6.78 Win Rate Shifts in Charts

With DotaBuff appearing to be back in order, I can now offer you this chart of the biggest movers of the patch:


Overall my eyeball estimates held up pretty well.  The only big exception is Elder Titan who I estimated at 1% but shows up here at 3%.  My suspicion is that the 1% estimate is more accurate.

When I made my estimates, I was looking at a weeks worth of daily data.  This chart uses an entire months worth of the 6.77c patch period.  This is important because Elder Titan was released in the middle of 6.77c, so the chart includes the initial week of his release.  Some heroes experience a win rate bump after the first couple weeks of their release because the population of people playing the hero goes from “Whoever can lock them in first” to “Mostly players who actually know how to use the hero.”  Slark experienced the same kind of exaggerated boost going from 6.76->6.77.

Some other things worth mentioning.  It’s entirely possible that shifts of less than a percentage point are just noise.  Conversely, it’s extremely likely that shifts above two percentage points are significant, but just because a shift is significant doesn’t mean it’s uniformly significant.  Because this is a sample of all games, these results are mostly from Normal games.  It’s entirely possible that some shifts are understated, exaggerated, or even the reverse of what we might see in Very High games.  It’ll still take some time before looking into that is possible on a wide scale, but it will be coming in the future.

Also, before anyone sees that 7% Huskar shift and declares him overpowered, keep in mind that pre-6.78 Huskar was generally pretty bad.  That’s not to say that this particular iteration will survive the next patch intact.  All we can say is that so far the new Huskar appears to be positively average in the context of low level pub games.

Anyway, some of you might want a unabridged version of the chart, so here you go.



13 Responses to 6.78 Win Rate Shifts in Charts

  1. icarus says:

    When the new patch came out I felt that hero selection became more diversified and heroes overall more balanced. I can see several shifts towards the average (50%), which seems to be in line with my perception. In this context it would be interesting to have readily available: 1. the std dev from the mean, range and/or other deviation metric; 2. median value. These values may be used as proxies of balance in the pub scene.

    • icarus says:

      note: I mean the deviation and median value for the win rates of all heroes, not for the win rate of each hero (although that could also be interesting/informative but for other reasons).

  2. jimmydorry says:

    Nice to see this stuff compiled.

    On a side note, I am not sure if you are aware, but more than 75% of the user base does not share match history. :(

    I’ve been collecting all of the match data for the last 6days. I’m not collecting the match details at the moment though, as I am busy doing something else with the data.

    For all of these stats, I can’t help but wonder how they would look if you average the win-rates per person per hero and then as a whole per hero. It should stand to reason that the people that play a hero more, will probably have a higher win-rate, and as they play more… the rates are biased more heavily towards them. Just a wild thought though.

    • Sheep♥ says:

      On the other hand, if you become better with a hero, you get upper in brackets, so maybe your win rate remains unchanged.

    • phantasmal says:

      On a side note, I am not sure if you are aware, but more than 75% of the user base does not share match history. :(

      I did see the Dev forum thread. I also don’t find this terribly surprising.


      Opt-in vs Opt-out makes a huge difference when it comes to participation. I suspect the privacy settings being opt-out is very much an intentional choice on the part of Valve. This likely also plays into why AP is by far the most popular mode and is definitely part of the reason why Solo Matchmaking has lower usage rates.

  3. aaa says:

    Oh man troll -7%. Icefrog has a bit of a problem balancing how effective to make global spells it seems, between that, Living Armor, Drow’s Aura and Centaur’s Ult.

    • phantasmal says:

      Balancing global abilities can be problematic, but the Battle Trance nerf is only half of the story. Maybe even less than half, but it’s hard to say.

      For the sake of comparison, Ursa dropped at least 5% solely off of giving Overpower a cast time, and that was in conjunction with a major Earthshock buff. Adding (or in this case restoring) a cast time to a hero vulnerable to kiting is a pretty big deal.

  4. ALegg says:

    How do You differ between 1 anon user, from another?
    Or 1 anon user playing 100 matches/week, from 100 users playing 1 match/week?
    Without those information I’d not make any statement about “% of userbase”.

  5. Krans says:

    Cool stats. Why not consider using a Brewer perceptual palette for colour coding values!

    • phantasmal says:

      Because I had never heard of them!

      I actually don’t use these visualizations much myself, so I have trouble seeing the difference, but if it’s something other people would benefit from I can look into ways to switch up the palette coloring.

  6. Paradigm says:

    Wow, that Whirlwind Axe “nerf” & Shadowblade nerf really hurt Troll. I felt before the patch the the nerf to Ult didn’t matter much, and after playing in the new patch I’d say it’s true, the nerf to his Ult doesn’t matter that much. It was the combo of Whirlwind Axe nerf and Sblade nerf that really makes him a lackluster hero, can’t really carry as hard, still can a tad.

  7. trenchie says:

    I would really appreciate more analysis that excludes players seemingly pressing buttons at random and buying recommended Vanguards. Quite a few of the heroes are way beyond the grasp of so-called Normal players and it distorts actual balance picture.
    In other words, less of thousand monkeys with typewriters observations and more of dota scene analysis, por favor.

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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: