Search
  • Matt Kimball

COD League Stage 2 Data Analysis

Updated: Apr 25

By, Matthew Kimball, Zach Rabuffo

Last modified: 5/3/21


Research Question

Before diving into Call of Duty League (CODLeague) data, one must fully understand how the league is structured. The CODLeague is an up and coming esports league (2nd year) that includes 12 teams and constructed with 4-player starting rosters. The tournaments are split up into 5 “stages” this 2021 season. Each stage is split up into 2 6-team groups (AKA divisions) where every team will play head-to-head within their groups (so 5 matches total). Those 5 games determine the seeding going into each “major” (AKA tournament). All matches are best-of-5 series in which 3 game modes are played in order: Hardpoint (HP) (maps 1 and 4*), Search & Destroy (S&D) (maps 2 and 5*), and Control (map 3). The grand finale of each major is a longer, best-of-9 series, where there are 3 HP’s, 4 S&D’s, and 2 Control maps.


Looking more into CODLeague data from the most recent Stage 2 and major, there have been many misconceptions between analysts and fans about overall player/team data as well as roster changes. All but 3 teams have made a roster change since the start of the season several months ago. Why is that? These unusual amounts of roster changes were very hard to understand amongst fans because many of the players had a good kill/death (K/D) ratio average. This is the most common statistics seen on broadcast, but there are many flaws with it and it has been debunked by many analysts (Mitsche, 2020).Considering all of this, we want to 1st assess which hardpoint statistics better determine team wins and losses. Then, we want to analyze specific player statistics in hardpoint and assess some of the many roster changes that occurred during stage 2.


Analysis/Findings


Impactful Statistics that Determine wins and Losses (Stage 2 and Major)

Methodology


Using hardpoint stage 2 and major data from the “CDL2021 Yr 2 Team Stats by @IOUTurtle dataset, winning percentages (in Hardpoint game modes only) are used to determine the correlation between winning and certain statistics. While Team Hill % seems as though it would be the most useful statistic (because it shows the amount of time a player is on the Hill, or hardpoint location, which is how you get points), our data showed no clear correlation between that and winning. Rather, we found that 2 other statistics Damage per Death (Damage pD) and Engagement per 10 min. (Eng p10) were much better indicators of a team’s success. There are a total of 11 teams included in this data. After finding the win percentages of each team, the positive correlation of these statistics show how they may be the driving force behind wins, rather than Kill/Death ratio.


For context, Damage pD looks at the amount of shots, and thus damage (from 150 health), a player puts into other players per life. Eng p10 assesses player speed by seeing the number of gunfights (or engagements) a player gets on a per 10 minutes. As described earlier, K/D is a very misleading stat.


Analysis

While the average Damage pD is not the only statistic which should be used to determine winning probability, in Hardpoint, it proves to be the most dominating factor

  • The main outlier in this group, FaZe (217.29), is nearly 20 points higher than the next team (Ultra). To put into perspective how big of a gap this is, every other team besides 1 (Thieves) is within 20 points of Ultra. This clear advantage in this category over the rest of the teams is shown heavily once looking at the winning percentage as well. Faze stands with an 85%-win rate in Hardpoint, almost 30% higher than the next teams (Subliners and Surge). While this may seem like an outlier, the consistency of this stat proves that they were clearly the better team in this game mode, and their dominating win percentage amplifies this.

  • The next group of teams with above-average winning percentages (Subliners, Surge, Empire, and Optic) are ranked 9, 4, 3, and 6 respectively in Damage pD. Excluding faze, ¾ of these teams are in the top half of the league in this category, with Subliners as the only exception (who lead the league in Eng p10).

While this statistic alone is much less indicative of how a team performs in the game, our data shows that the combination of this along with Eng p10 can create clear assumptions and relation to a team’s winning percentage.

  • According to the below scatter plot detailing how these 2 statistics interact with each other, there are 6 teams which are considered among the best in one of, if not both of the statistics. This is shown as the mediocre teams (by winning %) mostly find themselves located in the middle of the graph, unable to separate from the pack.

