The Heroes With the Greatest Improvement in High Skill Games

An ugly mish-mash of words that title, but it gets the job done.  Today we’re looking at the heroes whose win rates improve as we move into higher level games, and we’ll be using our fresh sample of roughly 33k/39k/39k games collected over the period of January through March up until the 6.77c patch.

First, a word of warning.  This sample size is the largest yet, but it’s still not perfect.  Some heroes are very underrepresented, such as Chen in Normal only having 927 games.  Win rates also tend to cluster around 50%, so a percentage point of error is far more significant here than it is in usage rates.  So keep in mind that things are still quite fuzzy.  We can be fairly certain of the largest trends, but the smaller trends may just be noise and the precise rankings are not set in stone.

With that in mind, let’s move to the table:


(The slope is just a representation of the relative strength of the increase in win percentage, and Wisp could have just as easily been 5.51 or .055.)

The big mover should be no surprise to anyone.  Wisp (or Io, but I’ll be damned if I’m going to fix all my name references every time Valve changes a hero name) is possibly the most potent hero in the game for turning a small coordination advantage into a win.  Wisp is simultaneously one of the most difficult heroes to play for those in the Normal bracket.  To drive the point home, I suspect Chaos Knight at #11 and Tiny at #16 made the list solely on the basis of being tether buddies for Wisp stomps.

The biggest repeat category on the list are the timing dependent carries, which would include Clinkz at #3, Storm Spirit at #6, and Anti-Mage at #7.  Clinkz is the clearest example as he’s quite simply dependent on farming a quick Orchid so he can dominate the game before his power curve falls off, and  I feel Storm Spirit behaves similarly.   Normal players simply are incapable of pulling this off regularly, and most would lack the sense of aggression necessary to use the Orchid properly even if they could farm it.  Given the level of play in Normal, I suspect that anyone capable of playing Clinkz correctly will soon find themselves no longer playing in Normal games.

Anti-Mage is a somewhat different case, but again, he needs good early farm to get a quick Battlefury and/or Manta Style.  Normal players cannot typically hit this level of farm, and Normal teams are incapable of working together to provide him the space his weak early laning demands.

Chen at #2 and Visage at #5 are simply two rather strong heroes that go unnoticed because of their micromanagement requirements.  Chen is also very dependent upon early pushing, which lower skill teams will typically fail to capitalize on.  I would say that an aspiring support specialist might want to consider practicing Visage, as there’s a good chance he’s very underrated right now.

Nyx Assassin at #4 is interesting in that he is one of only two heroes in the top 10 with over a 50% win rate in Normal.  Like many of the other heroes on this list, he thrives in a more aggressive playstyle than most Normal players are capable of fulfilling, but it’s a testament to his current state that even in the passive land of Normal his win rate is above 54%.  To be honest, I suspect 6.77c wasn’t the end of his patch note presence.

Zeus at #9 is a surprise for me.  I honestly expected his win rate to fall off in higher skill games, and so far it appears that it does not.  Perhaps I am underestimating him.

Shadow Demon (#8), Bane (#10), Pugna (#13), and Rubick (#17) are most likely on this list for having strong but unintuitive kit elements.  Disruption, for example, is a skill that I suspect many Normal players can’t quite wrap their heads around.  Another example is that to use Rubick’s Spell Steal effectively you need to have a fairly good knowledge of every heroes’ spells.  Magnus (#12) also likely belongs here due to how badly people screw up the mechanics of Skewer and Reverse Polarity.

Finally we have Clockwerk (#14), Bounty Hunter (#15), and Queen of Pain (#18).  To play these characters effectively you need to have the map sense to be constantly on the lookout for opportunities for aggression.  You also need a good idea of what kind of solo pickoffs you can mathematically get away with, particularly for Queen of Pain.  They can be reasonably effective heroes in passive games, but they all shine if the player behind them knows precisely what they can and cannot get away with.

I will say that as my slope measurement approaches 100 I feel less confident that the trends are significant.  It’s nonetheless curious to me that Shadow Shaman made the list at #19.  He seems to fallen out of favor with the Western scene, but Chinese teams did favor him a lot during TI2.  Is he still used often in Chinese and SEA tournaments?

I’m not going to comment too much on the bottom 20 heroes.  For one, the bottom trends are just weaker.  Besides that, a lot of what you see in the bottom trends is that heroes that win their lanes through attrition (Kunkka, Silencer, Necrolyte, Viper, etc) tend to outperform in Normal games.  No surprise here, and it’s also not terribly informative.

I will point out that Meepo’s status as worst improvement in high skill games is likely just a result of Normal teams being less coordinated in punishing a Meepo pick through aggressive laning and focus fire.  The only other observations I have is that Sniper and Phantom Assassin’s low positions speak poorly of their viability as carries, and it’s interesting that Centaur Warrunner sees a 56.5 -> 54.5 ->52.5 decline.  This would suggest to me that he might be appropriately tuned for an inclusion to Captain’s Mode soon.

