Breakdown of Overwatch Analysis
First off, an introduction is in order. My name is Kyle, but A_D1R3W0LF is usually what I go by in the online world. My journey in Overwatch started on console a few months after release. I made the transition to PC soon after that to play with friends. I joined a team with randos where I played for about about a year. We were diamond, we didn’t do anything to great success. When I got too busy to dedicate time to playing, I transitioned to coaching. I soon joined Mizzou Esports as the analyst for their Overwatch team. We won NACE and finished top 10 in Tespa that year which was a good result. I was also the analyst for Lunaris when it first started and stayed with them through their two Contenders Trials runs. I left Lunaris just before the Qara rebrand as I could no longer dedicate my time between two teams and keep my productivity high. Several people have asked me what I do as an analyst, so I decided to answer them in the form a of a series of blogs/op-eds about my opinion on analysis in Overwatch. Hopefully this can provide some insights (hehe, joke you will understand soon) as to analyst work and analysis in general.
If you were to compare Overwatch League teams to other Overwatch teams (T2, T3, Collegiate), outside of the obvious difference in funding, the largest difference is usually found within the coaching staff. Most teams will utilize a head coach. Some might have an assistant coach for macro or micro level. However, few teams will have an analyst, and those that do are usually T2 or OWL. This is because the analyst occupies an unusual role within the coaching staff as a luxury most teams cannot commit to. But when the analyst is of high-caliber, they can be an invaluable addition to a team.
An analyst’s job is to gather, maintain, and utilize data to support and facilitate the advancement of team and individual play. You can almost think of it as the Moneyball approach to Overwatch from the Oakland Athletics’ baseball team in the early 2000s. There are a wide range of things that an analyst can do, but for simplicity’s sake, I have broken it down into three levels of analysis: Reactive, Active, and Innovation. Each of the three levels builds and progresses on the previous, so much like Maslow’s Hierarchy of Needs, you need to be strong enough at one to move onto the next. Most analysts will never progress past the base level because they do not have the time or resources to dedicate to developing their analysis and moving up the pyramid. In this post, I will define these levels and describe the functions and benefits of each. My hope is to give future analysts a solid base to begin while providing the necessary tools to develop their craft.
The base layer of the pyramid, and easiest level to utilize, is Reactive. Here, the person in charge of analysis does basic tracking of some statistics. Within Overwatch, most teams already track some things at this level, such as the win rates of maps. Additionally, player win rates and composition win rates are relatively common for teams to record because it is simple and necessary since Blizzard fails to make these stats readily accessible for teams to reference. Map record is the easiest to track as the individual responsible for data collection only needs to scan through VODs to end of maps to see results and note them down, typically utilizing Excel or Google Sheets to do so. Similarly, to track player win rates, one only needs to see the end result of the players for each map. Compositional win rates, on the other hand, are more time consuming as one must notate the result of each fight. However, this becomes easier with practice by watching VODs in double or faster speed to shorten the time needed to complete the task. Additionally, with the way Overwatch metas work, a team might only need one composition as this is usually strong for a long period of time in competitive settings, being able to understand the patterns with statistics becomes invaluable.
The purpose of these statistics is pretty obvious to most. Map win rates are used when picking maps in official matches because you want to play on the ones you are best at to increase your team’s chances of winning. Player win rates side with map rates for similar reasons, utilizing different players on the maps they do better on will boost winning probability. Furthermore, looking at win rates of players in conjunction with each other will not only lead to a higher probability of success but also increase cohesion within the team. However, this is not possible for some teams, such as those that may not have more than six players, because they do not have the luxury of swapping out players for specific maps. Compositional win rates allow teams to identify potential meta comps before other teams can.
Within these three types of statistics, there can be variants made by individuals to better suit their team’s needs. From making a differentiation between scrim map win rates and official match map win rates, to comparing the win rates of two players to decide who the starter should be, to tracking composition win rates against another composition, each of these variations requires slightly more dedication than their basic form. However, none are too different or advanced enough to be considered part of the second level of the pyramid. Unfortunately, it is often difficult for teams to go into such detail without a dedicated analyst because advanced statistics take more time to complete than is available to someone with other duties on the team.
Like Maslow’s Hierarchy of Needs, once the requirements of the Reactive level are met, a team is able to focus on the next level of analysis — Active. The switch from reactive to active is usually signaled by having a dedicated staff analyst. This is because active analysis requires more time to do correctly and draw value from. At this level, the analyst and data take active roles in the shaping the direction of the team. Although reactively tracking map and player win rates can improve the team’s chances of success, these statistics are basic and can be done by anyone at any time. The goal of active analysis is to shape and improve team coordination and individual play.
The value of this level comes from the ability to quickly learn, adapt, and improve on the meta and individual play within the meta. Active analysis can do this by allowing teams to identify strategies or events in a composition that will lead to more fight wins. Think of this level as the Moneyball layer of Overwatch analysis. This is the gaming-the-system part where the analyst tries to get maximum value out of the most efficient input required. While the terms efficient and minimum are similar, they are not synonyms. Efficiency requires more work up front to make the process easier, while minimum work would imply only completing this analysis once and using the results for all future analysis. Becoming efficient is more beneficial in the long term as it puts systems in place to take on more advanced analysis.
