CFCL Analysis - By the Numbers

Hey guys! I'm a big stats guy, so I wanted to highlight the numbers so far in CFCL Elite. (https://imgur.com/a/kgMEaX1 for full size)

[IMG2=JSON]{"data-align":"none","data-size":"full","src":"https:\/\/i.imgur.com\/GbTknjT.png"}[/IMG2]

First, the ELO of teams. It was determined by head-to-head results with teams starting out at 2100 ELO. The K factor was 100 in 2-0 victories and ties, but 50 in all other circumstances (including DQs). Once a team is above 2350 ranking, their matches have a K factor of 40 (this discourages massive shifts due to big upsets). Players added in from Open League will begin with the ELO of the lowest-rated team in top 8.

The final numbers:

ruLegends
2446


Carbon
2320


Penta
2308


getBrikt
2273


Flipside Tactics
2133


Mindplay
2111


Final Gaming
2107


Instinct
2063


LCK / ANV
2024


Rawness
1884


Wafflemakers
1874


FuryOus
1773


(removed after month 1)

Team Cursive
1902


XQ Intense
1894


So let's look at these. Foremost, ruLegends has been absolutely dominant, and their ELO reflects it. The only team above 2400 has only had two matches (!) result in a loss of ELO this entire season.

Carbon made a massive surge toward the end, for the better part of of the ranking period resting as the 4th / 5th ranked team in the event. However, closing out last month with a 7-0 finish secured their spot as the #2 rated team in the league.

Penta has held onto their #2 rated placement for the majority of the cup, only briefly falling behind getBrikt in the end of July (then again dipping barely below Carbon's final surge). The exhibited monumental consistency, and close out the period as the #3 team in the league.

getBrikt started off incredibly poorly, falling as low as the 8th rated team in the league in the first few weeks. However, after that period, getBrikt had consistently strong performances, peaking briefly as the 2nd rated team in the cup, before falling behind Penta & Carbon in the final pull.

Mindplay have been consistent, though not impressive. Mindplay remained within the same 90 ELO range through most of the ranking period, with few upsets in either direction (aside from a draw with getBrikt). It will be interesting to see whether or not the new additions and change to "Teamw0w" will change this trend moving forward.

FinalGaming peaked briefly at #3, then crushingly fell through a series of DQ losses to the second-worst rated team. They finished fairly strong, raising their ELO up significantly with a series of 2-0 victories, and barely sc****d by into the top 8, securing their spot in August.

Instinct makes a strong case as the team-to-watch. While they were consistently one of the lowest-rated teams throughout both months, they made an aggressive push at the end, rising more than 200 ELO in only a handful of games.

Two of the sadder stories are Flipside Tactics and Rawness. Flipside started off incredibly strongly, initially taking the lead then following closely behind Penta at #3. However, an abysmal finish to the last month resulted in a more than 150 ELO drop. For Rawness, the fall was even more pronounced - briefly peaking as the #6 ranked team after a solid June circuit, Rawness ultimately would fall nearly 300 ELO points to not make the August cut.


A side effect of this analysis is that we now can (fairly accurately) gauge the likelihood of upsets, and track the relative win % chances of teams to one another.

[IMG2=JSON]{"data-align":"none","data-size":"full","src":"https:\/\/i.imgur.com\/7S9xkyi.png"}[/IMG2]


ruLegends
Carbon
Penta
getBrikt
Flipside Tactics
Mindplay
Final Gaming
Instinct


ruLegends

67%
69%
73%
86%
87%
88%
90%


Carbon
33%

52%
57%
75%
77%
77%
81%


Penta
31%
48%

55%
73%
76%
76%
80%


getBrikt
27%
43%
45%

69%
72%
72%
77%


Flipside Tactics
14%
25%
27%
28%

53%
54%
60%


Mindplay
13%
23%
24%
28%
46%

51%
57%


Final Gaming
12%
23%
24%
28%
46%
49%

56%


Instinct
10%
19%
20%
23%
40%
43%
44%




By these numbers, ruLegends' dominance is even more apparent. Their closest competition, Carbon, only has a 33% chance of winning a game.

