Understanding the Data of Dota 2 with Kpoptosis

by Bryan "Kpoptosis" Herren Feb 6 2015
Thumbnail image courtesy of Beyond The Summit

A common conversation in a traditional sports-watching household looks like this:

“Oh hey you’re watching the Super Bowl? What’s the score?”
“Seahawks are winning, 24-14”
“Oh cool.”

A common conversation in the Beyond the Summit Dota studio house looks like this:

“Oh hey you’re watching the Na`VI vs Team Secret Game? What’s the Score?”
“Um... Secret is leading in kills, but Na`Vi has the net worth lead, but Secret has a barracks advantage.”
“Well who’s winning?”
“Hard to say. Na`Vi I think, they have the better late game lineup and we’re nearly 50 minutes in. Secret could win if this push goes well though.“

A major factor that makes ARTS/MOBA games stand out from traditional sports (and even other eSports) is that there is no single metric that inherently correlates to winning the game. In American Football, you win by scoring more points. Likewise, in Counter-Strike: Global Offensive, you win by taking the most maps. 

These are metrics that update throughout the match and give a relatively clear picture of the game’s development. In Dota, the only thing that wins you the game is killing the enemy’s ancient. The problem is, that only happens when the match is over.

This affects aspects of Dota even beyond the fan experience. What drew me into my current role in Dota was a love for statistical analysis in traditional sports, specifically baseball. 

For years, baseball was stuck in a rut of prioritizing bad statistics to correlate with performance. For instance, Batting Average was seen as the main metric for determining offensive skill, despite it not taking into consideration such important aspects as walks and extra-base hits. 

In the 1980s, a man named Bill James came along and turned the baseball world on its head by suggesting a surprisingly simple idea: scoring is everything. By studying how effective players were at scoring or protecting scoring from happening, he was able to usher in a new age of analysis that focused on statistics that were directly correlated to winning games. Due to Baseball’s one dimensional scoring, James was even able to show how many wins a player contributed to over the course of a season.

I got into Dota statistics dreaming of doing what Bill James did. The reality was a bit harsh, as Dota is a much more complicated game. As I’ve already stated, no metric in Dota correlates to winning the game 100% of the time. Kills, Gold Farmed, Towers Taken, Roshans Killed, Damage Dealt and Taken —  all of them give insight into who has the upper hand in the match and how, but none of them are surefire ways to win every match. This affects Dota in a very powerful way when it comes to metrics and box scores. 

Therefore, the following are three things you should bear in mind when watching a Dota match:

1) Dota live game stats are holistic reports.

When watching a game live, reading a box score, or a live report for a Dota match the only way to see how a teams is doing is to take it all in. Are they behind in kills but they have a Medusa that’s the most farmed hero in the game? They probably have a chance to pull it out. 

In my opinion, the big three metrics that have the most insight into the game trajectory are Net Worth advantage, Tower advantage, and last (and least) Kill advantage. If one team is leading in all three areas, there’s a very good chance they’ll go on to win the game. But in the end, it’s Dota — nothing is guaranteed, there are also other things to consider.

2) Individual player metrics are insightful, but often misleading

The player with the highest career Kill/Death/Assist ratio (KDA) in Dota 2 is Sylar, an accomplished Carry player considered by many to be one of the best in the world. In second place is Corey, an exceptionally talented Dota player in his own right, but one who has had an unsuccessful Dota 2 career when compared to the International 4 runner-up. 

There are multiple factors that go into determining a player’s KDA, such as the play style of the player’s team (Sylar has often played in “four protect one” lineups where every other player on the team worked to make him look good) as well as standard of competition (Corey played in Korea for a years, where the competition pales in comparison to more Dota developed areas such as China or Europe.)

Additionally, when it comes to roles such as support, stats do a particularly poor job of describing their contribution to the team. For example, you would think not dying would be important as a support, but the Maelk Award exists to disprove that that is always true.

That's not to say that player specific stats are meaningless — quite the opposite — it just means you have to look at them like you look at live reports and box scores, holistically. A high GPM (gold gained per minute) may not correlate to winning 100 percent of the time, but it does illustrate that a player farmed particularly effectively. Additionally, high assist numbers on a support may mean he was very active in team fights and teleporting to help allies fight instead of prioritizing his own farm.

3) Almost every statistic in a game is relative to the draft

One of the few constants in Dota is the hero pool. For every pro match, before anecdotal situations and non-quantitative value is shown, the draft occurs. The same 109 heroes (for now) are available to all teams in every match, and the metagame in Dota is a vibrant and constantly in flux. 

For some heroes, such as Storm Spirit, having a very good KDA early in the match is a big indication that the game will go well for his team. If a team has Medusa, being behind in Towers and Gold late in the game is a lot less scary as long as the Medusa has farmed decently well. A hero like Vengeful Spirit has two abilities that reward putting yourself in harms way for an allied hero, therefore a high death score for that player is usually not a bad thing. 

Once again, nothing is certain! Medusas do lose 60+ minute games from time to time, and sometimes Vengeful Spirits aren’t helping their team — they’re just feeding.

Hopefully now you have a slightly better idea how to review and follow games on recap sites and apps, such as theScore eSports! The fact that Dota matches are harder to read than traditional sports might seem a bit overwhelming, but it’s representative of just how deep, complicated, and beautiful of a game Dota 2 is. Happy watching.

Brian "Kpoptosis" Herren works for Beyond the Summit as a producer and statistician. You can follow him on Twitter, and watch the Dota 2 Asian Championships on Beyond the Summit's channel.