Breaking Down the Alliance Draft Strategy (G1 League spoilers)

May 29, 2013

The 2013 G1 Champions League is over and the victors were Swedish team Alliance (formerly No Tidehunter) in a rather dominating 7-0 performance.  Much of the talk of the tournament has been centered around this being some kind of symbolic victory of Western teams over the formerly dominant Eastern scene.  I’m going to ignore all that and take this time to look at Alliance’s drafting strategy to see what makes them tick.

First we have this somewhat sloppy spreadsheet detailing the Alliance’s drafts in both the G1 league and many of their games in the WePlay league from earlier this month.  In the interest of focus, I’ve only included the first round bans for both teams and Alliance’s draft order, though I do occasionally make notes on the lineup of the opposing team.  If you want to look at the games in more detail check out datDota’s Alliance match log.

So let’s first describe Alliance’s overall draft strategy.

I’m not sure any currently active team makes as much of an effort to stick to signature hero picks as Alliance does.  First and foremost in this is their off-laner Admiral Bulldog who is somewhat known for only regularly playing a handful of heroes in matches (and a potentially unhealthy fixation on Lone Druid in matchmaking!), but the team has really seemed to embrace specializing in a handful of heroes and being able to mix-and-match multiple strategies from the same core.  In particular, Alliance highly values grabbing Magnus for S4 in mid and Lone Druid for the aforementioned Bulldog.  They also like to grab Chen for their support player Akke, but this pickup is less highly contested and might be considered more situational.  Their other support player EGM and their carry player Loda tend to draw from a somewhat wider selection of heroes, likely in response to draft conditions or intended strategies.

In some cases having a signature hero can be a bit of a liability; after all, your enemies know you want that hero and will go out of their way to ban it.  The problem is that Alliance has a lot of signature heroes and you only have two bans.  If you take out two of their heroes in the first phase of bans, they’ll respond by grabbing Batrider, Lifestealer, Wisp, KotL, basically many of the heroes you would otherwise be spending your early bans on.  All of those heroes can easily fit into their lineup strategies, and they’ll be happy to take any of them if you’re willing to leave them in.  Wisp in particular is a revealing case.  Alliance will only ban the hero if a) they actually fear your potential to Wisp and b) they are Dire.  In every other case they force you to make the decision to either ban out Wisp or face him, and from what I’ve seen they run Wisp quite well.

And now onto the hero specifics

I think the correct place to start is with the offlane.  In almost every case Alliance wants Lone Druid if they can get him.  There are only 3 games in the entire set where Lone Druid did not get either picked by Alliance or banned, and in all three it is simply because their opponents took Lone Druid with their first selection in the draft.  When Alliance can’t get Lone Druid, they will often grab Nature’s Prophet, usually in the first round of picks.  When this is not possible, the only other selections they have used are either a late Clockwerk pickup or an offlane Magnus.

Mid is somewhat more flexible, but their highest priority selection here appears to be Magnus.  In at least 2 of the 3 games that their opponents got Lone Druid, Alliance were Radiant and used their initial pick on Magnus.  In5 of the 15 games they ran Magnus as mid, and in 3 of 15 they ran him as offlaner.  It’s possible that they like him in part because they can always pass him over to Bulldog if they can’t get Lone Druid/Prophet or want to do something different with their midlane.

Their next priority in mid is to grab Batrider.  Actually, it might be their top priority, but the hero is so often banned it’s rather moot.  In the 15 game sample, Batrider was banned 11 times in the first round, and Alliance grabbed him 3 of the remaining 4 times.  Aside from Magnus and Batrider they can run a variety of other mids.  Templar Assassin was the most prominent at 3 appearances, and Qop, Puck, and Storm Spirit also had games.  There was also a Beastmaster in the gimmick strat they ran against DK to kick off G1 that you should probably see for yourself if you haven’t.

Supports are a bit tricky.  As I’ve mentioned, they will grab Wisp whenever they can, but don’t typically get the opportunity to.  Wisp was banned by Alliance’s opponents 9 out of 15 times, banned by Alliance 3 out of 15 times, and subsequently only played by Alliance twice.

The most noteworthy support trend is that Alliance loves to run either Chen or Enchantress.  Akke played those heroes 9 times in the 15 games.  This is particularly interesting as most jungle supports have fallen out of favor with the shift to a trilane vs trilane meta.  But we’ll get back to that in a moment.  Also of note, if they grab Chen it will be in the first round.  If they grab Enchantress it will always be in the second.

