Spellsy's Kill Charts - Heatmap analysis of all teams at Worlds
This article, Spellsy's Kill Charts - Heatmap analysis of all teams at Worlds, originally appeared on onGamers.
While Worlds feels like forever ago (hell, the winning team has already broken up), I wanted to do some final experimental analysis using the advanced kill data. The first area of data I wanted to explore is the coordinates for every kill/death at Worlds. At first, we simply just plotted it, but this did not give us many usable conclusions:
While this is a visually interesting graphic, it doesn't say very much (other than blue team won more than red team). The overlapping nature of the dots makes it somewhat hard to even tell which are the truly hot spots of the map.
To breakdown this data I looked to how Kirk Goldsberry, a contributor to Grantland, approached this similar data visualization challenge. He took the shot chart that we had seen for years on the left, and turned it into a frequency and efficiency data graphic on the right.
Source of Left Image. Source of Right Image.
Kirk took a bunch of dots and turned it into a readable map where it shows not just where Lebron is great on the court, but also it shows his style, how he loves the left side.
I wanted to take this same approach for the worlds teams, in order to break down both where teams are good but also look at the style of teams, and whether they focus on dragon or bot lane or baron or jungles.
The first step to doing this was breaking the map up into blocks, and then organizing them into areas representing each region of the map. After carefully breaking each map area into blocks we came up with this chart which is the distribution of kills across the map.
Now after breaking the map into 17 regions and getting a reference point for how much action occurs in each region we are able to look at it on a team by team basis. To do this we created what I am calling the "Spellsy Kill Chart", a packed bubble chart representing a teams performance in each area. Since these graphs are information dense we created a quick how-to-read graphic to help.
North American Teams
The first team to look at is Team SoloMid. There are two big relationships here that stand out to me. First is looking at the Tier 1 towers. They have done poorly Top and also that's where they fight the least (-25%). Meanwhile Mid, where they fight above-average (+15%) they have a positive ratio, but Bot is surprisingly their most successful lane early game, with a 14-10 kill to death score. Despite previous articles showing WildTurtle's weak lane phase, they did well in the kill department at Worlds (maybe due to ganks).
The second relationship I wanted to highlight was looking at the Dragon area and neighboring Red Golem jungle, compared to the Baron area and the neighboring Red Lizard jungle. In the NA LCS we’ve come to see TSM have a heavy focus late game on vision control, especially in this Top Lizard jungle, and we can see the increased frequency paired with the successful K/D ratio shine here. However, once things get down into the Baron area they have a much weaker success rate, with a really bad .4 K/D ratio. We can see how they like to play the vision game but when they don’t have control they often take some questionable Baron calls. On the other hand, we see the opposite relationship around Dragon, where they have negative ratios in the areas surrounding Dragon, but an average K/D in the actual Dragon pit. This I think points to them needing a bit better vision control early, especially when on Red side.
For Cloud 9, the biggest thing that stuck out to me was their weak Dragon area. We have considered this team a great teamfighting team, and especially at Dragon in the NA LCS, but here they have one of the weakest Dragon areas with a .6 K/D ratio and above-average fighting frequency. This is telling us that not only do they pick poor fights but they do it repeatedly and it’s telling as to why C9 got behind in so many games early. On the other hand though they had a slight positive K/D on Baron, which as we will see with more graphs is actually somewhat rare.
Their bot lane on the other hand was extremely successful, with a strong 18-9 K-D. This really shows the strong performance that Sneaky (and LemonNation) had at worlds, and how he was one of the surprise stars out of North America after being called average for so long.
The last thing that surprised me was looking at C9’s performance in the Red base. Now, I don’t put much stock in the base kills normally cause they are most often when a team is about to lose and gets desperate, but they had 5 wins, the base score should not be that bad. This actually reaffirms some of the crazy tactics that C9 went through in order to win games. Or in this famous clip, their attempt to win but still wiping in the Red base. Despite having a relatively successful Worlds appearance, getting out of a tough group stage, Cloud 9 had a really messy performance, and it’s shown here.
LMQ had a weak performance at worlds, going 2-4, but these stats are truly staggering when looking especially at the diagonal of river. Including Top, Mid, Bot t1s and the Dragon/Baron rivers they were a combined 30-75, or .4 K/D ratio. The other noteworthy thing is that all of these sections are above average frequency, meaning that not only are they not winning these fights, but they are fighting very often. This shows their aggressive risk-taking style, but unfortunately at worlds it did not pay off for them.
Europe
Alliance has a really interesting looking graph because they won all of their games on Blue side, but lost all of their games on Red side. This made their Blue base and Blue inner towers all have zero to low activity as they never even got onto Alliance’s side of the map. Their wins were extremely definitive, unlike the Cloud 9 chart with a crazy mess all around the map.
The second really interesting thing from this graph is the extremely positive K/D ratios in the lanes, despite Alliance being a 3-3 team at Worlds. This shows their awkward playstyle where if they are ahead they are aggressive and confident, winning the lanes definitively. But if they are slightly behind they get extremely passive, give up objectives and rarely fight, leaving their K/D statistics high despite losing half their games. The only team that fought less in mid lane was Kabum, which is surprising given the hype that Froggen gets for being the leader of the team you would expect more combat to revolve around his region (which is also one of the most popular kill areas as we saw in the percentage graph way above).
While we saw in the last graph how Alliance won all Blue games and lost all Red games, here Fnatic was the opposite. Fnatic won 2 out of 3 games on Red side, but lost all 3 Blue side games. Fnatic also similarly had the situation where when they were losing on Blue side they rarely fought until things reached their inner base. In their 3 Blue side losses they only really lost one big teamfight outside their base and it was Top lane.
