[SBA] Huskar and Ancient Apparition -or- Maybe Burning Spear Isn’t So Terrible After All

When I changed methodologies on my skill build analysis a week or two ago, one of the driving reasons was that the old version didn’t work so well when it came to heroes with very similar early game builds but divergent late game builds.  Huskar and Ancient Apparition aren’t particularly hot in either pubs or tournament play at the moment, but they’re useful to me as examples of how this new analysis method is way more flexible.  In these cases, the big question for both of these heroes is whether their often neglected skills Burning Spears and Chilling Touch are even worth taking at all.  The answer, particularly in Huskar’s case, is a bit less than straightforward, but I’m getting ahead of myself.

Let’s start with Huskar(dota2wiki page):

Q – Inner Vitality: Scaling Single Target Heal

Scaling — Rate of Scaling

W – Burning Spear: Magical DoT Orb Effect

Scaling — Damage per Second

E – Berserker’s Blood: Stacking Damage and Attack Speed Passive per HP Missing

Scaling — Attack Speed and Damage per 7% HP

Now for the table:


There are two big stats I need to draw your attention to:

The first is that Huskar’s win rate drops pretty significantly the higher you go in the bracket (48.23% -> 45.78%).  This is hardly surprising, but it becomes important later on.

The second is that maxing Berzerker’s Blood(E) first definitely appears to be the way to go.  This should be even less surprising, but just to really hammer home the point, I made a quick little chart comparing the win rate of Huskar in games where he maxes E by level 8 to games where he doesn’t max E .


So Huskar’s in the Normal bracket who do not max E see their winning percentage drop by 8.5 percentage points, which is a pretty huge dip.  It also points out that Huskar’s performance gap between Normal and Very High is ~5% when you look exclusively at builds that max Blood first.

With Huskar’s ideal primary easily determined, the real question about Huskar builds is how he should spend the rest of his points.  To examine this more closely, I created a modified version of the Win% by Skill Point table from the previous SBAs.  But instead of looking at Q/W/E by level 8, I restricted the search to builds that max Blood first and looked at Q/W/Stats by level 10.  Since these builds have all spent 4 points in E and the vast majority one point in Huskar’s Ult, this leaves 5 points available spread between our three options.  So how did things shake out?


At all levels, Blood(E) into Inner Vitality(Q) is the most popular.   Blood into Burning Spear(W) is the next most popular, but loses ground as you move up through the brackets.  Blood + Stats is always fairly rare, but still reasonably popular.  The win rates paint a rather curious picture.  At Normal and High, E+Stats is the strongest followed by E->Q, but in Very High E->W looks like the clear winner.  In light of this, I did another focused test comparing E Prime builds with 4 points in Burning Spear by 10 to E Prime builds without 4 points in Spears.


It doesn’t occupy a huge percentage of his Very High usage, but E->W definitely experiences the least Very High Win Percentage Decay of any of Huskar’s major builds.  It’s not definitive since these builds make up a relatively small portion of the entire sample (~487 games total in Very High), and it could be a trick related to E->Q just being more popular.

On the other hand, it could be that E->Q and E+Stats do well in the Normal bracket because the HP edge helps Huskar win 1v1 fights reliably and snowball from there.  In Very High the importance of 1v1 fights is diminished and builds that max Spear second have better overall damage output for skirmishes, early teamfights, and non-Ancient Neutral farming.  Maybe.  Whatever the case, there appears to be nothing wrong with a 0/4/4/1 build.  Other builds appear to outperform it in the lower brackets, but it remains resilient in all 3 brackets.

Now Ancient Apparition on the other hand paints a much more straightforward picture.  For the sake of time I won’t describe his abilities, but you can check them out on his Dota 2 Wiki Page


Cold Feet(Q) -> Ice Vortex(W) is the unambiguous winner here.  So is Chilling Touch(E) still a skill worth skipping for stats?  To test this we just replicate the tests we did with Huskar, only this time we look at QW builds at level 10 and see how they fare when they’ve put points into Chilling Touch.


It’s not a huge effect, but it does look like Chilling Touch is might still be a bit lackluster.  It does do well enough in Normal though, and it’s at worst not an active hindrance.

What needs to be emphasized though is that this is not an argument that Chilling Touch should never be skilled; it’s at best an argument that it usually does not pay off to skill it early.  As always, there could be situations where early Chilling Touch is actually quite viable but that are statistically rare enough that they aren’t showing up in my analysis.  What this means is that if you have a good argument for trying to integrate Chilling Touch early, by all means test it out.  If you don’t any reason to believe that Chilling Touch is going to work particularly well in a given game/lineup/lane matchup, then you’re probably better off playing it safe with a standard Q->W->E build.