  • The 3 teams with the highest winning percentage (Faze, Subliners, and Surge) all find themselves as outliers in the scatter plot, proving that excelling in these categories is correlated with winning (for the most part).

  • The bottom group of teams (Legion, Royal Ravens, and Guerillas) have performed the worst (in terms of W%) yet are also ⅗ of the lowest teams in terms of Damage pD. However, unlike the other 2 teams (out of the 5) with a low Damage pD, these 3 also have average-below average scores for Eng p10.



Evaluating Roster moves using Hardpoint Player Data (Stage 2 and Major)

Methodology


Using hardpoint stage 2 and major data from the CodLeagueStats dataset, statistics such as damage per death (Damage pD), engagements per 10 minutes (Eng p10), and KD+ were used to analyze 51 different players. This analysis focuses on the 7 players who were either “benched” during or after the stage 2 season. These statistics were copied into an excel spreadsheet and analyzed using PivotTables, charts, and different formulas. After looking into several stat comparisons, the best ways to understand player pace and overall impact was by comparing Damage pD with Eng p10 as well as K/D+ with Damage pD.


For context, we made a stat called KD+ that scales the data to 100 points and assesses the league average K/D given the player’s role (AR/SMG/Flex). The formula for this is 100(Individual K/D / League avg. role K/D). For example, a player named Simp has the highest KD+ at 131.31. This means that he’s 31.31% above the league average within his SMG role.


Analysis


Starting with players benched during stage 2, 3 veteran players by the names of Temp, Slasher, and Majormaniak were all replaced by younger talent:

  • Temp, an AR player who played for 100 Thieves, averaged 45.42 eng p10, 161.11 damage pD, and a 78.43 KD+ in hardpoint. He ranked 5th worst in eng p10 and in KD+. Compared to other AR players that he’s matching up against, he simply isn’t as impactful in slaying. This caused the 100 Thieves team to bench him mid-season, and it resulted in a solid top 6 finish in the major.

  • Another AR player on this 100 Thieves team, Slasher, had much more formidable stats compared to Temp. He had a 96.08 KD+, 181.74 damage pD, and 47.24 eng p10 in hardpoint. Referencing the scatter plot below, one can conclude that he was playing way too fast for his AR role. He is 1 of 2 players ranked in the top 10 for eng p10 while also ranking in the bottom 1/3rd in damage pD. A culmination of playing too fast while also lacking a synergistic pace with Temp led to both of these players being benched.

  • Majormaniak is another AR player who was benched mid-stage by the Rokkr because, unlike Slasher, he was playing way too slow which hindered team performance. In hardpoint, he averaged a 100.98 KD+, 206.32 damage pD, and 38.20 eng p10. He clearly had good slaying numbers as he had comparable stats with another top AR in this esport, Octane. However, one can conclude that he was playing way too slow compared to his team. The average engagements p10 in the AR role is 42.85, so he was severely underperforming in this aspect.

Looking at players benched after stage 2, 4 players by the names of Fire, Slacked, Dylan, and Vivid were replaced:

  • Fire was a FLEX player who played for the Legion. The team didn’t perform to expectations, and he was the only player that was benched after stage 2. In hardpoint, he averaged an 81.72 KD+, 155.96 damage pD, and 38.46 eng p10. He ranked as the worst or 3rd worst in all of these statistical categories. It's very justified and smart of Legion to bench him because his impact and overall pace just wasn’t there especially considering he had to play an SMG role sometimes.

  • Slacked, an SMG player for Mutineers, averaged a 90.91 KD+, 171.03 damage pD, and 39.60 eng p10. He’s another player who lacked in all of these statistical categories. Amongst averages by the SMG role, he should have about 10 more damage pD and 5 more eng p10 to be considered an “average player.” Overall, he lacked the pace and slaying numbers alongside his team during stage 2, leading to him on the bench.

