After months of R&D, “PEPS+” has finally been revealed just before PUBG Continental Series 1 (PCS1).
We are very pleased to see our efforts gain so much attention from the PUBG Esports scene and are happy fans get to see more stats about their favorite pros. We wanted to go in-depth about the PEPS+ systemwith you, and hope this dev letter can be an opportunity to share our team’s thoughts on the new stats system, which was developed to help you have a deeper understanding of PUBG Esports pro-players’ sophisticated tactics and strategies.
Development goals of PEPS+
Based on analysis of various official tournament data accumulated from 2017 Gamescom PUBG Invitational up until PGI Berlin, we devised Esports rulesets and metrics, named “S.U.P.E.R,” that can provide the data with a level of fairness and tempo suitable for the PUBG pro scene.
[Pic.1] The entire match graph of PGI 2018 used as the basis to develop S.U.P.E.R
With S.U.P.E.R, we designed general rules that determine the final winner of a tournament on the macro level and play guides within a match on the micro level to establish the foundation for professional-level PUBG Esports gameplay. Now that we had a baseline of what great play “looked like” in the data, we have been developing PEPS+ to achieve two goals:
We want to provide meaningful stats on professional players based on in-depth R&D on PUBG Esports gameplay.
Gameplay in PUBG Esports is completely different from that of any other genre of Esports. A massive number of players (16 teams, 64 players) are supposed to make countless decisions within a single match and their choices consistently affect each other with external influences of random item spawns and circles. This makes every session unfold differently, requiring teams to execute different tactics and strategies in every match,. even on the same map. PUBG even needs two different scoring mechanisms to fairly judge and score the end of a match: the kill point and placement point. To better understand the performance of battle royale Esports players, we came up with Player Stats by grouping kill-related stats into the Battle category and placement-related ones into the Strategy category to translate player performances into metrics and numbers.
We want to decipher situations and team performance that can’t really be explained with just primary stats to help our fans better enjoy PUBG Esports.
While doing our research, we were fortunate enough to meet lots of people interested in PUBG Esports, but quickly learned there are some information gaps going on between Esports experts and general players – Esports analysts and avid Esports fans know all about the core fun elements of PUBG Esports whereas normal users might not.
Football fans can better enjoy watching football games by understanding their favorite team’s strategy and tendency (e.g. formation, position of the key player, playing time management for starters, substitution timing, etc.). Likewise, we believe that PEPS+ will help you a lot to better understand your favorite PUBG Esports teams and players.
For example, most Esports teams tend to have their favorite landing spot or patterns (of course they may change plans depending on the plane path). This is one area that the data can help you understand each team’s strategy and better enjoy PUBG Esports. Here’s an example of two teams that take quite different approaches in general to the game:
Team Liquid in Europe favors landing on central areas (Pecado in Miramar and Pochinki in Erangel) and taking a tight formation making tactical moves after looting is completed. This can be viewed as their strategic judgment to take the risk of early engagements, since engagements tend to happen in the middle of the map, and get themselves better positioned in circles and concentrate their power on later/end game combat. And that strategy proved to be effective in the first week of PCS1 Europe.
Element Mystic in Asia, on the other hand, tends to land on southern Severny in Erangel and La Cobreria or north of Water Treatment Miramar as their landmarks. After looting is over, they take a 2-2 formation or all 4 players split out widely for tactical rotations. This data can be interpreted as a careful, defensive strategy, where they try to avoid early engagements, gear up safely until the next circle, and mutually communicate risk factors around them by observing broad areas of the map.
As you can see, we can derive a ton of information and assumptions about the teams by simply looking at the numbers provided by PEPS+. Also, we can further understand what kind of role a certain player is playing in the team by looking at the PEPS+ of the individual. So Player Types of PEPS+ aim to provide quantitative stats that can give PUBG Esports fans deeper insight on their favorite teams and players to better understand their decisions and strategies and more enjoy watching PUBG Esports.
Our thoughts on the potential risk of misunderstandings that we “grade” players with PEPS+
PEPS+ evaluates players on a scale of 0 to 100 in a relative manner, and we never intend to make arbitrary, subjective judgments about Esports players. PEPS+ is not a test score or a player score. It’s not intended, nor should be seen as, a “player’s number” or their “score.” It’s a relative look at a player’s performance based on the data we choose from events. We aren’t trying to score players like they’re video game sports characters; you can’t boil down a player’s total skill, ability, and potential into a number. But you CAN compare players against each other relatively, and that’s what PEPS+ does.
If Pio gets an overall 91 at the PCS1 and an overall 90 at PCS2, that doesn’t mean Pio went down from a 91 to a 90. It means that Pio’s performance at PCS2 was one point lower relative to everyone else’s performances at the tournament. Likewise, PEPS+ is designed to give you interesting insight on a player’ performance and to make relative measurements of their skills and decisions shown under a variety of circumstances, rather than evaluating and categorizing them into grades.
We also hope PEPS+ can help those players with brilliant performances in international tournaments receive the attention they deserve based on reasonable and factual stats.
Story about PEPS+
PEPS+ was developed by 3 PUBG members (Hyojun Je, Jeongho Lee and myself) who spend most of their time and effort studying PUBG Esports data. To achieve the two overarching goals explained above, we have opted for 8 stats out of 20 possible options for PEPS+ and selected a total of 8 type variables that can be further broken up into 256 (4^4) sub-types.
We will make efforts to make PEPS+ more reliable based on data accumulated from future international PUBG tournaments. In-game raw data stats (e.g. Kills, DBNOs, damage, time survived, position vector, etc.) are refined into primary statistics (means, min-max scaling, standard deviations, mutual ratios, etc.). Derived from our formula by factoring in those primary figures, Player Stats and Types are relative stats among the teams and players that played together in the same match. That means it’s extremely difficult for a player to get high scores in all stats. If a player’s stats exceeds 70 on average, that means the player showed an outstanding performance.
* Please note that the current stats are relative values within each region since there hasn’t been an global inter-region tournament so far in 2020 due to the COVID-19 pandemic.
1. Esports Player Stats – Radar Chart
Players Stats consist of 3 categories: Battle, Strategy, and Experience.
Each category has up to 5 variables that can explain the stat based on a unique formula.We came up with optimal weights for the stats through case simulations to make them more convincing.
First off, the Battle category. This is a performance stat closely related to kill points which is one of the decisive factors in PUBG Esports. Battle consists of 3 sub-stats as follows:
: This stat tells you about whether a player was able to knock down an enemy or secure a kill on an enemy he dealt damage to(here we will simply call knockdowns and kills combined “downs”), measuring the number of groggy cases, kills, and the ratio of downs to damage dealing cases in the entire tournament.
(Special info: A higher weight is applied to the down ratio than to the number of downs.)
: Firepower measures the most effective damage dealing cases of a player within a short period of time by factoring in the most damage dealt, most downs, and highest number of successful hits in a specific time period out of all matches.
(Special info: A larger weight is applied in the following order: Highest damage dealt > highest number of successful hits > most downs in a specific time period.)
: Combatability tells you about the successful combat engagements out of a player’s entire combat time by calculating the winning rate based on the damages exchanged with enemies (we will simply call it the “damage-winning rate”) and the winning rate based on the player’s downs and deaths (the “down-winning rate”).
(Special info: A higher weight is applied to the down-winning rate than to the damage-winning rate.)
Next up, the Strategy category. This stat is related to placement points, which is another decisive factor in PUBG Esports. It has the following 3 variables:
: Though Survival is calculated through the simplest way compared to other stats, it is difficult for players to get a high survival rating. It is measured by factoring in the time survived and the amount of health recovered over the match.
To get a higher Survival rating, a player has to achieve a high placement point total as well as surviving for a long time period. This is why the player who earns most points gets a relatively low Survival stat while showing outstanding performance in other stats because point earners are likely to be on the front line, putting themselves at risk. (e.g. Gen.G Pio and TSM_Fexx)
(Special Info. A higher weight is applied to the time survived than to the amount of health recovered.)
[Pic.2] As point earners of their team, GEN_Pio and TSM_Fexx got high combat-related stats by jumping into lots engagements and presenting good tournament performances but they got relatively less survival stats:
- Tactics (related to circle play)
: This is the stat that came from the deepest case study throughout our R&D efforts and gave us a lesson that an excessive data-oriented approach may end up distorting player stats.
The definition of Tactics used here is a stat that tells you about ‘how well a player performed in advantageous or disadvantageous circle positions.’
It factors in many variables like the time survived, downs, size of Safe Zone, distance from the center of Safe Zone, the number of players alive, and the number of players present within the square area between a player and the center of Safe Zone. Weights are applied to the time survived and downs in the Tactics formula.
(Special Info. A higher weight is applied to the time survived in various circles than to the number of downs.)
[Pic.3,4] The world map below shows the divisors applied to the players in relation to their distance from White Zone.
(A higher number shown on the map means the location gives more positioning advantages, so the player there needs to perform better to get a higher stat. / The actual figures used in the formula may not be the same as below.)
: This stat measures how well a player performed in terms of teamplay under various circumstances. It factors in counterattack/damage trading on an enemy who attacked teammates, team attacks on the same enemy target, and how many times a player revived teammates, with outliers removed.
