How trading stats can show a team's structure — and why Heroic stand out

We visualise Heroic's unique style, as well as show why NiKo, YEKINDAR, and Ax1Le are more similar than you might think.

Breaking down a Counter-Strike team into clearly defined roles is a near-impossible task. Every team's structure is different and it's rare for teams to have strict definitions for each player. On top of that, roles are fluid from round to round and players can occupy the same position with vastly different styles.

Yet, it is still useful to have labels for players from the outside. Simplification is inherently flawed, but it can reveal a huge amount as long as we acknowledge those problems. A common method is to divide teams by aggression, usually with opening kill attempts, but there is a key difference between opening kills and entry kills: Players like Shahar "⁠flameZ⁠" Shushan may have fewer opening attempts than Mareks "⁠YEKINDAR⁠" Gaļinskis, but are first into a bombsite just as often.

Read more

Is the entry fragging role dead?

The way we quantified entry fragging in our last article was by manually seeing which player went into a bombsite first, which is simply impossible to do over a huge sample size or for every team. The by-product of this approach, however, is that it confirmed two things. One, opening kills are a fairly weak predictor for how often a player is first into a bombsite. Second, both entry kills and opening kills are poor predictors for whether a player is in a team's 'entry' pack.

Another is by position, which was used in the HLTV Award Show for the panel awards. This is better, but still fails to take into account the exact style of players; Sergey "⁠Ax1Le⁠" Rykhtorov lurks on six out of seven maps, but many people were still surprised that he was put in the 'Closer' category. Players define their own game, rather than letting positions define theirs.

Lurkers go in first when the pack comes to their site; some space takers go so far ahead of their pack that it is nearly impossible for them to be their team's designated bombsite entry. So, can stats do a better job?

The primary method for this article will be to look at trade involvement. If a player is lurking or aggressively searching ahead of the rest of their team, they are less likely to both be traded or to trade a teammate. Therefore, looking at trading statistics excludes both aggressive and passive lurkers from the 'pack,' unlike opening kill attempts.

We can expand beyond trade kills and traded deaths, too. Another similar stat is what is labelled as 'saved teammates per round' on site, measuring "how many times this player killed an opponent who was attacking a teammate, within 1 second of the last attack." The reverse of this is how often a teammate was 'saved', which can synchronise with deaths traded.

So here's the formula. We combine trade kills and deaths — as raw numbers and as a percentage of total kills — in addition to 'saving' teammates to get a rough idea of how often a player is around their teammates.

This overall 'Lone Wolf' rating gives us the results we might expect: The players that score highly are those that are comfortable on their own, in both an aggressive and passive sense.

Robin "⁠ropz⁠" Kool and YEKINDAR have identical scores — the Latvian's aggression means he is around his team as much as the poster boy for passive lurking in 2022. Similar players follow. Nikola "⁠NiKo⁠" Kovač, Robert "⁠Patsi⁠" Isyanov, Kristian "⁠k0nfig⁠" Wienecke, and Evgenii "⁠FL1T⁠" Lebedev are all aggressive, but — according to this metric — outside of their team's pack.

Now we should address a flaw of this approach. When we only take into account traded deaths and teammates 'saving' players, we are slightly punishing 'successful' entries. If someone swings into a bombsite and kills two players, nothing is recorded trade-wise. They do not need to be traded, they did not take any damage — but they were still at the tip of the spear.

Some players will succeed in these low-probability rounds more often, but it will be very rare for them to not rack up trading stats at the same time. There is also, at the moment, no way to directly distinguish between an opening kill that came from a lurk or a bombsite entry. Using trade statistics combats that effectively enough that it is a lesser evil compared to the greater flaws of opening kill attempts.

Another flaw is the lack of side-differentiated data, at least for now. HEROIC's players do fairly poorly, with all five players closer to the pack than being a Lone Wolf, probably caused by how cohesive they are on CT-Side. OG, too, score poorly, which might be because of their reliance on grouping up early in the round.

Other teams are punished the other way, for spreading out too much. It is probably not a coincidence that international teams like FaZe, ENCE and MOUZ operate more as 'Lone Wolves' than as trade-heavy packs. This is all to say that a player's individual 'Lone Wolf' rating is more useful when compared to their teammates than across the whole field. Because of this, we are not combining every player into one chart. Instead, we're splitting it up.

First, let's do it with four teams that fit the meta, our model accurately denoting a team's entry pack in accordance with the eye test. There will be four groupings made to each team: The 'Lambs' that are thrown into entries, the 'Crows' that focus on trades, and the pure Lone Wolves. These names are just for fun; remember that we are not measuring effectiveness here, but style. There is no correlation between high fraggers and these zones.

FaZe are the team with the most consistent and intuitive roles in the world and that is reflected here. Finn "⁠karrigan⁠" Andersen and Håvard "⁠rain⁠" Nygaard are the main force in the entry pack (Lambs), with Russel "⁠Twistzz⁠" Van Dulken and Helvijs "⁠broky⁠" Saukants as auxiliaries (Crows) and ropz as the main lurker (Lone Wolf).