If you’d like to see what I’m talking about here, click here for the complete data from the 6.77 sample.  Win rate trends are on the third tab and are sorted by my slope measurement (which I didn’t include (because I am lazy)).


8 Responses to The Heroes With the Greatest Improvement in High Skill Games

  1. Ewig Custos says:

    These winrates improvements by brackets are quite logical, since the coordination of teammates is way higher there (stacks usually go to high and very high, after all), and it’s nice to see statistical proof of that.
    Regarding Zeus: First of all, his normal winrate is probably affected by fact that his ult is global nuke, and that actually dramatically affects winrate. Just look at the centaur or remember that HUGE decrease of Treant’s winrate when his living armor was reworked. Regarding Zeus on high and very high brackets, I can assume that the reason of it lies in higher team coordination. What can be just harassing with spells on normal can have a follow-up from teammates on higher brackets. Andd since his spells do such a huge amount of damage, this follow-up probably ends in more early kills, and the game just snowballs from there. Even if it’s not the carry which got this additional gold, it means that Zeus got it, and farmed Zeus with Aghanim’s and/or Refresher can destroy everyone in mid game, even from another corner of the map.
    I take my hat off to you, keep up the good work. Awesomeness of stats can never be overestimated.

    P.S. If you have some free time, can you look at what changed to Phantom “Cancer” Lancer since the last dota update where bug with him having 1 more illusion above maximum and another bug with diffusal blade damage were fixed? If you could do it, it would be much appreciated.

    • phantasmal says:

      P.S. If you have some free time, can you look at what changed to Phantom “Cancer” Lancer since the last dota update where bug with him having 1 more illusion above maximum and another bug with diffusal blade damage were fixed? If you could do it, it would be much appreciated.

      I can’t do this very quickly since I need real life days to pass before I can make a sufficiently large sample of a new patch period. I will say though that dotabuff does provide daily win rates and Phanton Lancer’s past week doesn’t look especially noteworthy.

      Also, depending on how braindead they are, global abilities can be much more effective in lower brackets than they are in high. I tend to believe Zeus’ ult is more on the braindead side of things, but perhaps I am wrong on this.

      jimmy: keep in mind that Meepo’s Normal win rate isn’t especially good (42ish); his High and Very High Win Rates are just more terrible. I believe this to be driven by Normal players being so passive that Meepo manages to accidentally get out of control more often in that bracket.

      • jimmydorry says:

        The stats speak for themselves, but it is still hard to swallow.

        I’ve always steered away from gimmicky heroes like Zeus, but maybe I should look to see how I can build semi-stacks around him, and how to possibly convince pubs not to rage when a Zeus either initiates (to weaken the team) sacrificing mop-up ability… or finishes heroes which can be construed as kill-stealing (not the obvious kills where team is there hitting the hero).

        Really the whole lack of co-ordination works against some heroes more than others.

        As for Meepo, maybe I will try and master him a bit more on an alt account, and try pub stomping with him. Then again, he is not particularly “fun” to play. Not the same as an invoker or chen.

        The meta has to eventually shift when more people can manage to do something reliably with Meepo. It is still a very high risk, low reward proposition though.

      • Zeus is a very positional sensitive hero. If he gets caught out of position he casts one, maybe two, spells and dies. If he is in a good position he deals a truckload of damage. Learning where to be and not to be is one of the key things that you learn as you become a better player.

        You may also want to look at possible sampling bias.If any given hero has a much higher rate of participation in various brackets there may be externalities that are affecting win rates that aren’t immediately obvious.

  2. jimmydorry says:

    Interesting stuff, but I think there are two clear anomalies.

    I am not sure about Meepo in the bottom of the rankings though (you have mentioned sample size being an issue). It is weird to see his win rate DECREASE in the higher tiers. He is not exactly pub-stomping material, and he is very easy to shut down… but I still wouldn’t expect pubs to play or win with him in normal tier.

    And Zeus… Almost 60% win rate in higher tiers? *_*

  3. s_1 says:

    Meepo has a higher win rate in lower brackets because people will often make smurfs to practice him. This is relevant just because the amount of players that play meepo iin lower brackets is small so even a couple players making a new account and playing meepo would affect this sort of stat.

  4. xdv says:

    I always thought Zeus was meant to be played a teamfight initiator / counter initiator – your ult disables blink daggers which counters a lot of initiators, as well as giving your team the confidence to dive the enemy since it reveals the enemy’s positioning. And it can be done across the map as well. It sounds like a very teamwork dependant hero whose value increases with the effectiveness of the enemy’s teamwork, so it sounds like it should shine at higher brackets.

  5. Some Guy says:

    Broodmother no increase :(

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

%d bloggers like this: