Mythbusters: Is the Hero Pool Really Getting More Diverse?

There’s an old theory with Dota heroes: every hero is viable in pubs, but far less are viable for pros. More recently casters and players have been noting how almost every hero is viable now, how every patch seems to expand the hero pool to make more and more heroes legitimate picks. But is this is all psychological? They’ve seen games come and go, but they’ve never sat down and compared the numbers. So I’m going to see whether the hero pool is statistically more viable now. Is the hero pool more diverse than years gone by? Or is it all in the mind?

The 40 percenters

Here on Zota I like to use the term “top picks”, that is a hero who has more than 40% ban/pick rate in the pro scene. I record them each month in an easy to find and search page, and like to consult it when curious as to the last time a hero was popular. So let’s start with this, I have 25 months of data here, dating back to December 2012.

What we see is that for the last 25 months of professional Dota, there has been about 15-20 top picks every month. At the moment Brew, Razor, and Lycan, whereas 12 months ago it was Visage, Lifestealer, and Dragon Knight. We see a constantly rotating roster of top picks, but the raw number of them doesn’t change much. They’ve gone down ever so slightly, but the trend is awfully mild from this metric.

Month to Month to Month

I want to compare December 2012 with December 2013 with December 2014. Three consecutive years of professional Dota could give us an insight into how the hero pool has evolved over the years. The difficulty with doing this exercise is that the hero pool has expanded over the years. In 2012 we had only 83 heroes, 2013 was 97, and 2014 is 106. This makes it a struggle to compare, but I think the best way to do split them into percentages of the hero pool based on their pick/bate rate. Top picks are 40% or more, Popular are 20-40%, Unpopular are 5-20%, and Outcasts are less than 5%.

In December 2012, we had 25.3% of the hero pool as top picks, but heroes fell off hard though, with only 13% being popular. We then ended up with 25% of the hero pool as unpopular, and an absurd 36% contested in less than 5% of games. You can see why I’ve got this ultra low outcast category, because in 2012 a third of the pool were very rarely seen. This is what I refer as a skewed hero pool: the top 25% of heroes are responsible for the 70% of all picks and bans. It’s like the Dota equivalent of wealth inequality, we have hero inequality where the chosen few are constantly favored while a third of the heroes live in the slums; outcast to the world. But let’s see how this compares to 2013 and 2014.

The top picks decrease over time. You see, while the raw number of top picks is staying similar as shown far above, since more heroes are being added to the pool it becomes a lower percentage of the pool. But all these new heroes aren’t changing the dynamic of the hero pool too much.

Looking at the raw data, you can see that the new heroes just end up padding out the bottom of the pool. The first tier of heroes are very similar, while we do see a big increase in second tier heroes, which is a big positive. But the majority of heroes seem to end up being rarely picked. In 2012 60% of the heroes are in the bottom two tiers, while in 2014 66% of the heroes are.

The Verdict

Are more heroes becoming more viable? Yes, but it’s very mild. We still have about a third of the pool rarely picked, and we still have around 18 heroes being top picks. It seems to me that the only thing about the hero pool that is more diverse, is that when someone is not picking a top hero, they have a bunch of heroes to choose from. In that regard you are seeing a lot more diversity, the problem is that we still are seeing a large amount of drafts central on the top 40 heroes. In 2012 we had the top 25% of heroes taking 70% of all picks and bans. In December 2014 we had the top 25% taking 66% of all picks and bans. The hero pool has expanded so you do get more heroes in the top 25%, which should be noted. It is more diverse than before, but it’s a very mild trend.

Basically you have chicken for dinner 10 nights a fortnight, on the other 4 nights you alternate between beef, pork, salmon, and kangaroo. You eat a wide variety of meats overall (moreso than the 11 chicken nights two years ago), but there’s no denying that you’re still primarily a chicken guy. That’s the Dota 2 hero picks in a nutshell, mostly the same stuff, but a lot of variety on the odd chance you do want to mix it up. The hero pool is 27% bigger than 2012, but that’s just 27% more stuff to choose from on the non-chicken nights. That’s a pretty solid analogy of what’s happening.

Percentage of the hero pool that is picked outside of even distribution, with 10% leeway each side of even distribution classified as ‘Just Right’.

Preconceived Notions?