By these numbers as well, a clear "4 tiers" are obvious, with players from each tier coalescing around certain points.

Tier 1: ruLegends - class of their own, undisputed best team in our event at the moment.
Tier 2: Carbon, Penta, getBrikt - they are, statistically, further from the 5th ranked player than ruLegends is from them.
Tier 3: 5th-8th. These teams are, again, fairly obviously above their competition outside of the top 8. However, they clearly fall short of the top 4 echelon.
Tier 4: everyone else. These guys have a large mountain to climb before they can really be considered top or even mid level threats.

On a side note: while Instinct's odds of beating top teams are incredibly slim, their chances against the mid-tiers have been steadily rising, while most of the mid-tiers are on the decline (save for Mindplay / teamw0w, which has remained consistent).

Anyway, I hope you guys enjoyed reading it, I thought the narrative provided by these was really interesting.

Comments

  • How can Mindplay be over us "Final Gaming" when we win them easy? How you rate this elo? Penta played with ringer, we played with ringer u can´t make any analytics from teams who are playing with ringer
  • So only one team can be tier 1? Guess the international team for NA is not a tier 1 team. I'm pretty sure every team that goes international should be a tier 1 team.
  • nakin1234 wrote: »
    How can Mindplay be over us "Final Gaming" when we win them easy? How you rate this elo? Penta played with ringer, we played with ringer u can´t make any analytics from teams who are playing with ringer

    The ranking is by total results only. Everyone started at 2100 ELO evenly, with the K-score adjusted by the strength of the win (2-0, 1-1, DQ, etc.). It does not take into account substitute players; if teams want to win more / raise their ranking, they should play with their s5 more. Your team had several forfeit losses last month, and these lower your ranking as well.

    I don't really get how you can say you expect to beat them easily, though. Your last match against them, they had 2 substitute players. You won 10-5, and lost 10-6, clutching out the map by a single round. "Easy"?

    9902321412 wrote: »
    So only one team can be tier 1? Guess the international team for NA is not a tier 1 team. I'm pretty sure every team that goes international should be a tier 1 team.


    This is by the numbers, taking an objective look at their performance. ruLegends are twice as likely to win as they are to lose against their nearest competition, and they are clearly in a tier of their own. For context, they are roughly as likely to lose to Carbon as Penta is to Final Gaming.

    When more teams join that level of dominance, they can be additions to the top tier. However, as of now, there is a clear gap between ruLegends and the rest of the field.
  • On another note, here's how the ELO rankings compare to the actual rankings of the ultimate finishing placements of the teams last month. Should be noted that the 3/4 players in the middle were the ones most subject to upsets, with their ELO being close to 50% matchups. This means that Instinct had some key upsets to place over Final Gaming and Mindplay, but that the rest of the matches have played out more-or-less how we can expect.

    .

    ruLegends
    2446

    ruLegends
    0


    Carbon
    2320

    Carbon
    0


    Penta
    2308

    Penta
    0


    getBrikt
    2273

    getBrikt
    0


    Flipside Tactics
    2133

    Flipside Tactics
    0


    Mindplay
    2111

    Instinct
    +2


    Final Gaming
    2107

    Mindplay
    -1


    Instinct
    2063

    Final Gaming
    -1


    LCK / ANV
    2024

    LCK
    0


    Rawness
    1884

    Rawness
    0


    Wafflemakers
    1874

    Wafflemakers
    0


    FuryOus
    1773

    Furyous
    0




    And for the first month. The first month WOULD have been more accurate, with Flipside being at third and Final Gaming being 1 position lower. However, Flipside took a DQ loss to Final Gaming due to an illegal item.

