Analyzing game data through correlations
How important is economy in Counter Strike: Global Offensive?
In a previous post we looked at correlations between DotA and LoL objectives and winning the match. This time we look at what it takes to win a CS:GO round. For this analysis we took 3551 professional CS:GO matches starting March 16 2019*. That corresponds to a total of 92,596 rounds! In total, terrorists win 48.25% of the rounds. Bomb is planted in 53.84% of all rounds, defused in 24.71% and explodes in 32.67%. 69.11% of all round end with 5 kills.
There is plenty of information that we can see in just one plot. For instance, look at the strong correlation (0.71) between terrorists winning the last round and having an equipment advantage in the current round! It is then hardly surprising, that there is a significant correlation between winning the last and the current round (0.33). What we also see is that bomb exploding and terrorists achieving five kills are two almost mutually exclusive events (though of course it can also sometimes happen), resulting in a negative correlation (-0.31). That's also why the correlation between winning and achieving five kills is also "only" 0.69 - you can win a round without making a single kill! Here is a histogram of round win reasons, showing how comparably few rounds are decided by the bomb as opposed to kills:
Looking at feature correlation is just a first step in data analysis but a very important one. There is a lot more information about the game in these plots than explained above. What else can you find?
*This day was chosen due to the fact that a significant update to the CS:GO economy happened on that day.