For the past 5 weeks, ending last Sunday, I have been collecting data on hero popularity in Very High Ranked games, to see how much they change from week to week. Players tend to lust for MMR, and will do almost anything they can to try and increase it. This is especially so among the top 9% of players, aka Very High. With DAC running in the past few weeks, I’ve noticed dramatic changes in the hero pool from week to week. And all of it seems to result in far less enjoyable games.
Gainers
Above are the 6 biggest gainers in popularity for the 5 week period, including their increase from week to week. Someone like Shadow Fiend was only in 15% of games at the start of the experiment, but as the weeks went on he skyrocketed (by 26 raw percent) to being in 42% of games. You’ll notice across the board that the heroes that become more popular, are the heroes that people believe to be the most OP right now. People want to gain MMR, and the easiest way to do that is to pick a strong hero. Unfortunately this results in utterly one dimensional games, where the same heroes appear constantly with little to no variety in hero picks. But how one dimensional are they?
One Dimensional
So 5 weeks ago, Juggernaut was in 56% of all Very High ranked games, and it fell off sharply to Axe in 38% and Slark in 32%. But as the weeks went by, you can see how one dimensional the hero picks become. There used to only be 1 hero with more than 40% pick rate, now there are 3. While there used to be 8 heroes above 20% pick rate, and now there is actually 11. What’s happening is that people are deserting heroes across the board, and picking more and more of the top heroes. Week by week, we lose massive amounts of hero diversity, as players prefer the top 20 heroes over the other 86. Note that this experiment was done prior to Winter Wyvern.
Frequency
At the start of the year, the top 10 most picked heroes were picked the same amount as the bottom 63 heroes. In just 5 weeks, this has increased to 74. So the top 10 heroes are now picked as much as the bottom 74 heroes. The top 10 heroes now make up 33% of all picks, 22 heroes make up the middle 34%, and 74 heroes make up the bottom 33%. This means that you’re likely to see at least 3 of the top 10 heroes every game. Hero diversity is plummeting recently, and every week that went by made it worse. The actual heroes varied as people shifted to the new 6.83 meta, but as far as hero diversity goes on a daily basis, we are seeing an ungodly amount of the same stuff.
The most bizarre thing is that hero win rate is sometimes irrelevant to a hero’s popularity. Across the board, the heroes that aren’t played very often generally have terrible win rates, with the least played hero Enchantress boasting a terrible 36% winrate in Ranked Very High. Alchemist is closely ahead at 39%. However we still see popular heroes that also have bad or mediocre win rates. Doom was popular for weeks, even though he only had a 44% win rate. Slark is also similar, he is insanely popular, but only has a 47% win rate in Ranked Very High. What appears to be happening is that people felt that Slark and Doom were OP, even though their win rates have been mediocre for a while. They give into the circlejerk of OP/Dumpster heroes, without actually looking at their own games or pubs as a whole. Luna, Treant, and Night Stalker are almost never seen in these games, but all have over 50% win rates. Luna is a carry that fell off extremely hard in the pro scene post TI4, but in pubs she still wins about the same as Axe and Shadow Fiend, and wins more than the popular Slark, Storm Spirit, Void, Tidehunter, and Skywrath.
I averaged the pick rates for every group of 10 heroes, and compared the weeks (explained below)
Who’s Sapping my Fun?
With so many games of Dota 2 under my belt, I have found that the game is more enjoyable with variety. We all want to win, but we’re not playing for a million dollars. We are ultimately playing in our free time, and part of that is wanting to have fun. I personally don’t have much fun when one hero is in 60% of the games, and then another is in 56% of them. Having to play every second game with the same 2 heroes gets old really fast. Pudge was the most popular hero in Dota 2 for about 3 years straight, but he almost never was in more than 40% of games. Three heroes top that right now. In all Skill brackets Pudge has actually been bumped down to third by both Sniper and Juggernaut, as 6.83 swings into full force.
In the graph above, I averaged the win rate for the top 10 heroes, and then compared it to the average win rate of heroes 11-20, 21-30, etc You can see that the top 20 heroes gained in popularity, while the bottom 80 dropped. The 30-50th most popular heroes were hit hardest, really gutting the ‘often picked’ heroes.
How To Solve?
After looking at all the numbers for weeks on end, I think the BanPick game mode is honestly the best solution. That is a game mode where everyone gets ~20 seconds to ban 1 hero each, and after that we go to the normal picking phase. If you didn’t ban, then no one is banned. I think this would not only make the game more enjoyable, but make the pool more diverse. People are likely to either ban heroes that are played way too much, or have insanely high win rate. You could ban counters, but more often than not you’re just sick of the ‘spin-to-win trio’. Besides, if matchmaking does it’s job then similar size parties should be matched together, and thus will have the same advantage in any possible ‘counter-banning’. Besides, a hero counter will always slip through the cracks.
BanPick also has the distinct advantage of being similar enough to AllPick that people will actually play it. Matchmaking in CaptainsMode is generally far worse than in AllPick, especially on smaller servers like Australia, since there is a much smaller amount of people searching in that mode. BanPick being similar to AllPick works in its advantage, as people will actually queue for it, giving the gamemode a big enough player pool to get close MMRs quickly.
Closing Thoughts
This patch, 6.83, has resulted in many things to dislike. I think overall though, the plummeting hero diversity is what I dislike most. People are still shifting heroes from old meta to new meta, so we do see some changes in heroes week to week. The heroes are starting to stabilize though, as DAC ends, and overall the heroes picked are becoming increasingly one dimensional; creating some awfully repetitive games. People want to raise MMR, so they are picking what they think works (the same stuff), which I believe decreases the enjoyment of the game. This appears to be pushed to the extreme in 6.83, more so than in previous patches. Dota does have pretty good hero diversity, but every week of 6.83 that passes, is a week where we lose hero diversity. In week 1 there were only 3 heroes appearing in less than 2% of games, this increased to 9 heroes. Both high and low, we are seeing massive shifts towards more of the same. Don’t forget that the old Dota strategy is to stop the snowball, before it becomes too big to handle. At the moment, we seem to be ~20% less diverse than we were 60 days ago.
Zota 