The same pattern is here for the other three teams. On G2, Rasmus "⁠HooXi⁠" Nielsen, Aleksi "⁠Aleksib⁠" Virolainen and Audric "⁠JACKZ⁠" Jug are the Lambs, with NiKo as the Lone Wolf. On Cloud9, it is Vladislav "⁠nafany⁠" Gorshkov and Abay "⁠HObbit⁠" Khassenov. On Liquid, Nick "⁠nitr0⁠" Cannella. This lines up with what we see both in opening kill attempts, positions, and entry attempts.

A full interactive chart for more teams can be found here. Simply click your desired team on the right side of the chart to create another version of the chart.

We can see clearly which players are usually first in, second in, and on their own across the map. But this is on easy mode: What about the teams that don't fit the pattern, or have less defined roles? Here are a few more teams:

This last set of teams is more difficult. Natus Vincere are not too irregular, though lacking a true Lamb since Kirill "⁠Boombl4⁠" Mikhailov's departure. Outsiders, too, still retain the Lamb-Crow divide of most teams. What is strange is the lack of a true Lone Wolf; all the players are fairly close to each other on the chart.

The same is true for Vitality, despite them being centred in the opposite corner of the plot. Dan "⁠apEX⁠" Madesclaire and Emil "⁠Magisk⁠" Reif are nearly Lone Wolves, but not quite. Peter "⁠dupreeh⁠" Rasmussen and Kévin "⁠misutaaa⁠" Rabier are nearly Lambs, but not quite. Mathieu "⁠ZywOo⁠" Herbaut seems to be the only trader, turning their plot into some kind of strange constellation.

It is not too different to Cloud9's chart, with Dmitry "⁠sh1ro⁠" Sokolov and ZywOo clearly marked in the top left. Yet, as with opening kill attempts, Vitality shares a lot of the other duties, creating what is essentially a straight line along the riflers. What this means is that all four riflers from the start of the year are trading and getting traded at similar rates; dupreeh veers off slightly towards the Lamb side of things, but so do the other three.

So what does this mean? Is ZywOo baiting? Maybe, but that is a good thing. Most teams have AWPers in the top left, with HEROIC and Casper "⁠cadiaN⁠" Møller as the exception, which is exactly the zone you want your best player to be in. This data is also from the entire year, meaning that Lotan "⁠Spinx⁠" Giladi is considered part of ENCE. When we just use data from ESL Pro League Season 16, where Vitality were victors, the chart looks more conventional.

At BLAST World Final Natus Vincere were more defined than their year-long averages suggest, too. Valeriy "⁠b1t⁠" Vakhovskiy and Oleksandr "⁠s1mple⁠" Kostyliev both operated in the top left as traders, with the other four players in the bottom right as Lambs. Given that they shuffle lurking roles between maps and do not have a super aggressive player like YEKINDAR, it is no surprise that they do not have a classic bottom-left corner Lone Wolf.

Where we come unstuck is with HEROIC. But, arguably, this just shows the system is working: HEROIC are in many ways a unique team. Take it from Ismail "⁠refrezh⁠" Ali: "A lot of us were sharing the same role. We all had to take responsibility and be able to make calls, to come with a lot of input. cadiaN as the AWP and IGL has… a really defined role, same with Martin "⁠stavn⁠" Lund as the second caller. But the other three have the same responsibility."

"Other teams have an entry fragger, for example rain and karrigan on FaZe. On Heroic it is pretty different. Everybody is going to entry-frag at some point. There are not clear lurkers and clear entry fraggers, it's more dynamic. When they are firing on all cylinders Heroic are really hard to play against because everybody is capable of doing anything."

It is a feature, not a bug, that all of HEROIC's players occupy the top right of the chart. The Danes consistently top the deaths traded statistic on both sides of the map, it is only natural that their graph is different to most teams. stavn's star status is still illustrated by veering to the left, the only player to leave the diagonal trend line; but even he is traded more often than the average player.

And this is where this approach actually has the most value. It is confirming the narratives that are already present from analysts' eye tests, and assigning players to groups based not on overarching ideas of the meta composition, but their own specific team and system.

What is so impressive is that HEROIC are not cheating their numbers. They are not grouping up as a five-player group out of spawn. They still default and are able to throw in slow rounds. What separates them from the rest is the ability to group up so efficiently round after round. No wonder they are so hard to play against when their communication is on song — it is not often you get a 50-50 gunfight against a team like this.

For 2023, the challenge is to maintain the effectiveness of their quirky style. HEROIC are notorious for their aggression, particularly their double or triple set piece pushes on CT-side. They share the frags, the entry duties, and the limelight. It is all about the system, about the team. They are one clenched fist rather than five fingers and that approach has gotten them to No. 1 in the world.

But that is the easy part. It is one thing to reach the summit and another to stay there. One thing is for sure, though, they will live and die by their demanding style. The approach will not change, whether they stay at number one or not — cadiaN's machine will turn, the pushes will keep on coming and the trades will continue to arrive. Can those trades be converted into more trophies, or even an era? We will just have to wait and see.