TSM ADC Case Study: Doublelift vs. WildTurtle

TSM ADC Case Study: Doublelift vs. WildTurtle

1.0 Introduction

After Worlds 2016, Team SoloMid faced a unique dilemma as their superstar ADC, Doublelift, decided to take a break from the NA LCS in favor of streaming. Although the team had other options on the table for the 2017 Spring Split, Team SoloMid (TSM) ultimately decided on bringing back WildTurtle, who had previously played for TSM and had performed well for Immortals (IMT) throughout 2016.

Despite having great chemistry with the team out-of-game, WildTurtle had big shoes to fill and consistently ended up being the focal point of criticism whenever TSM failed to meet their high expectations.

It’s easy to go back and forth with water cooler discussions on Reddit and Twitter, but we want to dig deeper and substantiate the real truths of the matter. In this article, we will be combining the power of our analytics platform with the data that our friends at Stage.gg have collected in order to empirically tackle the following questions:

  • Did Wildturtle deserve the heavy criticism and did his play actually hinder the success of TSM during the 2017 Spring Split?
  • Is TSM really better with Doublelift? If this is true, what are the specific factors that can help explain Doublelift’s better fit with the team?

2.0 Roadmap

League is a complex and ever-evolving game, not to mention a lot happened between Summer 2016 and Spring 2017 besides just a roster change. To answer the questions posed above, many factors need to be taken into account. Here is our approach for tackling the two different TSM seasons:

First, we’ll be comparing win records between the splits so we can understand the different circumstances that govern the statistics we are looking at. Unlike traditional sports, stats for winning and losing games in League of Legends are inherently more polarizing, so the difference between a winning and losing season is even starker.

Second, we’ll examine strategy-related decisions such as resource distribution, champion picks, and meta shifts in order to determine the impact on statistics that we derive from the games played.

Finally, we’ll measure performance factors such as resources obtained, damage output, deaths, and death severity to determine differences in team performance with two carries.

3.0 Split Record Comparison

To begin, we’ll be examining the win-loss records of TSM’s 2016 Summer Split and 2017 Summer Split. TSM had a rocky start in the Spring Split with WildTurtle, losing their first series to Cloud 9, and losing some games (but not the series) to Team Liquid, Dignitas, and Immortals.

TSM win percentage

Figure 1 – TSM’s weighted running average win percentage for each season graphed against the strength of schedule. Greater values for strength of schedule correspond to tougher opponents for the week.

Looking at Figure 1, you will notice that TSM’s Spring Split win percentage (again, for individual games not the series) started off at a low 40% and progressively improved throughout the split. Their Summer win % started very high and had nowhere else to go but down, but the overall win percentage still finished ahead of the Spring Split.

Note that this is a weighted average which weighs win percentage closer to the week in question more heavily, to better illustrate the team’s performance changes over time. Also note that we have layered a strength of schedule function to tell you what caliber opponents was being played each week (this is based on the rankings of the team at that week).

4.0 Strategic Factors

4.1 Meta Changes

League of Legends is not a static game so it’s important to take into consideration some of the major changes that happened between splits. Season 7 started, which brought with it a bunch of meta changes, not to mention roster changes.

To give you a better idea of the major shifts, we compiled Table 1 below, showing the top 3 champion picked during the season for each player on TSM. We also added in WildTurtle’s Picks on Immortals (green) and Doublelift’s picks on Team Liquid (blue) during the respective seasons they were not on TSM for good measure:

Champ picks

Table 1 – Top 3 champion picks with number of games played during each season for TSM players, IMT WildTurtle and TL Doublelift.

In the preceding Summer Split, the most popular ADC picks were Sivir, Ashe, and Lucian with the backup dancers being Caitlyn, Ezreal, and Jhin. It’s worth noting that even with Ashe being so popular, Doublelift only played 4 games with her, preferring high DPS, self-sufficient champions like Lucian and Sivir instead.

Meta changes go way beyond champion picks though, but knowledge of these basic facts can at least prepare you to think critically on any differences that might pop up while we continue with the analyses.

4.2 Resource Distribution

The next thing we want to look at is how the gold was distributed in each one of TSM’s seasons. Despite League being a team game, not all players receive equal resources, with teams opting to give the bulk of their gold to a few players – the carries. Who you funnel gold into generally depends on meta and how that game is panning out.

If you’re a team with one-dimensional players, you risk being hamstrung and less able to adapt to the meta. This often forces you to focus gold through certain players, but generally teams should be focusing more on getting the ideal champions ahead for that game instead of players.

Doublelift is known to be a carry-style player and although he loves to carry, he has mentioned before that he is willing to do whatever is needed to win.

So let’s take a look at how the gold was distributed to each player on TSM (Figure 2):

Gold share

Figure 2 – Summer and Spring gold share across TSM players with WildTurtle with Immortals in Yellow.

What we see from Figure 2 is that the gold share remained pretty equal across the board. The largest shift happened between Doublelift to WildTurtle, but even that was only about 1%, which translates to 500g in a 50k gold game (little more than a Doran’s Blade). We also threw in WildTurtle’s gold share with Immortals for good measure, and they are all within the same ballpark (although a little higher).

However, gold share alone is not enough to fully answer the amount of resources being given to each player. Figure 3 takes a look at the gold per minute for each player as well as the entire team.

TSM gpm

Figure 3 – Gold per minute for each player on TSM with WildTurtle Immortals in Yellow.

4.3 Gold difference

In Spring 2017, TSM, on average, earns about 40 gold less per minute, which translates to 1200 gold in a 30 minute game. This would mean little if it was evenly distributed to every player instead of allocated to one, however, most of this drop in gpm comes from WildTurtle, who earns about 35 less gpm than Doublelift. Consistently in most matches, this means WildTurtle would be, on average, 1k gold behind from completing his next major item by the 30 minute mark.

If you are curious where the gold difference is coming from, Doublelift earns about 35 more: 20 of those gpm comes from kills, 3 from assists, ~2 from minions. The final 10 is probably from objectives, seeing that they got 1 objective more every 13 minutes in Summer (Doublelift) when compared to Spring (WildTurtle).

Since TSM’s performance improved considerably in the late part of the Spring season, we decided to compare the two halves of the season as separate. Although this reduces our sample size, it strengthens any differences between poor and good performance and can allow us to compare the better performance favorably with the Summer Split.

4.4 Gold per minute (gpm)

Figure 4 shows us that towards the end of the Spring Split, TSM’s total gpm numbers are much closer to those in the Summer Split (1889 vs 1899). However, this gold is not going to WildTurtle; this gold is going mainly to Bjergsen and Svenskeren, who see a gold increase of about 25 gpm and 20 gpm respectively, or, alternatively, Sorcerer’s Boots upgrade and 2 Cloth Armors.

Gpm spring

Figure 4 – Gold per minute across TSM for Early 2017 and Late Spring 2017.

It is also important to notice that across TSM the gold per minute stats increase from Early Spring (pre-Feb 19, 2017) to Late Spring (Feb 19 and onwards), which may also be due to them winning more in the later parts, but it also shows that they are becoming better as a team in capitalizing on gold sources.

This extra gold income comes from extra kills, with there being 1 more kill every 10 minutes in Late Spring when compared to Early Spring. This also translates into assists, with there being an extra assist every 6.25 minutes in Late Spring when comparing to Early Spring. Considering how many kills happen in a competitive game, 3~4 extra kills a game is around 900 -1200 just in extra kill gold.

Add to that the extra gold from assists and in turn an easier time getting resources or objectives since the enemy is dead. The CS per minute for WildTurtle in Spring 2017 and Doublelift in Summer 2016 are 8.6 and 8.7 respectively, further supporting the hypothesis that the gold difference is not being generated primarily through farm.

4.5 Emphasis Shift to Other Lanes

One question that immediately comes to mind when comparing both halves of the split: Were there any significant team decision making changes accompanying the stat differences? The data paints a picture of a change in TSM strategy, with less gold being invested into the ADC and more into the mid and top lane.

This may be due to either a deliberate strategic decision to focus the resources away from the AD carry, originating from a player behavior change, or a change in meta, which made those the most favorable gold investments.