Aside from that, Alliance will mix and match.  They’ll make grabs for high value supports like Keeper of the Light, Shadow Demon, and Rubick, but it’s relatively rare for them to grab both in the first round.  They also have some pocket Ogre Magi strats, including what they ran in the DK game I linked earlier.

As for Carry, they’ll make a play for Lifestealer if he’s available, but the most noteworthy trend is Phantom Assassin with 4 picks out of 15.  That might not seem especially high, but their carry selection is much more flexible than Mid/Offlane, and carries also receive a large portion of the 2nd round bans.  They’re very much willing to run a semi-4 protect 1 strategy with either Phantom Lancer or Anti-Mage, and they’re also relatively fond of Alchemist.

Alliance almost exclusively picks up their carry in the 2nd round of picks.  This was true in 12 out of the 15 games with 2 of the 3 exceptions being the overall highly valued Lifestealer.

Moving to their laning strategy

Alliance drafts with the intention of having either a strong early game with as much carry potential as you, or stronger carry potential with a roughly equivalent early game.  They do this primarily by having either their off-laner or mid be a hero with a relatively strong carry potential: Lone Druid being the strongest, Templar Assassin being somewhat in the middle, and Nature’s Prophet being the weakest.  Their goal is to try to make you choose between containing this hero or being able to put pressure on their primary carry, and they react by playing extremely defensively on whichever target you choose to pressure.  Then in the midgame they’ll start roaming their supports and mid to buy farm room for the hero you had been up to this point containing.

The most striking example of this is when Alliance ran Chen or Enchantress in the G1 league against an offensive trilane.  Instead of making futile efforts to protect Loda, they just paired him with EGM on support and played extremely defensive in lane while trying to get as much experience as possible.  In the meanwhile, Chen or Enchantress would go to an offensive jungle and put pressure on what the opposing team was expecting to be a relatively 1v1 matchup.  Through this they would get good farm on their 4 support and offlaner, along with tower pressure and potential kills, and then use this advantage to free space for their carry to catch up in farm.  I can’t recall another team doing something like this as an answer to trilane pressure, and it really gives Chen and Enchantress a stronger position in the meta than they would have had otherwise.

(If you want to see some examples of this, check out the final game of Alliance vs LGD, as well as [I think] Alliance vs Orange)

Going back to the basic strategy of wanting to outcarry you given equal farm, it should be noted that Alliance’s one loss in this entire set was the Dignitas match where they ran Gyrocopter+Nature’s Prophet as their carries against a Lune+Templar Assassin team.  Based off my hero farm dependency test, it’s my opinion that Gyrocopter is not an especially strong hard carry.  He’s much more akin to a Juggernaut in that he has some carry viability, but his real specialty is the early game presence he has from his abilities.  Unlike Juggernaut he does some shocking late game potential in Flak Cannon.  Unfortunately, this is often dependent on having nearly a full set of items and possibly a Rapier on top of that.  Luna can arguably accomplish the same thing earlier thanks in part to her higher movespeed and +Damage passive, and earlier is a big deal–even for hard carries.

Also of note, Gyrocopter+Nature’s Prophet does work for Alliance in the final game of their set vs EG, but it works largely because the opposing team only has Weaver as a significant carry threat.

So what can we take from all this?

Let’s start with the problem of who to ban.  In my opinion it is absolutely Lone Druid+Wisp (assuming you aren’t Radiant and want to select either of those with your first picks).

Wisp is obvious.  If you let him go they will pick him up.  If you don’t have a strat ready to counter Wisp (and no team really seems to), you don’t want to face that.

As for Lone Druid, part of it is that I feel it is Admiral Bulldog’s strongest hero.  Part of it is that I feel Lone Druid is much more flexible than Nature’s Prophet, Clockwerk, or off-lane Magnus.

But a lot of this is that Lone Druid is really good right now, and I strongly suspect Armlet is the reason why.  Take a look at Lone Druid’s Win% and Relevance graphs on Dota-Academy.  In the past 3 months he’s seen a pretty noteworthy surge in their metagame relevance measurement while also having the highest 3 month winrate since his inclusion in Dota2.  This coincides fairly well with the discovery of Armlet bear, and I suspect Armlet is driving it.  On a side note, I hope that if Lone Druid sees changes in 6.78 it targets the efficiency of the Armlet on bear.

Granted, you then have to be ready to deal with Batrider, Magnus, and Chen, but the good news is that it’s rare for them to pick up both Batrider and Magnus.  The other good news is that Chen, as threatening as he is, declares a lot more above what their lineup will look like than many of their other selections.  You can threaten an offensive trilane, ward for Chen’s location, and roam your supports to punish him.  NTH actually did this back in late 2012 when they picked support Gyro+Naga to hunt down an enemy Chen.  At least I think I remembered this happened with NTH but maybe I have the teams wrong.

Basically, I think instead of asking who you don’t want to let Alliance have, you have to ask yourself who you’re prepared to let them have.  Pick your bans in such a way as to give them what they want while being prepared with a lineup that can handle what you expect them to throw at you.

But the other thing to learn from this is that part of Alliance’s strong performance is that they know the heroes that they’re good at playing and know the lineups they need to run to make them work.  Too many people get caught up in counter-picking and trying to win every lane.  It’s more important to have a set of lineups that work and that you’re very comfortable with, and then gradually expanding from there to deal with metagame shifts or certain opponents.  Don’t get so clever in your drafting that you end up defeating yourself.


DD’s Treant Dominance in the Western Qualifiers

May 21, 2013

Yesterday we looked at how mouz nullified a pair of Treant/Anti-Mage comps.  Today we look at the other half of the Treant story: DD Goblak’s utter dominance with the hero.

First some basic stats.

  • Treant was picked 9 times in the Western Qualifiers and won 5 of those games.  All 5 of those wins were on DD.
  • DD played 14 games.  Treant was picked by DD or banned in 13 of them.
  • Treant was banned in 8 games overall.  7 of those were targeted at DD.

Those were a bit of a pain to collect, so I might be off a game or two there.  Regardless, DD’s play basically defined what the hero was capable of in that tournament, and it warrants looking at in more detail.  I should also mention that I’ll be exclusively looking at how DD used Treant in a strategic sense.  It could very well be the case that Goblak right now is a superior Treant player from a tactical and mechanical standpoint, but that’s an entirely different analysis.

Rule #1: Play Treant as a support and get greedy with your carry, mid, and solo picks

In every game DD played, they used Treant as part of a trilane.  In pubs or lower level games this isn’t such a big deal, but when the quality of the competition is higher you need to maximize the number of useful Living Armor targets.  Let’s look at the non-support heroes DD ran in their Treant wins:

Outworld Devourer/Enigma/Gyrocopter:: Log // VOD

Queen of Pain/Batrider/Phantom Lancer vs mouz:: Log // VOD

Lone Druid/Batrider/Luna vs RoX.KIS (feat. Luna vs Zeus mid):: Log // VOD

Lone Druid/Batrider/Phantom Lancer vs RoX.KIS:: Log // VOD

Queen of Pain/Razor/Weaver vs mouz in Grand Finals:: Log // VOD

The basic formula is that you have a primary carry, a damage dealer, and an initiator.  The 4th game listed breaks this a bit by just going 3 damage dealers, and the 5th is kinda weird in that the tri-lane carry wasn’t actually their primary carry, but they still follow the overall pattern.  You really want at least two sources of early damage output in a Treant lineup in addition to whatever initiator you’re using.

And DD’s single Treant loss in the final game of the tournament?

Magnus/Dark Seer/Gyrocopter:: Log // VOD

Here they violate their own rule.  Magnus and Dark Seer are initiator/utility characters.  Gyrocopter is a big early damage dealer, and a potentially huge late game carry threat, but he’s not a strong midgame carry.  They really needed to either pick Magnus or Dark Seer, and replace the other with a big damage threat, and preferably one with a stronger midgame autoattack than Gyro.

The other factor at play in this match is that DD chose to run a defensive trilane around the Gyro.  Maybe the Alchemist/Lina/Bane combo forced this, but Alchemist beats Gyro in a straight farm-fest and Treant is absolutely helpless to do anything about this.

I should mention that all of this is in no way proof that Treant will never work outside of a trilane support.  For instance, if we see a metagame shift away from Trilane vs Trilane back to 4 protect one defensive trilanes, Treant might function as a suicide solo pick.  Maybe Treant could even work as solo now.  But regardless of where you stick him, you can’t play for a passive early game and expect to win consistently.  Given current drafting trends, trilane support Treant is the easiest way to fit him into an early aggressive lineup.

Rule #2: Pick Damage Dealers with Strong Midgames

Treant falls off after 20 minutes right?  Then you want to have a team that’s going to be really strong in that 15-20 stretch.  OD, Razor, and Weaver are all carries with questionable late games, but if you’re running this type of strategy none of that matters.  QoP and Batrider shine in this stretch of the game.  Luna and Lone Druid have late game potential, but they also have midgame strength that rivals Lifestealer.

The one possible exception is the Phantom Lancer pick that DD went with twice.  Of note here is that they always paired the pick with a strong midgame damage/initiation duo, Batrider/QoP in the first game and Batrider/Lone Druid in the latter.  This allows Treant to function better in a 4 protect 1 strategy.