So to summarize:

  • Always max Berzerker’s Blood first on Huskar
  • Blood -> Stats appears to be the strongest Huskar build at low levels of play, followed by Blood->Inner Vitality
  • If you’re doing either of those builds, only grab 1 point of Burning Spear if you need it for specific orb-walking purposes (such as attacking through CM’s Frostbite)
  • On the other hand, 0/4/4/1 looks perfectly viable for Huskar, especially at higher levels of play.
  • Ancient Apparition appears to generally perform the best with a straightforward Q->W->E build order.  The E vs Stats comparison is inconclusive, though Chilling Touch does appear to be the stronger option in lower bracket play.

6 Responses to [SBA] Huskar and Ancient Apparition -or- Maybe Burning Spear Isn’t So Terrible After All

  1. Ewig Custos says:

    Awesome as always.
    But one question: just how large are that “dangerously low sample size” on AA? Win rates on primary W in high and very high bracket seem a little bit off. W – E jump from 35.71% to 62.50% and then to 0% does not seem legit. High winrate with not-so-good synergy is almost twice bigger than normal and is bigger than Q – W, and 0%, like, seriously? I don’t think that there is any build of any hero that ALWAYS fails.
    Also, it’s quite interesting that there is consistency in split builds’ lows winrates. Looks like it’s rarely viable in comparison with focusing on 2 skills.
    By the way, it would be nice to look at synergy statistics for heroes. Like, winrates for AA with teammate(s) with stun(s), slow or without them, since it could make a HUGE impact on early game. Yeah, I know that this is enormous amount of work, but who knows, maybe someday that can be done automatically? More stats, more happiness. :D
    Anyway, I take off my hat to you. Keep up the good work.

    • phantasmal says:

      AA’s W->E in High is 6500 * .0429 * .0287 = 8 matches. In Very High it’s 6500 * .0242 * .0064 = 1 match. So yeah…that’s something.

      Hero synergy is a tricky thing to test for. For me right now it’s not really viable since I can only get around 20,000 matches a month if I want to do bracket divisions. And then you need at least hundreds of thousands of Ancient Apparition matches to have a statistically significant sample for every possible teammate.

      Dotabuff actually has that many matches and more and they tried something like it before, but then the problem you run into is say, suppose you think AA and Chaos Knight have good lane synergy and then you look for it statistically. You might still have trouble finding significant patterns because of all the AA + CK matches you have they might only be laning together less than a third of the time. Basically a synergy or counter might actually exist, but it can only exist if the heroes are actually running into each other during a point in the game where the outcome is still in question.

      I think it’ll be somewhat easier to do something like lane counters and synergy tests if you can use replay parsing to create a lane matchup system so you can create an exclusive list of games featuring the precise laning scenario you want to test. Until then you need a ton of data to get anywhere, and even then there’s going to be a lot of noise.

      • Ewig Custos says:

        Actually, I think that usage rates will be pretty much the same with dotabuff’s (or valve’s) database: 6500 is more than enough for that purpose. I believe only winrates will be affected.
        Well, these low sample sizes for those builds just mean that on high and very high skill brackets players don’t usually go for builds that are considered useless (or far less useful than others), and there is no reason to believe that situation will change if you’ll increase primary sample size.
        Regarding replay parsing: if it will be available, again you need a lot of resourses to go through them to get accurate results. Oh, and some old replays are not available, so that won’t make the task easier (this can be a temporary dota bug, though).
        And by the way, your work really should be more well-known, I found this site completely by chance.

      • phantasmal says:

        You’re right. I’m reasonably confident in usage rates. I just don’t want to give anyone the impression that a build is actually secret op because it won 5/8 games in the sample.

        It also plays into some hesitation in other parts of the analysis. E+W vs E+Stats is an interesting comparison, but neither selection has a huge number of games to draw from. With a larger sample I feel I could offer a more definitive answer on that front.

        Parsing at that level does take a lot of resources, and that’s part of why I’m sticking to the API for now. I don’t know if old replays being missing is too much of a hindrance though. I prefer to treat each path period as a separate entity, so as long as you’re getting a significant number of the current patch period you’ll have something interesting to say. There’s certainly things in older patches I would have liked to explore, but I’m not going to lose any sleep over it.

        And yeah, I’m a pretty terrible self-promoter. It’s probably something I should address, but given the choice between advertising and doing anything else I feel like I’ll always go for the latter.

  2. jimmydorry says:

    I read this yesterday, and could see how E + stats could theoretically help you snowball earlier in the game.

    So I put it to the test, and it seems quite effective. This is not a definitive test, as I will play around with it a bit more. I don’t think I would have even thought about mixing stats in earlier for Huskar without reading this.

    This is the game I tried it in. Typical scrub match-up, and I was definitely not playing safe (in fact I often suicided into the whole enemy team just to see how much damage I could inflict).


    • phantasmal says:

      Yeah, I think E+Stats is perfectly legit if the situation calls for it. My general feeling at the moment is that too much of Inner Vitality’s value is held back early on. You just don’t have the strength or survivability to get much bonus healing, and you also won’t have the mana available to use it very often. Stats and Spear are both flat boosts so they’re most effective if picked up early, but you’re usually going to be better off committing hard to one or the other and then picking up Inner Vitality later when you have the base survivability to make it effective.

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: