Hero Winrate Differential 7.07 vs 7.06

With the patch out for ~48 hours we can start to begin to see who the biggest winners and losers are. Edge Analytics is able to provide insight into major patch changes within 24 hours of a patch. This particular analysis focuses on how each hero’s winrate changed from the 7.06f patch to the 7.07 patch. The focus of this analysis is on 5k+ MMR pubs.

7.07 was particularly disruptive, and a lot of our notions about the patch will turn out to be incorrect. This is why it takes many weeks for certain strong heroes to enter the meta, and we can expect this phenomenon to occur even more with a patch like 7.07. Some of the prevalent trends in high mmr games include increased picks of Bane and Antimage. We have also seen the dominant 7.06f dominant heroes such as Bloodseeker lose their standing in contested picks. The data confirms that Antimage was buffed pretty strongly and strangely enough Bloodseeker, a hero who was expected to receive a nerf, received a small buff in the form of new talents. Ironically, Bane took a hit after the patch which makes the requests of nerfing Bane a moot point.

Take a look at the visualization below to compare how well heroes were doing in 7.06f compared to the 7.07 patch. This graph shows the difference in winrate from each hero’s 7.06f winrate and their corresponding 7.07 winrate. Positive values indicate an increase in winrate from 7.06f to 7.07, a “buff”.

DOTA2 Hero Counters

One of the core components in DOTA is to pick heroes that work well against your enemies. In order to simplify our discussion, let’s assume we are talking about counterpicking against a single hero. There are two ways to think about doing this. One is to select heroes that win more often than the enemy when they are matched together. This is often a good method to predict the result of how that matchup plays out, but it  leads one to the conclusion that certain heroes are good “counters” when they actually aren’t. Take, for example, Necrophos in the 7.06f patch. He boasts a winrate higher than 50% in almost every single matchup, and by our previous definition we would come to the conclusion that he has almost no counters.

In an effort to understand which hero’s kits work well against each other, regardless of whether they are flavor-of-the-month, another approach was taken. A hero’s advantage was determined by the impact they had on the enemy’s expected winrate. Going back to our previous example of Necrophos, if we want to see whether Antimage is a counter, we compare each hero’s expected winrate across all matchups and see how much they are impacted by the matchup itself.

Necrophos’ expected WR across all heroes: 56%
Antimage’s expected WR across all heroes: 50.7%

Antimage’s winrate vs Necrophos: 49.2%
Necrophos’ winrate vs Antimage: 50.8%

Even though Antimage loses this matchup > 50% of the time, we can see that Antimage is a counter to Necrophos due to his impact on Necrophos’ winrate, which on average should be 56%.

Using match data from 70,000 5k+ mmr games, a graph was created of each hero’s average “advantage rating” against the top 5% , 10%, 20%, and 35% biggest counters. It’s important to note that the rating is negative due to the fact that the each hero has a negative rating against counters. The better the counters the worse that hero’s advantage rating will be.

For clarification, when we refer to something as the top 5% of counters, we are referring to the 5% quantile of all of the heroes. It does not refer to an average of the top 1-5%, but serves as a marker.

The purpose of dividing it into tiers is so that we can look at how they fare against their best (top 5%) and their general counters (35%). Some interesting conclusions come out from this tiered division. For instance, we can look at Dragon Knight who doesn’t particularly stand out against his general counters (35%, 20%), but faces huge problems against his top 5% best counters. This is something that holds a lot of weight in pro level matches because it can develop into several strategies. The apparent one is to completely avoid heroes that posses very low advantage scores against their top 5% of counters. Two, it’s possible to draft around the idea that you are picking high risk heroes and need to make the necessary ban/pick adjustments to take those heroes outside of the pool. This information is also applicable in pubs where you want to minimize the picks with a lot of counters or a few really strong counters. These picks can be reserved for later in the draft where you have a more clear idea about who has been picked and how many picks are left.

We also have a very interesting outlier in the graphs with Spirit Breaker. He appears to hold an advantage against the top 35% and top 20% of his “counters”. This is due to the fact that he has such a small amount of counters that over 80% of the heroes in the pool he holds an advantage against. This leads to heroes that aren’t actually considered counters showing up in the quantiles of 35 and 20. Spirit Breaker’s versatility against heroes is further accentuated by how well he does even against his top 5% best counters. This advantage score in the 5% quantile is matched mostly by other heroes when facing their top 35% counters.

 

— Update — 

A new method is being using to calculate advantage rating to address some of the short comings of the one illustrated in this article.