The other factor to consider is that Phantom Lancer isn’t precisely helpless outside of the lategame.  Like Anti-Mage from yesterday he’s definitely squishy pre-items, but Spirit Lance and illusions allow him to contribute to larger fights without putting himself in as much danger.  Phantom Lancer is also more resilient to disruption (the general kind, not Shadow Demon) than Anti-Mage.  His early game is much less item dependent, thanks in a large part to his high agility growth.  With just levels and Tranquils he can transition to the jungle once the lanes get too dangerous.

In short, DD’s Treant/Phantom Lancer lineups were much more resilient than the Treant/Anti-Mage lineups yesterday.  One thing I should mention here is that perhaps part of this is the Anti-Mage players making the mistake to go for a Battlefury build in a Treant comp.  Building Anti-Mage for early aggression while treating Living Armor as a free Vanguard might be a better strategy for making sure your team doesn’t get bullied around before “I have finished farming” can kick in.

Rule #3: Living Armor allows you to violate lane and metagame conventions — take advantage of it.

The first game of the tournament where DD could actually get Treant before he was banned, DD ran an Outworld Destroyer mid and Enigma suicide lane.  Both of these are risky options (the latter a bit moreso) but Living Armor allowed them to get away with it.  I suspect a lot of teams don’t even consider mid OD a serious option because of the risk involved, but Living Armor mitigates that risk in a big way, which in turn opens up your drafting options.

Take Weaver.  I generally don’t like Weaver as a pick.  He’s counterable to burst, and preventing this often encourages an early Linken’s purchase which just accentuates Weaver’s tendency to fall off in the late game.  But he does have a ton of early game potential, especially when it comes to frustrating Lifestealer trilanes.  Living Armor amplifies this potential for early aggression and turns Weaver into a completely different pick than he would be without it.

But really the crown jewel of this strategy was when RoX.KIS picked a Zeus mid and DD countered by sending Luna mid.  Who the hell studies the Luna vs Zeus 1v1 matchup, let alone the Luna + Living Armor vs Zeus matchup?  It was a complete trouncing, and I can’t really blame the Zeus player in the least.  Living Armor allows a completely unparalleled element of unpredictability when it comes to laning, and good Treant comps will take advantage of this.

I think that does it for Treant, but while we’re here, let’s talk about a few of the other interesting hero trends in the Western Qualifiers.

  • Slark doesn’t really need much of a mention given the hype he’s been getting.  His picks were restricted to a couple teams, but he’s definitely a developing threat.
  • People have been talking about Magnus’ 5-10 record and how teams have finally figured him out.  Teams are definitely adapting, but one tournament result isn’t enough to rule him out as a threat.  Afterall, no one in their right mind would say that Nature’s Prophet has been “solved” just because he went 5-8.   I will say that Magnus may have played a role in Puck’s 10-6 performance, but I’d have to go back and look at the match-ups more in detail to check that they ran into each other as often as my memory tells me they did.
  • What does stand out to me as a potential trend is Lone Druid’s 13-4 performance.  I have to wonder whether Armlet bear has significantly boosted his effectiveness.  It’s definitely taken over as the de facto build for the hero.
  • And while we’re talking about Lone Druid, DD as a team seems to prefer to try to jam popular hero picks rather than fight for them in the draft.  For example, if I counted correctly they only used Lifestealer twice in their fourteen games.  Nonetheless, they definitely compete for Lone Druid, but when they couldn’t get him in the finals they brought out an interesting counter in Razor.  He actually was drafted twice by DD and I only saw the first game, but Razor definitely frustrated Lone Druid with very liberal use of Static Link.  I don’t know if it will have legs, but it’s something to keep an eye on.

Anti-Mage and Treant Protector in the Western International Qualifiers

May 20, 2013

(Part 2 on DD’s stunning performance with Treant in the Qualifiers can be found here)

The Western side of the International Qualifiers has finished, so I thought I’d take some time to talk about the appearance of Treant Protector (a hero whose latest incarnation I wrote about recently) and the way he interacts with the overall strategy of both teams.  As of this writing the event concluded quite recently so what follows will likely be considered spoilers.  Consider yourself warned.

Spectators have taken to jokingly referring to Treant Protector as Living Armor, but the joke isn’t actually too far from the truth.  The majority of Treant’s contribution comes from that single spell.  It can allow you to escape death from the complete opposite side of the map, and it more importantly allows the rest of your team to attempt aggression that would otherwise be suicidal.  He is, effectively, an amplifier for the rest of your picks, so he needs to be drafted with a very specific set of strategies in mind.