We put together Table 2 to help us answer that question:

champ picks spring

Table 2 – Top 3 champion picks for TSM roster in Early vs Late Spring.

So what we’ve seen so far is that WildTurtle is not taking on the same hard carry roles as Doublelift did, but rather he’s adding in a lot of utility, and the team is playing around other angles to get their advantage. This was not specific to WildTurtle, as most ADC’s adopted this strategy in their respective teams during that season.

5.0 Performance Metrics

5.1 Damage output

Now that we have a better understanding of TSM’s strategy and how they’re allocating their resources, let’s take a look at the effect on their damage output. At Mobalytics, we love the power of data and numbers, but we know that that power comes from properly examining them. As such, we have created the Champion Bias Coefficient (CBC). This coefficient applies a weight to metrics that are reliably based upon a champion’s specific kit to better assess a player’s performance across different champions.

Let’s use an extreme example to illustrate what we mean. Karthus is able to readily output damage to champions throughout the game, and so, a Karthus player will often have a very high damage done to champions score. Lulu or Zilean, on the other hand, don’t deal as much.

Does that mean the Karthus player is automatically better? We don’t think so. This became incredibly important when measuring DoubleLift and WildTurtle across different metas.

The results of this can be seen in terms of damage dealt to champions per minute in Figure 5 below:

scaled dpm summer

Figure 5 – Coefficient adjusted damage per min to champs for each player on TSM.

TSM Damage output decreases: What we can deduce is that TSM’s damage output dramatically decreased from Summer to Spring, dropping by almost 400 DPM from TSM with Doublelift to TSM with WildTurtle. The largest drops are seen for Bjergsen in the mid lane, doing an average of 132 less scaled dpm in Spring compared to Summer, and Doublelift/WildTurtle as the AD carry, whose scaled dpm dropped by almost 200.

If we take a closer look into the Spring Split, comparing Early to Late Spring just like we’ve done above, we get the following result (Figure 6):

scaled dpm spring

Figure 6 – Scaled damage per min to champs for each player on TSM in Early and Late Spring.

What we see in Figure 6 is that the total scaled damage done per minute to champions for TSM did not dramatically change, however, we do see an internal change in damage. Bjergsen puts out about 60 more scaled dpm, whilst WildTurtle’s damage decreased by 80 scaled dpm. We note though, that TSM’s performance was significantly better in the Late Spring than Early.

The increased damage may well be the result of the team working together better, and also underlines what we’ve already discussed before, with WildTurtle taking on a more utility role, and Bjergsen more of the carry role. Of course, the damage output of Bjergsen is still not even close to what it was in the Summer (it’s actually still off by 100 scaled dpm), but nevertheless, it seems that whatever change occurred in TSM was successful at elevating the team’s damage and performance.

5.2 Deaths per minute

Let’s take another look at TSM from another angle, namely in terms of how often they die. Deaths can be a much more solid indicator of how careful or careless a team is playing.

Figure 7 shows us the comparison of deaths per minute for TSM in the Summer and Spring Split:

deaths per minute summer

Figure 7 – Deaths per minute for TSM in Summer 2016 and Spring 2017 Split including WildTurtle with Immortals in Yellow.

The first thing that stands out is the “dramatic” increase in total deaths per minute from Summer to Spring. However, upon close inspection, this increase is actually only 0.06 deaths per minute which, for clarity, corresponds to about 1 extra death every 15 minutes, so really not as dramatic as it initially seems but still significant at the professional level.

Generally, almost everyone keeps their death statistics from the last season, with Svenskeren seeing the biggest change at about 1 more death per game on average. So for those of you viewers who constantly have been stating that Svenskeren has been getting caught out a lot more as well as getting punished for risky invades, pat yourself on the back!

It becomes interesting though when we not only look at the amount they die, but rather at the severity of their death. We’ve developed a metric for measuring detrimental deaths, which identifies deaths that lead to a loss of an objective. This allows us to get a better understanding of how severe a player’s death really was.