One thing that surprised me though was Fnatic’s strength on objectives. At baron they had the 2nd highest K/D ratio at 1.8 and a high K/D ratio at dragon with 1.6. This combined with their low fight frequency in top and bot lane showed their objective-focused play.
SK’s graph looks just as we would expect, with a large red circle in the dead center. Jesiz had a really weak performance this worlds and it shows clearly here. Similarly bot had a weak performance too, going 9-17. While I’m sure the jungle had a significant influence on the team morale and synergy neither the mid lane nor the bot lane put up good performances on their own. Top lane on the other hand was positive with a 1.6 K/D ratio, reaffirming that fredy122 is the strongest part of the SK team.
Despite the weakness of mid and bot, I think this shows one of their biggest flaws which is around the dragon area, a brutal .3 KDA for a team that was in contention for their group.
Korea
This Samsung White chart after seeing all of the other teams truly shows the dominance that they had on this year’s Worlds. They are the only team to have a K/D ratio greater than 1 in every section, and often have >3 K/D. Their only ‘weak’ spot was Baron and even that had a positive K/D ratio.
Also worth highlighting is their dominance in the Dragon pit and Top lane. In the Dragon pit they were 42-17, the second highest K/D in that area EDG had just 17-10.
SSW’s sister team, Samsung Blue, has a lot more blemishes on their map, despite going to the semi-finals. SSB was much more dominant on Red side winning 5 out of 6 Red games but only 3 out of 7 Blue games.
One noteworthy thing to look at is the frequency at which SSB fights, especially on the Blue half of the map. They look almost like a Chinese team in their large bubbles throughout the jungles and Mid lane. Unlike the European teams we saw in the last section, SSB fights a lot when they are losing and we can see this in their large red circles around the blue base.
Najin White Shield has a really odd looking chart, with generally very small bubbles. Najin White Shield had the 4th least Kills+Deaths per minute overall (KBM, ALL, FNC with less), by far the least out of the winning teams.
The other surprising thing shown in these graphs is the weakness NWS had top lane, which is supposed to be where their star player Save resides. Throughout all of top lane both at Tier 1 towers and Tier 2 towers they had weak K/D Ratios. Instead where they won their games was over Dragon, as they had one of the stronger records in the Dragon area.
China
Despite making it to the Finals, Starhorn Royal Club’s kill chart doesn’t look that spectacular, instead it is rather average. To me, this establishes many people’s thoughts that SRC had a rather unremarkable advancement to the finals; looking at this team’s chart compared to the other finalist Samsung White shows no real match.
I also expected there to be more success in the bot lane. We see an above-average frequency, as expected, but they were almost forcing too hard as they went 34-36 overall. But in a trend we will see for these Chinese teams, their real success was at the objectives where they went 1.3 K/D at dragon and 1.4 K/D at baron. This is where SHR’s straightforward shotcalling came into play and they showed their teamfight prowess.
While with most teams we see a mid-lane focused strategy, with most kills being at the Mid tier 1,2, or 3 towers, with EDG we see a strange side-lane focus where there were actually more kills in each sidelane than there were in Mid lane (which is unusually if you look back up to the kill % by area chart we had near the top). EDG didn’t even really fight in the jungles, as they are all below-average frequency. Their big money-maker though was Baron, where they had the highest K/D ratio of any team, and a significantly larger frequency too. They fought often and successfully around Baron, and didn’t do too bad at Dragon either. Lastly looking at the lanes, despite the flak that Namei has gotten for his disappointing Worlds, Bot lane was actually their most successful lane, although that isn’t saying much.
For OMG we see almost the opposite that we saw in the last chart, where they have a Jungle and Mid lane focused style and stay away from the epic monsters and side lanes. The biggest surprise here though to me is the success that they had in the bot lane, despite having DaDa777 support half of the games they still came out with a 1.6 K/D ratio. Also I was rather skeptical of their mid laner Cool, who I thought underperformed despite his hype, but this shows that his mid lane presence was instrumental to their success, as that was the center of their operations.
SEA and Wildcards
TPA, as we will see with most of the teams in this category, had a weak performance at worlds with a 1-5 win-loss. However, despite the prominence of red on this chart their best spots seemed to be around Dragon and Baron, showing they at least have the potential to teamfight well.
Unlike TPA who had relatively low frequency as a whole, AHQ as we see really likes fighting and have lots of high frequency areas. Unfortunately for them, they are really bad at Dragon fights which set them behind in a lot of their games, forcing them to have comebacks by picking people off in the jungle or at towers.
Kabum is as expected a very red team. With no real promising areas, one thing that stuck out to me was just how small all of their bubbles were. Not only did they lose most fights, but they also avoided fights as much as possible, while you would imagine that the worst teams have the least to lose and thus play more aggressively. Instead, Kabum had the least Kills+Deaths per minute out of any team (the only team with less kills+deaths than Alliance).
Lastly, Dark Passage who is the opposite of SSW’s graph: they died equally everywhere.
Conclusion
This chart and analysis is at its heart an experiment to see if we can better measure a team's strengths and weaknesses in ways we might not see in a live broadcast. Let me know if you think this is a worthwhile attempt or if there are any suggestions.
One big concern I had with these graphs is the Blue/Red side difference, but to double up each of the graphs above would make this article take up even more space than it already is. So, if you are looking for more information I made an imgur album with the Blue and Red side for each team in a less graphical format. CLICK HERE TO SEE IT (or click any of the images below).
Data collection, analysis, and design by Spellsy. Help designing by Benson Li
Related Articles