    .

    ruLegends
    2309

    ruLegends
    0


    Penta
    2276

    Penta
    0


    Flipside Tactics
    2193

    Carbon
    +1


    Carbon
    2183

    getBrikt
    +1


    getBrikt
    2162

    Flipside Tactics
    -2


    Rawness
    2102

    Final Gaming
    +1


    Final Gaming
    2096

    Mindplay
    +1


    Mindplay
    2084

    Rawness
    -2


    ANV Gaming
    2033

    ANV Gaming
    0


    Instinct
    1941

    Instinct
    0


    Team Cursive
    1902

    Team Cursive
    0


    XQ Intense
    1894

    XQ Intense
    0


    .

    Here are the results if that DQ loss were not administered.
    .


    ruLegends
    2309

    ruLegends
    0


    Penta
    2276

    Penta
    0


    Flipside Tactics
    2193

    Flipside Tactics
    0


    Carbon
    2183

    Carbon
    0


    getBrikt
    2162

    getBrikt
    0


    Rawness
    2102

    Mindplay
    +2


    Final Gaming
    2096

    Final Gaming
    0


    Mindplay
    2084

    Rawness
    -2


    ANV Gaming
    2033

    ANV Gaming
    0


    Instinct
    1941

    Instinct
    0


    Team Cursive
    1902

    Team Cursive
    0


    XQ Intense
    1894

    XQ Intense
    0


    .

    So all in all, the ELO tracking system is pairing exactly with what we would expect from results. Interested to see where it goes!
  • Interesting largely because it tells a story of consistency - outside of the DQ win shuffling things up month 1, there have only been 2 real upsets each month, with Rawness / Mindplay in month 1, and Instinct over-performing in month 2.

    Storylinez
  • EX0T1C wrote: »
    On another note, here's how the ELO rankings compare to the actual rankings of the ultimate finishing placements of the teams last month. Should be noted that the 3/4 players in the middle were the ones most subject to upsets, with their ELO being close to 50% matchups. This means that Instinct had some key upsets to place over Final Gaming and Mindplay, but that the rest of the matches have played out more-or-less how we can expect.

    .

    ruLegends
    2446

    ruLegends
    0


    Carbon
    2320

    Carbon
    0


    Penta
    2308

    Penta
    0


    getBrikt
    2273

    getBrikt
    0


    Flipside Tactics
    2133

    Flipside Tactics
    0


    Mindplay
    2111

    Instinct
    +2


    Final Gaming
    2107

    Mindplay
    -1


    Instinct
    2063

    Final Gaming
    -1


    LCK / ANV
    2024

    LCK
    0


    Rawness
    1884

    Rawness
    0


    Wafflemakers
    1874

    Wafflemakers
    0


    FuryOus
    1773

    Furyous
    0




    And for the first month. The first month WOULD have been more accurate, with Flipside being at third and Final Gaming being 1 position lower. However, Flipside took a DQ loss to Final Gaming due to an illegal item.

    .

    ruLegends
    2309

    ruLegends
    0


    Penta
    2276

    Penta
    0


    Flipside Tactics
    2193

    Carbon
    +1


    Carbon
    2183

    getBrikt
    +1


    getBrikt
    2162

    Flipside Tactics
    -2


    Rawness
    2102

    Final Gaming
    +1


    Final Gaming
    2096

    Mindplay
    +1


    Mindplay
    2084

    Rawness
    -2


    ANV Gaming
    2033

    ANV Gaming
    0


    Instinct
    1941

    Instinct
    0


    Team Cursive
    1902

    Team Cursive
    0


    XQ Intense
    1894

    XQ Intense
    0


    .

    Here are the results if that DQ loss were not administered.
    .


    ruLegends
    2309

    ruLegends
    0


    Penta
    2276

    Penta
    0


    Flipside Tactics
    2193

    Flipside Tactics
    0


    Carbon
    2183

    Carbon
    0


    getBrikt
    2162

    getBrikt
    0


    Rawness
    2102

    Mindplay
    +2


    Final Gaming
    2096

    Final Gaming
    0


    Mindplay
    2084

    Rawness
    -2


    ANV Gaming
    2033

    ANV Gaming
    0


    Instinct
    1941

    Instinct
    0


    Team Cursive
    1902

    Team Cursive
    0


    XQ Intense
    1894

    XQ Intense
    0


    .