We can see this spread in Figure 8:

Detrimental deaths across Splits

Figure 8 – TSM’s detrimental deaths in Summer Split and Spring Split.

Despite dying marginally more during the Summer Split, when a player does die the repercussions are more severe. Player death was twice as likely to lead to a lost objective in the Spring Split than it did in Summer.

Generally, in competitive play, high detrimental deaths should be a results of one of two things: not having lane priority or dying only in the later stages of the game. In the first case, if you’re dying when your team does not have priority in lanes, then it makes it very easy for the enemy team to take objectives from a kill. In the second case, obviously there are longer death timers in the later stages of the game and teams have access to much more DPS. This makes it much easier to take many objectives off a single kill.

Alleviating the detrimental deaths may have been one of the key roles that Doublelift fulfilled on TSM, with his well known decisive play style and shot-calling guiding the team in situations where they’re down one, or several, men.

In Figure 9, we’ll take a closer look at how this changed over the Spring Split, again comparing Early to Late Spring:

Detrimental deaths spring

Figure 9 – TSM’s detrimental deaths in Early Spring and Late Spring.

Although the team is learning to play together better, adapting to fill the potential guiding role that Doublelift played, they were not able to fully adapt to bring their stats down to what they were before.

6.0 Conclusion

Now that we have all poured through all this data, let’s revisit our original questions:

  • Did WildTurtle deserve the heavy criticism and did his play actually hinder the success of TSM during the 2017 Spring Split?
  • Is TSM really better with Doublelift? If this is true, what are the specific factors that can help explain Doublelift’s better fit with the team?

The numbers point to WildTurtle being a fantastic AD carry in his own right. According to the awesome folks at Stage.gg, he did the second most damage out of all ADC’s in the 2016 Summer Split, when he was playing for Immortals (Doublelift on TSM coming in at third). Perhaps, however, he just isn’t the personality that TSM needs.

Our analyses show that Doublelift’s play elevates TSM’s performance and allows the team as a whole to perform better. His decisive playstyle and shotcalling influence (which can be seen on shows like TSM Legends and Liquid Lol Squad) seem to be what fits TSM’s style and allow the players to perform at their best.

There are definite limitations to our analyses despite our attempts to mitigate these limitations. League is a complex game and it is difficult to account for all the nuanced variables that differ from game to game, especially across different metas.

One of the more obvious limitations of our analyses is that the team who puts out more damage isn’t necessarily playing better, but may be just taking more fights.

Also, it’s difficult to compare a season with more wins to one with less, but our separation of the Spring Split into two halves should help alleviate that. However, combined with the less detrimental deaths and the more gold the team had with Doublelift, it becomes a harder set of facts to ignore when comparing the team’s performance in both splits.

There are so many other things we wanted to look at: comparing performance at MSI vs Worlds, looking at other team stats during the splits as a frame of reference, and so on. However, we are limited by time and resources. At the end of the day, TSM with WildTurtle repeated as NA champions, and WildTurtle is headed to Flyquest for the 2017 Summer Split.

We hope you found this article informative, and it will serve as a good foundation to launch a discussion where like-minded League stat geeks can discuss their theories and present their arguments (here’s our Discord). By the way, here are the original graphs if you’d like a more in-depth look at the data!

Credits: This article took a lot of time and effort from people at both Mobalytics and Stage.gg. Fortunately, no Yordles were harmed in the creation of this article. A big thank you to:

  • Max, our data scientist, did the bulk of the analysis with the help of our analyst, Hewitt, and our in-house coach, Adam.
  • James and Guilherme from Stage.gg provided data and insights to drill down the topic and make the article more robust.
  • Ryan, our data architect, helped us organize and compute the data so we could make sense of it.
  • Agilio, our content creator, made sure that our article looks and reads like an actual article.
  • Stasya, our designer, without whom we would have no pretty purple graphs.
  • Amine, who just talks a lot, writes one sentence, and says “this looks good”.

Mobalytics is the first personal performance analytics platform for competitive gamers. The platform measures players performance, helps them define their strengths and weaknesses and provides personalized advice on how to improve. Feel free to sign up for Mobalytics Private Beta here!