(Special info: A higher weight is applied in the following order: Counterattack on the enemy who attacked teammates > team attacks on the same enemy target > the number of times a player revived teammates.)
The Experience category is designed to deliver a different type of information compared to the other stats above – We are planning to break it down into the following 2 sub-stats:
: Potential is based on the top n% of the matches where a player showed outstanding performance across a certain tournament. While Firepower represents performance in certain moments and Combatability explains overall performance across the tournament, Potential quantifies a performance based on the best-played matches of a player compared to their other matches.
: We couldn’t roll out Career in PCS1 but will introduce it after accumulating PEPS+ data later on.
Starting in 2020, we are planning to reflect player careers into their stats in a quantitative manner. As many of you already know, experience in large-scale tournaments is one of the most important elements to Esports players as well as sports players. The tournament atmosphere is different from regular competitive play; I’ve even competed in a small tournament but remember how nervous I was playing there! The Career stat won’t reflect the overall stat of a player. But we believe it can be a useful factor when you want to see if a certain player could continue to show outstanding performance fueled by experiences in large international tournaments and if the player should be in the PUBG Esports scene spotlight.
2. Esports Player Types – Horizontal Parameters
Player Types consist of 4 categories to help you understand the tactics and strategies of each Esports team and player:
1) Landing Position and Drop Timing to understand playstyle in early phases;
2) Team Density and Looting Variation to understand the strategy of a team or player;
3) Combat Distance and Warlike to understand the way a team or player engages in combat; and
4) Performance by Map (PBM) and Point Distribution (kill and placement points) to show the per-map performance of a team or player.
These Types of PEPS+ are based on actual data, which are quantitative results from the decision making of a team or player. It means Types can be used to derive insight on players’ actions through inductive reasoning. Similar to the Player Stats calculation, each Type is gauged between 2 contrasting parameters (100% in total) by normalizing all players’ play styles into 0 to 1 to show which player has a stronger tendency in that specific Type in relation to other players.
As developers of PEPS+, we had to check on the Types of each and every player and verify them among players of all skill levels. For PUBG Esports fans, however, we hope Types can serve as the grounds for storytelling about a certain team or player, unlike Stats that measure player performance in a quantitative manner.
Types will give you insight on a story behind a player’s action even if you don’t have background knowledge.
- Landing Position can tell you why the team or player decided to land on a certain landmark by observing whether they favored landing on the map center or outskirts.
- Drop Timing can explain which player tends to drop earlier or later in a team to monitor other teams’ landing spots.
(ex. Radiance’s Miccoy landed faster than any other teammates in 80% of the PCS Charity Showdown matches, presumably playing a role to check on enemies’ landing locations.)
- Team Density tells you about whether a team took a tight or loose formation in general across the tournament, giving you insight on their favorite movement tactics.
- Looting Variation gives you a hint at what type of items a player favored to pick up to prepare for later engagements.
- Combat Distance can tell you whether a player performed better in close-range combats or long-range engagements.
- Warlike helps you understand how frequently a player attempted to attack enemies compared to other teams and players.
- PBM shows on which map a team or player performed better between Erangel and Miramar.
- Point Distribution tells you whether a player contributed to the team more with kill points or placement points.
We hope the comprehensive figures provided through Types can help you better understand the playstyles of over 64 players presented in the entire matches and the gameplay decisions that your favorite team or player made as well as the results that followed.
Message to PUBG Esports fans
During the development of PEPS+ and our efforts to make data more accessible, robust, and give deeper analysis into PUBG Esports, we have always been careful to minimize potential misunderstandings that might be caused from rating real-life players, not in-game characters, with quantitative stats that we designed.
This is why PEPS+ rating tends to be lower than other commonly used stats. PEPS+ is designed to make relative assessments among the top-skilled Esports players based on a variety of Stat dimensions.
Once again, just over 70 in a certain stat is high enough to show that the player presented outstanding performance in that category. As for the overall stat, just exceeding the average (which we set to 50) means that the player performed well. We hope you understand that PEPS+ requires a broader approach of interpreting data compared to the character stats of other games.
We will continue to make R&D efforts on PEPS+ for higher accuracy and effective delivery of data, and will build up its credibility by accumulating official international match data.
Thank you for your continued support and we look forward to hearing your feedback and thoughts on PEPS+.
– Gerald Kim, Game Designer from Esports Dev Team
– Hyojun Je, Data Analyst from Esports Dev Team
– Jeongho Lee, Data Analyst from Esports Dev Team
– Hanna Shin, Project Manager from Esports Dev Team