    So all in all, the ELO tracking system is pairing exactly with what we would expect from results. Interested to see where it goes!

    penta won Month 1 you do realize that right
  • Did they? It doesn't change the rankings, 'cause those are head-to-head records, but it would be an upset by the ELO. I thought ruLegends won month 1?
  • EX0T1C wrote: »
    Did they? It doesn't change the rankings, 'cause those are head-to-head records, but it would be an upset by the ELO. I thought ruLegends won month 1?

    Both had the same amount of wins as they do with losses too but Penta had a round advantage in month 1 so that's how they came out on top of ruLegends ...
  • Interesting, then. So three upsets in the entire event, by the ELO, so far. Again, that doesn't change any of the rankings, since it's by head-to-head, but it's good context to keep in mind.
  • We won the first Month because we won against ruL , head-to-head won us the Month. And between ruL and all of the others isnt a clear gap bro :) We also smashed them in the first month and we didnt even need to play serious , and 2nd month we had a ringer that didnt touch the game for 6months.
  • [USER="18807299"]EX0T1C[/USER] what are you doing with your life bro ?
  • bAZiNGAK93 wrote: »
    We won the first Month because we won against ruL , head-to-head won us the Month. And between ruL and all of the others isnt a clear gap bro :) We also smashed them in the first month and we didnt even need to play serious , and 2nd month we had a ringer that didnt touch the game for 6months.

    The beautiful thing about numbers is that perception doesn't matter.

    You give us an excellent example of how bias can happen: you did not smash them the first month of CFCL. In fact, you barely beat them - you went to overtime, and had a narrow 10-7 win. You won by an average of 2.5 rounds per map - you'd need to win by nearly twice as much for it to be a blowout. Yet you feel you "smashed" them. Even if you were purely trolling, that's an undeniably close finish. Their performance in month 2 is a better example of what smashed actually means - a brutal 20-10 over your team. But, because of your personal perception, you feel that it was a "blowout" month 1, and that month 2 should somehow not effect your team's ranking (as if teams in any sport discount rankings just because of substitute players).

    That's why I decided to show my ELO rankings this time around. You guys let personal bias influence your perception of teams way too much. It really makes me doubt your perception on other things (top players, teams, etc.)

    ___

    So I hate to break it to you, but here's the reality:

    ruLegends is roughly as far ahead of you as you are ahead of Mindplay. ruLegends has a 70% chance of beating you by the ELO.

    They have lost one match the entire event - your team has lost 3. Your head-to-head favors them - they beat you by a much wider margin than you beat them. They have beaten higher rated teams than you have, by a wider margin than you have (your record against Carbon is 1-3 in maps, ruLegends' is 3-1). They consistently do better than you against top teams, and are winning your head to head.

    Hence their higher ELO, and hence them being in a tier above you - above all of us.

    You can use excuses, "Oh we had a ringer", "oh we didn't play well", "oh Snox had to go", but that's self-serving bias. You attribute your failures to external, and you attribute your success to internal. Thus, your super close win over ruLegends becomes "we smashed them", while your blowout loss to them is "oh well we had a ringer that would never happen in a serious game". Ultimately, the only thing that matters is the final result.

    ___

    I know that's a bit brutal, maybe too brutal, but like...I don't get your argument here. They're ranked way higher than you, by a lot, and they absolutely deserve to be. Don't like it? Start winning, and keep winning.


    @Grecc I work in esports regulation / data world, and play CF ~20 hours a week on the side. That's my life xD
  • I didnt say anything about them being top1 , its just funny how u say they are far ahead of others , because they won 2 matches more than us (1-1 between us as a teams and 1 bo2 with a stand in that didnt play for 6months, dont tell me it doesnt affect a team, i wanna see how u beat a topteam without finaldown as an example , it gets much harder). Also , just watch the Game on Stream (ofc smashed is a bit harsh , but we purely trolled and won 23-18 or something like that) . Just sounds stupid to me call a team far ahead of others , when its a league with several topteams that can beat each other.