There are a wide variety of statistics and patterns that can be analyzed within this level. Items such as fight win rate when a certain hero is eliminated first or fight win rate when a certain target is hacked to initiate a fight. In GOATs meta that dominated for over a year, teams started to key in on fight win rates when the Reinhardt was the first death. Looking back now, it is obvious that teams would win more fights on average by eliminating the enemy Reinhardt. However, this was not as apparent in the early days when teams were still learning the meta. The original composition used Moira to eliminate this character, but as the meta developed more teams started to utilize Zenyatta since metrics revealed that the extra damage from discord orb allowed teams to remove the enemy Reinhardt from the equation quicker. To counter this adjustment, teams started running Winston as the main tank in place of Reinhardt because he was better able to handle the pressure of the discord orb as well as build Primal Rage quickly due to the stacked nature of GOATs. Teams that were able to key onto these statistics were more successful within the GOATs era, and most of these teams had a dedicated analyst on the staff to identify these trends in a timely manner.
The highest value an analyst can provide is found at the top level of the pyramid because this is where they have the opportunity to create their own metrics to utilize. This comes from working closely with either the coaches or the players to develop statistics that they believe will help improve the team or themselves. Coined Innovative — due to the fact that all analysis developed here is brand new or tackles a specific problem, issue, or train of thought from a different direction — this level is for creativity and the discovery of advanced ideas or philosophies of analysis within Overwatch.
There are endless possibilities for analysis within this level, at least when it comes to the analyst’s side of the equation. The only limit is the individual’s creativity and skill. The biggest win provided by this level is its ability to give teams the advantage against others through access to unique and innovative statistics. However, there is an inherent risk as the statistics you develop have the potential to be misleading and a detriment to the team, so caution is advised when developing stats from scratch.
If I were to follow the format of the previous two sections, this is where I would give a few examples of statistics that fall into this level. However, since the cornerstone of this final level is that all the stats are unique to analysts or teams, there are no concrete patterns. This level is hard to show examples because this is where an analyst’s true value is shown. Nevertheless, I can give a few pointers about potential analysis from this level. For instance, when developing stats, one thing I like to keep in mind is the use of a delta factor or being able to compare two or more components. Additionally, keep in mind what you already have at your disposal and how you can use it to your advantage. The pure potential provided by the Innovative level is enough to excite any analyst.
Since the start of Overwatch analysis, data collection has mostly been done by dedicated individuals in the form of filling out a spreadsheet during or after scrims. The biggest downside to this process was the time commitment required to track everything that you wanted to record. On the plus side, the data was usually accurate as it was done by hand and the analyst was able to rewind or fast forward to find and correct mistakes. Over the last year, artificial intelligence has been utilized to parse scrim VODs with image recognition and quickly scrub through to gather information. Unfortunately, the image recognition is not as good as a human doing it by hand, so what you gain in time spent, you lose in accuracy. However, in the long run, this method will prevail because the less time analysts have to spend on data collection, the more they can spend on actual analysis, which leads to them moving up the pyramid. Additionally, with recent Overwatch updates of being able to look at the workshop inspector logs, this auto-collection is now easier and more accurate than image recognition software because it pulls events directly from the game.
With such technological developments becoming more accessible, each level of the analysis pyramid will be shifted down. The bottom level, Reactive, will become something that all teams can do because it takes little to no expertise or time commitment to complete. The second level, Active, will become something almost any team can do if someone commits a little time. Although this level will probably still be gatekept by having a dedicated analyst, the fundamentals become more accessible to everyone. The final level, Innovative, will remain the same because it will still require teams to have a dedicated analyst who is able to develop stats that have never been seen before.
Overall, analysis within Overwatch is still the Wild West because the dedication resides within a small number of individuals and the technology is not available yet to make it easily accessible for all. There are the basic stats that everyone gravitates towards because they are simple; the slightly more advanced stats that will allow teams and players to develop quicker than others; and the final level of stats which is the great unknown as the only limit is the analyst’s brain power. The addition of new technology will make the analysis process easier and force analysts to slightly redevelop their craft if they want to remain a valuable asset to their team.
In the future, I would like to have full access to the API of what is tracked within the game. Unfortunately, as of now this still lies on Blizzard’s end, and the community has no influence. For what the community does have influence over, I do not know what to expect because there are a wide range of individuals with varying experience and philosophies on the topic of analysis. Regardless, I believe analysis is an important part of Overwatch, as it is with all sports, and being able to shape the direction it goes while still in its infancy is exciting. I hope to write more of these but actually looking at statistics. I’m currently analyzing the October Conteders tournaments from China, EU, and NA; so keep an eye out for that. You can follow me on Twitter at https://twitter.com/A_D1R3W0LF or Twitch at https://www.twitch.tv/a_d1r3w0lf if you liked what you read. I tweet a little about stuff I do and stream on occasion.
If you have made it this far, congrats. Seriously, people do not tend to find analysis to be exciting as its usually just a bunch of numbers and only want to see the fancy graphs and tables. I have a few shoutouts to people that I feel like I need to make. First is Kevin Reape, Mizzou Esports Director, for taking a chance on me and allowing me join and help out. Without the initial Mizzou Esports opportunity, I wouldn’t have met Hal (https://twitter.com/halplaysgames) and join Lunaris and get some T2 exposure. The next two people are Aud (https://twitter.com/MrCoachAud) and Skiritai (https://twitter.com/SkiritaiGG) for helping me with the coding side of my analysis as I am smooth brained in that department. Your help made a lot of my more advanced analysis possible. Shoutout to the team over at Insights (https://twitter.com/insightsgg) for taking me on as an intern and exposing me to their software and OWL teams. The last shoutout is to my sister for reading about analysis which she has zero interest in and editing this big block of text. Oh and Speakeasy (https://twitter.com/Speakeasyow) for asking for weird stats and pushing me to better my craft.
