Advanced analytics for CS:GO
In this article we discuss some features that could be added in the future to truly bring advanced analytics to our favorite game.
Though it's unquestionable that statistics have furthered our ability to analyze Counter-Strike, just like any sport, there are still plenty of critics who do not believe numbers capture the true essence of the game.
They aren't exactly wrong - there are a number of things that numbers cannot, and likely never will be able to catch - but features could be added that would make statistics much more powerful tools in analysis.
After tweeting a few ideas during the playoffs of SLTV StarSeries XII Finals, I have included some more suggestions in this article for metrics that if added would paint a better picture of players and teams.
When extending these statistics to teams as a whole, we could characterize teams using the advanced statistics and better understand what it is that makes each of them tick.
Statistics aren't meant to replace watching matches. However, no one in the world can watch every point of view of every match, and that's why we need statistics to better understand what we can't see.
Current statistics are good, but we shouldn't settle for good
Crosshair placement
Currently this is one of the skills that is only judged from watching the games. There is nothing wrong with it, but seeing as it is one of the most important skills in the game - and one that can be a huge difference maker among similarly skilled players - it would be interesting to be able to track it with more accuracy. We could get that metric by implementing something that calculated the number of pixels a crosshair is moved on average within X seconds before a kill. Similar info is already presumably used in the "X saved Y" feature.
On the contrary, high values for crosshair movements before a kill would showcase a precise flick aim, something that would surely allow us to evaluate more erratic players such as Aleksandr "s1mple" Kostyliev a little better. In either case, it would be interesting to analyze different aiming styles and when evaluating young players who are just getting started, it could be one of the future metrics for trying to figure out which of the youngsters could actually make it big in the future.
Ease of kills
Right now we know players like Christopher "GeT_RiGhT" Alesund and Vincent "Happy" Schopenhauer are primarily lurkers and get an awful lot of what we deem as easy kills. But those kills are easy, for the most part, only because their positioning is so smart that they get to shoot at people's backs or their sides. The way we could measure this would be to add a metric that would judge whether a kill was scored from outside of the opponent's field of vision, allowing us to calculate statistics around it.
A second way to interpret this statistic would be finding out which players, if any, are only able to get these easy, or smart, kills. There are surely players who get the most important and impressive kills - such as entry frags - and there are others who sneak by with the easy kills, yet wind up with similar numbers on the scoreboard. This kind of metric would help us separate those kills.
Kills versus saves, force buys and buys
With money spent now a common figure in the Counter-Strike spectator HUD, it should be a no-brainer to finally separate kills by how much money a player's opponent had spent on equipment and guns that round. We could then find out which players are eco-cobras in feeding season, and which ones actually tend to contribute most against well equipped opposition.
In addition to figuring out an opponent's equipment level, equally interesting would be to find out how much money the player scoring a kill has spent. By getting this data, we could then figure out just how effectie certain players and/or teams are on those second round buys in comparison to their opposition, whether Na`Vi really are the force-buy kings with their Galils, and more.
Flashes leading to a kill
One of the statistics that could showcase the value of role players much better than anything we have today, we should be able to calculate how many kills were scored while opponents were blinded by a player's flashbang. This way we could directly credit players for the so far unthankful job of holding out a pop flash for their teammates, who then get to enjoy the fruit of their labor.
In a case where the player scoring the kill were not the one who threw the flashbang, it would not be wrong in my opinion to also credit the flashbang throwing teammate with an assist, making that statistic more interesting and also much more relevant than it is today.
Time spent/deaths flashed
A fairly minor statistic that wouldn't be game changing, but if we knew the time in seconds a player spent flashed - and the deaths that took place while the player was blinded - it could give us a better understanding of players who are able to best avoid opponent's use of flashbangs. Position plays a huge role here - and this would need to be broken down by half - but it would give us more knowledge about individual players' styles.
Overall grenade damage dealt
Throwing flashes and smokes is a fairly unappreciated job, but the metric described above would let us at least judge players who help their teammates out. In addition, we should be able to judge the effectiveness of a player's HE grenade and Molotov/Incendiary grenage usage by tracking how much damage they deal using those grenades on average per round, and per throw.
The former would better showcase the effectiveness in the sense that sometimes you may buy them and not be able to even use them, so you would get punished for that - whereas in the second you could only see how much damage is actually achieved per each grenade you use. Both would be interesting metrics, and would give us a better understanding of one of the many variables that are currently very hard to judge.
Average Damage per Round statistics
This is a statistic that existed during GotFrag's GameSense days - and both CEVO and ESEA still include it - but was scrapped later on because while HLTV.org started parsing statistics from demos, getting the ADR information required server logs, which were a scarce commodity in the late Counter-Strike 1.6 days. However, today virtually all statistics are parsed from server logs, which should give us a chance to bring back the ADR numbers, thus showing us which players deal the most damage to the opposing team.
Entry frags statistics
Entry frag statistics are often quoted before important matches on our Twitter account, but they must become a standard in the future so we can get a proper breakdown of entry fraggers. Standard profiles should include at least a player's entry frag statistics on both sides - including the all-important ratio, which is what Adam "friberg" Friberg, a volume entry-fragger, is often criticized for - and possibly a weapon breakdown as well.
Most importantly, and this is something that I will touch on more later on, is we should be creating heat maps out of entry frag statistics to find out where players are scoring their entry frags and where they are dying. Building teams and recruiting players would be a lot easier if things such as this could be figured out from analytics before actually spending time together on the server, as it would give players a good understanding of others' capabilities beyond their selective memories and highlights clips.