    Or you seriously wanna tell me that we can't beat them rn because they are far ahead just for the reason that we threw a single game against INSTINCT (other losses were against ruL , Carbon first month)?
    And why you count maps , when all that matters if you beat a team? We are 1-1 against ruL and 1-1 against carbon in bo2 :)

    The only thing that stands between ruL and other topteams is that they are more consistent against lower teams. Otherwise i can see Carbon , FG , PENTA beating them.


    Maybe you're right about certain things , but im pretty sure that they are not far ahead of the other topteams.
    Thats my only point, to let you know that they are not so far ahead than you think they are. And tbh i find it harder to play against Carbon or you guys , than playing against ruL.
  • Of course you -can- beat them. Anyone can. However, it is unlikely. It would be a big upset if ruLegends lost to anyone. Your team, specifically, has about a 31% chance to beat them on any given day. Carbon's odds are slightly better, at 33% (my team's odds are 27%, with the next-closest competitor, Flipside, only having a 14% chance of an upset). Upsets happen, and even if you beat Penta this month it probably isn't going to suddenly make you the favorite. That's not how ELO works.


    If I go by pure "feeling", I would agree with you. If I went by pure feeling, I would say that Carbon and Penta are roughly the same if not better than ruLegends. However, this is where bias creeps in. ruLegends has a dominant head-to-head over Carbon and getBrikt, with a lead in the head-to-head over Penta. Not only is ruLegends winning by a solid margin against each of their nearest competitors, they ALSO are crushing the lower level teams in a more consistent fashion and by a wider margin. "Being more consistent against lower teams" is a HUGE factor in how good a team is.


    I didn't subjectively decide they were so far ahead because of the 2 match gap between you, or because any of this. ELO is impartial - it's a math equation that is equal for every single team. Every team started at 2100. Winners have their ELO increased based on the ELO of their opponents, losers have their ELO decreased based on the ELO of their opponents. Over time, players or teams get stratified into their own baseline level (usually, within 20-25 games). And, by the numbers, ruLegends is in a class of their own.


    Yes, there may be other factors - you had a ringer one match, you trolled the other one, Carbon feels harder to play...but these don't really matter. Why? Because, ultimately, you don't know what ruLegends' side is either. Maybe they trolled against you? Maybe they were sick? They have a massive ping disadvantage compared to European players (and they're still winning btw in spite of having the worst ping-total of any non-Egyptian team), maybe that tilted them the first month? Maybe ruLegends had PC issues, or just a really bad day? It's human nature to look at other teams and assume they are playing at 100% of their ability 100% of the time, then to look at ourselves and see all of the external factors influencing the game. This is called the fundamental attribution error. So, your win becomes "oh man these guys suck, we trolled and smashed them" without considering things from their view. Maybe from their view it was: "ugh these guys didn't even take it seriously, that made us lose our composure. It won't happen again."


    ELO ignores all of the external factors, and focuses purely on results and the stats, because ultimately any team can make an infinite number of excuses. My team's losses to Carbon? "we were up 9-5 and they came back to win in OT, we started trolling, that's why we lost"...so? We lost. The next month, we lost again: "me and FinalDown had 100 ping on the NA server that day, we're our hardest fraggers"...so? We lost. Your team's excuses? They don't matter - you lost. ELO doesn't care - it tracks nothing other than how likely you are to win.

    As a side note, your ELO changes by a wider margin if you win or lose 2-0 than if you win or lose 1-1 (draws can change your ELO too). That's just to balance it out (basically it treats 2-0 wins as 2-0, and 1-1 wins as 2-1 for ELO purposes)
  • This is the prettiest graph I've ever seen on these forums. And I love it.