  • Dylan was another SMG player who played for the Royal Ravens and he averaged a 95.96 KD+, 176.96 damage pD, and 48.72 eng p10 in hardpoint. Contrary to Slacked, he was playing at way too high of a pace as he ranked 2nd in eng p10. Although it’s great to have a fast SMG player, it's clear that he was way too fast at his role and the team needed a slower player to replace him.

  • Vivid, a player for the Guerillas, was the last SMG player to be benched after the stage 2 season. He averaged a 92.93 KD+, 176.51 damage pD, and 50.27 eng p10 in hardpoint. Just like Dylan, Vivid played at way too high of a pace as he ranked 1st in eng p10. This hindered the team’s overall performance because the rest of his team played at an avg pace based on avg eng p10 by roles. This roster change got the most flak from fans because he was regarded as the best player on the team. However, these statistics back up the benching decision.




Conclusions


In Hardpoint, the clearest indicator of a better winning percentage comes from a higher average team Damage pD. However, it would be irresponsible to consider this the only important metric. The difference between teams with average Damage pD was where they ranked in Eng p10. The Surge and Subliners both found themselves about average in terms of Damage pD but were the lowest and highest in Eng p10 compared to the rest of the league. This proves that, while the Damage pD is most important, Eng p10 can be useful for separating the teams as it relates more to pace and play style as well.


After evaluating Roster moves using Hardpoint Player Data (Stage 2 and Major), every player benching during and after stage 2 was justified using 3 different statistics: Eng p10, Damage pD, and KD+. These statistics further proved what many were seeing based on the “eye-test” of watching these players play. A fascinating trend from the 4 players benched after stage 2 is that they were all within the bottom 14 of 51 players in damage per death and had KD+’s of 96 or less. According to eng p10, coordination and overall pace is more important than ever in Call of Duty, especially in this past stage 2 and major.


Overall, there appears to be a small correlation between the first and second research questions in terms of how the teams averaged in hardpoint (based on the Damage Per Death vs. Engagements Per 10 Min (Team) scatter plot) vs. the 7 players who were benched on those respective teams. The 3 players who were benched on the Rokkr, Mutineers, or Legion respectively (Majormaniak, Slacked, and Fire) were benched due to being in the bottom 5 of eng p10. The Damage Per Death vs. Engagements Per 10 Min (Team) scatter plot showed that the teams these players were on hindered as a result of their poor performance (3 of these 5 teams ranked in the bottom 5 in eng p10).


Data Sources

Related Findings


Mitsche, Karsen. “Why K/D Ratio Is Deeply Flawed, and How We Can Make It Better.” Cod League Stats, 7 June 2020, codleaguestats.com/2020/05/03/example-post-3/.


  • This article helped us understand what’s wrong with the massively popular statistic, K/D ratio, in COD. Factors such as the role a player is playing, the amount of engagements there in, and the amount of objective a player plays all impact a players K/D. It also helped us get the ball rolling in terms of grasping more of an innovative line-of-thinking about other accurate statistics that can assess player impact.

Krez, JP. “The Inclusion of Damage to Black Ops 4 Stats.” CoDStats.gg, 6 December 2018, codstats.gg/articles/the-inclusion-of-damage-to-black-ops-4-stats.


  • This article talked about the inclusion of the damage statistic to COD and how new metrics were formed based on this stat. This stat has now been seen in every COD title since it was revealed in 2018. Once again, JP helped us understand more about the stat and statistics that can be made from it. If we didn’t read this article, we wouldn’t have delved as deeply into damage per death.

Sources


Created our own dataset in Excel. The data came from CodLeagueStats, CDL2021 Yr 2 Team Stats by @IOUTurtle, and the Official Call of Duty League Website. The CodLeagueStats data included all player data based on event and weapon type (or role). We filtered it and narrowed it down to stage 2 and the major in the hardpoint game mode. The CDL2021 Yr 2 Team Stats by @IOUTurtle included all match data up to this point. We only looked at stage 2 and major matches from this dataset (in the worksheet tab titled “Tracker”). Lastly, the Call of Duty League Website gave us the team standings, and we filtered it to just stage 2.

19 views0 comments