking. Please get in touch.

Otherwise, interested to hear what people think.

maybe you wont think that come the end of the season ^_^

yeah that sounds pretty accurate, i think teams will need to have strong teamplay to counter the big dm of the asia fortress players. Its really hard to say how they will perform against the other top teams because we have seen some super aggressive/super good dm teams do very well in the past and also some do not-so-well.

I think the team representing asia fortress is pretty set on going. Some of the Western Australian players have scrimmed some 6s matches against asia fortress in mix-teams. This included some of their players that intend to head to i58. They have a very different play style from any other region and some extremely talented players, so it will be exciting to see how they fare against the rest of the world.

listen to your breathing

in the same boat as you mate

have you tried all the ports? dvi/vga/hdmi? is the fan spinning?

Hi, i was looking for the demos of the experiment from season 10 if possible. The whole season would be great, or any of these games:

http://play.esea.net/index.php?s=stats&d=match&id=2460907
http://play.esea.net/index.php?s=stats&d=match&id=2442804
http://play.esea.net/index.php?s=stats&d=match&id=2532387
http://play.esea.net/index.php?s=stats&d=match&id=2492574

Thanks!

if you have mihalys flow regular season games that would be awesome. They played in esea season 11.

GeaR, Aporia, Blaze and A_Seagull are probably the best roamers. They seem to understand the game well and make the "correct" decision based on the situation. This seems to stem a lot from the style that relic used to play on mix^. Not always the most creative or flashy plays, but they can be relied on and they get the job done.

I've felt that ww- played a more aggressive style of roaming, where a lot of the time he would force the enemy into awkward situations with his strong mechanics and fast playstyle. He always seemed very comfortable on the flank without support and always willing to take a 1v1. He was much more likely to be proactive in situations, but this resulted in a lot of needless deaths too. You could see this when he played for the experiment in season 10. Lots of quite amazing plays and also lots of failing plays.

Harbleu also plays a very similar style to ww-, and was probably one of the reasons why mihalys flow was so dominant in season 11 (not counting chokes on lan).

anyways, just my 2c

Heres some movies i thought were great:

Lives of others
American Beauty
How to train your dragon
midnight in paris
city of god
before midnight
ghost world
kiss kiss bang bang
short term 12
the man from earth
the perks of being a wallflower
the place beyond the pines
50 50

maybe these are good, idno :S

https://dl.dropboxusercontent.com/u/81478998/gnat_badlands_56951.zip
https://dl.dropboxusercontent.com/u/81478998/gnat2_granary_1751.zip

the owl competition in Australian tf2 uses halves like esea and plays 2 maps per round (each round lasts 2 weeks). Both maps are first to 5, or 4 in the case of koth. We have had other competitions that only allow a 30 minute time limit and are first to 5 however.

The 5 round difference rule could create some interesting situations for teams and make the game more exciting, giving the losing team a chance to make a comeback. However, in a 30 minute time limit, getting more than 6 or 7 rounds played is very unlikely and the team that pulls ahead could just turtle to win.