Kills within X meters of a smoke
This metric could judge how effectively a player uses smokes to their advantage. In essence, it could show whether smoke play is a net-positive or a net-negative for a player by showing how often a player dies within close range of a smoke versus how many kills a player gets within close range of a smoke. This may not sound that interesting to some, but players such as Niklas "niko" Johansson exceled within smokes in 1.6, and in CS:GO they play an even larger role.
Passive or aggressive fraggers
A metric to measure how much a player moves - excluding strafing back and forth, but rather in terms of actual movement on the map - right before a kill on average could give us a better understanding of what type of players people are. This statistic would give us an idea of whether a player is the passive type that prefers waiting for kills to come to them, or the aggressive kinds who actively go seeking for kills around the map.
In addition, with a heat map system figured out properly, we could set up certain zones across the map - for example, by how long it takes to get there from a spawning point - and calculate players kills in all of those areas. It would then allow us to separate aggressive AWPers such as Jesper "JW" Wecksell from far more passive ones, such as Aleksi "allu" Jalli, using statistical measures.
Anti-AWP specialist
This metric could be split into two halves - one for rifles, and another one for snipers, including the big green - but in essence it would track who are the most efficient players in taking down AWPers. If we went into detail, it could still be broken down to two further halves, one in which you only count kills against an AWPer who has his scope out at the time of the kill, and another where he has taken his AWP out in the past X seconds to make sure it's still a relevant situation.
With many of the world's best players now at least adequate AWP users, if not full-on snipers, it's important to be able to win duels against the snipers or at least neutralize them using other measures. As such, a statistic to measure just that would give us valuable information that could be used to judge players, and perhaps even more interestingly, teams, who are best able to take out their opponents' AWPers.
Accuracy of spray
We often give huge amounts of credit for players such as GeT_RiGhT for their spraying skills, and rightly so - only selected few were able to control their bullets as accurately as him, but with more than a decade of Counter-Strike under our belts the best way to measure that is still the old eye-test. That should not be the case, as it should not be too hard to implement a statistic to properly measure this as well.
Instead of watching demos, replays or highlights to judge how goood players are at spraying, we could simply add a metric that would judge how many bullets - over a pre-determined limit of what is conceived as spraying - it takes a player on average to score a kill. Naturally this would have to be broken down per gun, but would it not be interesting to find out who are the best sprayers with a FAMAS, a Galil, an AK47 or either of the M4s?
Traders and tradees
One important aspect of terrorist side play is being able to trade kills. That involves two players - the one going in first and dying, as well as the second one trading the kill. Both are important roles in being able to take over areas of maps - which translate directly to round wins. It follows then, naturally, that we should be able to track how successful players are in both roles - and how often they play either role, giving us some team-wide information on styles as well.
If we know how often a player trades a kill we'd obviously understand his ability as a trader - though we'd have to carefully decide on the criteria of how close said player is to his teammate dying - and on the other hand, getting traded is not an issue on the terrorist side in even numbered situations, which means if a player never extends too far to not die without getting traded it's valuable for the team. This is one of the statistics that fans bring up to support entry-fraggers, so we should be able to properly measure it.
Pointman into bombsites
Fans often make excuses for players by saying they always run into bombsites first and therefore cannot be expected to put up big numbers. The trading kills statistic would already address it partly, but in general it would be interesting to know how often a player enter into a site from a given chokepoint first on his team. It's a statistic to be combined with e.g. ADR or KPR (kills per round) to paint a full picture, but it would already be interesting to know how late players in general enter bombsites to define roles more accurately.
Bomb plants in A and B
This is hardly an advanced statistic at all - yet it currently is not being tracked - and it would add valuable information to fans, casters and followers. Knowing how often any team plants the bomb in either bombsite would give us the split in how often a team attacks either bombsite. In any given match it wouldn't add much value, but we would learn overtime clear trends of which bombsites on each map every team prefers, thus giving us a better understanding of their gameplan.
Fake salesman
Another role that gets no mention on the scoreboard is being one of the players selling a fake in another bombsite, thus almost guaranteeing a sudden and unpleasant death, while his teammates gain free entry to the other bombsite and thus get to enjoy the advantage of opponents having to run to their screens. By setting up entrance trackers to bombsites' choke points, which would be used to track who is first into sites as well, we could see how often players fake other bombsites in a given match.
This metric would require some careful planning to make sure it tracks what we're hoping it would track, but once accurate enough, it would bring us a whole new level of detail to player roles and their importance to teams outside of what is visible on the scoreboards.
Area of kills and deaths
This is possibly the most interesting statistic out of all the ones I have listed here, and pre-existing heat maps could already provide us them. If we could generate heatmaps as we wanted from a specific set of matches, we could find out first of all which areas of maps players are most effective on, which they aren't so strong on, which players generate most of their kills as terrorists after entering bombsites, in after plant situations, and which ones clear the way for their teammates.
It would also help us see clear trends in teams' play, and would help teams' coaches come up with better gameplans against their opponents, thus leading to a higher level of play in general among the competition. Another way to use this data would be to compare players strengths when building teams to make sure everyone's favorite killing spots do not overlap, thus making sure that a team would be well rounded with that set of players.
Heat maps already exist - we just need to make them useful
Together with these new metrics and changes, it is also up to us to revamp our system so that it's easier to dig up all of this information - and filter it using a number of criteria, including country, age, and all statistics.
Got other ideas that would measure players skill or contribution to their team wins not mentioned here? Leave a comment below to share your thoughts and some may end up as reality.
Follow HLTV.org's @lurppis_ on Twitter.