I think the explanation as to why people feel that almost every hero is viable ultimately comes down to perception and psychology. Most people reading this know about those filthy Mirana pickers, but in December Mirana was only drafted 12% of the time. She fell off hard in the pro scene, but you still remember what she’s capable of. What I believe happens, is that because you remember when a hero was top stuff, you think better of the hero for months to come.

You’ve seen Miranas have massive game impact, so even when they fall off in popularity you still rate them highly. Given enough time watching pro Dota, the natural cycle of heroes will result in you seeing almost every hero have their day. When a Hero is incredibly popular, you see the hero’s strengths at work, and even when they fall off, you still remember their strengths. Rarely some heroes get nerfed into the ground like Alchemist, but many heroes like Sand King fall off massively without anything happening to them, or just very mild nerfs.

It’s about perception. You perceive a greater amount of heroes to be viable now than a year ago, because you’ve seen new heroes work in professional settings. In recent months, Zeus has come from 3 years of being ignored, to being contested 33% in December. Even if he falls off, you’ll remember his strengths. Mjolnir’s Static Charge had a similar fate, people ignored it for years, even though it had its moments. Suddenly it becomes top stuff with a buff, and even though it was nerfed later on, the Dota 2 community now has much more awareness about it. You becomes aware of strengths and potential when something is popular, and you remember this for years to come.

Previous Article: Juggernaut’s Fury, running the pub numbers of 6.83 Juggernaut


I made a bold move and created a Facebook page. If you would like to be notified when a new article is posted, or simply like the content that I put out, please consider liking the page. It will really help in keeping this site going.

5 thoughts on “Mythbusters: Is the Hero Pool Really Getting More Diverse?

  1. Uh that’s not how you measure diversity. Each game can only pick/ban 20 heroes, which makes the number for 40%+ pick rate heroes limited. And we will always have OP heroes which take more space in distribution.

    Let’s assume each heroes is picked/banned equally, which means max diversity, it will result in everyone has bp rate of 20/106=18.9%, none is 40%+ – which in your theory implies worst diversity.

    The way I can think to measure diversity is by calculating the standard deviation of pick/ban rate of every heroes.

  2. I’ve already calculated the gini coefficient for every month of pro dota. They’ll be on joinDOTA in my next Metagame Fortnight editorial. Until then suffice to say that the data that I’ve found indicates that compared to last summer, drafts are significantly more dominated by a wide arrange of lesser-picked heroes and fewer highly picked heroes in pro dota. There’s also a consistent downward trend to the percent of total heroes picked who are played in 75% or more of games.

    It’s not the number of heroes that matter in terms of the diversity of the hero picks. It’s the percentage of draft picks that is dominated by high-tier heroes. If 100% of heroes drafted are in 100% of games (perfect inequality), then the balance is very bad. If 100% of heroes are played in <50% of games, the the balance is relatively very good. Not every hero will be viable, but having lots of heroes see game time who are not dominating is a sign that most heroes are situationally viable, which is the goal.

    Currently, only 5% of draft slots are taken by heroes who see 75% or more of games (much lower than 2013 where it was over 20%). Almost 75% of draft positions are taken by heroes who represent fewer than 25% of games–this is an all time high. The game is currently DOMINATED by tier 3 picks when you look at the numbers–in fact, they are impacting significantly more of the draft than tier 1 and tier 2 picks combined (which has not always been the case).

    I'll have more on this with my next metagame fortnight on joinDOTA.


  3. Dont you think that the huge influx of games ,over last year and 2 years back, played in comp dota is a variable to your calculations? Especially on the subject of picks? All teams will try to copy and steal ideas from other teams but this is something we see alot of out of smaller teams all the way down to amateur leagues. While creativity and new ideas/picks can come out of anywhere, I believe that t1-t2 teams are showing alot more strategies than lets say 6.81 where the list of first ban heroes was like 5 heroes out of which 1 was banned almost every game,Lycan. Now we see teams first banning alot more on based on who they are facing, sure that was happening in the past too, but I believe the your data is skewed heavily by t4-t3 teams that are trying to break into the comp scene. Just some thoughts, good work, would like to see analyses like this in a format of a video or even a show(stream), maybe weekly or something. I believe that the scene needs this alot. Atm all we have are people who are involved in like 5 different games and all they talk about is drama and ”the esports community”.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s