Another common theme in the Treant discussion is how he falls off hard after twenty minutes.  I feel that this weakness is often exaggerated, but it is still a legitimate weakness of the hero.  You might liken him to Undying in that respect, but I believe Treant’s early game is much more versatile than Undying who can only effect a specific portion of the map at a time.  Undying is effectively shoehorned into strategies that seek to force early teamfights, whereas Treant can contribute equally well on teams that wish to spread the map and force ganks.

One of the interesting outcomes of the Qualifier is that the top two finishers included the test who best used Treant and the team who best countered Treant.  It’s already well known that Goblak of DD was the only player in the entire tournament to actually win a game with Treant, going 5-1 with the hero.  What might have been less noticed is that mouz went 3-2 against Treant, finishing it off by going 1-1 against Goblak’s Treant in the finals against DD (with their previous loss to Treant also against DD earlier in the brackets).  So let’s begin by looking at mouz’s first two games against the hero, both of which happen to feature an Anti-Mage as Treant’s principle carry.

QPAD vs mouzVOD link

Mid:

QPAD Nightstalker vs mouz Beastmaster

Bottom:

QPAD Anti-Mage, Shadow Demon, Nyx Assassin

vs mouz Lifestealer, Keeper of the Light, Rubick

QPAD in the Radiant safe lane

Top:

QPAD Treant Protector vs mouz Nature’s Prophet

Treant’s impact is felt early when the QPAD trilane gets first blood and a double kill less than 3 minutes in.  Anti-Mage is a weak carry in the early laning phase, especially when compared to a Lifestealer, but Living Armor greatly mitigated the harass that he took during early laning and kept him alive when he blinked into the disrupted Rubick for first blood.  Shortly after that the QPAD supports rotate mid to get a kill on Beastmaster.  Things so far are looking pretty good for QPAD.

There are some other relatively minor events after that, but the real changing point of the game is when mouz decides to commit to a full 5-man push (with occasional split pushing by Beastmaster and Prophet) about 7 or 8 minutes into the game, which culminates in mouz going up 3 towers to 1 before 10 minutes.  This includes taking QPAD’s mid tower and bottom tier 2, and only giving up the relatively irrelevant Dire top tower.

But think of the entirety of what this strategy accomplishes.  Obviously there is the advantage mouz gets from tower gold and map control.  Thanks largely to this extra gold, Beastmaster finishes a Mekansm in under 14 minutes.  By 10 minutes Lifestealer has a Drum.  He adds an Armlet around 17:30 and an Aegis at 14:30.  Even Rubick gets a 10 minute Urn.

QPAD cannot outrace the tower push without giving up Anti-Mage’s freefarm.  QPAD likely can’t even outrace them with Anti-Mage.  But QPAD also can’t feel confident about taking a teamfight.  Living Armor is great when there’s a single, obvious target, but an early 5v5 means that mouz has plenty of non-Living Armor options.  mouz knows that Lifestealer trumps Anti-Mage in an early teamfight, and they also know that Lifestealer is giving up a lot less relevance by showing up to the teamfight instead of farming.  Nightstalker also makes his impact by finding ganks during the first few nights, and by robbing QPAD of map vision and only moving forward as a group mouz forces the Nightstalker to try to make an impact on their terms.  Granted, there are legitimate criticisms you could make of the Nightstalker play, but mouz knows what the hero wants to accomplish and makes the concerted decision to not give him what he wants.

So while none of this is actually about Treant Protector, it still is about Treant because of what I said earlier.  He exists to amplify the other heroes on the team.  If you can nullify those heroes, then you nullify Treant.

  • Living Armor certainly did help Anti-Mage with his weakness of weak laning, but there’s nothing Treant can do about his weakness of needing at least 20 minutes of uninterrupted farm.
  • Treant can help sustain team pushes through Leech Seed, but QPAD didn’t have any significant pushes to speak of.
  • Treant can mitigate the damage towers take, but only if you can force the enemy team to back off.  Good luck doing that when your best tool for anti-push is Shadow Poison.
  • Can Treant contribute to early teamfights?  Sure, but who on QPAD is going to teamfight?  Anti-mage wants 20 to 25 minutes to gear up and not be squishy.  Nightstalker has to deal with Telekinesis, Sprout, Open Wounds, potentially a Roar, and kiting in general to do his damage.  Nyx is a threat, but mouz is staying as 5 and has sentries.  That leaves Shadow Demon and Treant who both can contribute to teamfights, but neither of which is going to carry on their own.

QPAD picked a lineup that wanted to aggressively gank until Anti-Mage became relevant.  mouz recognized that if they shut down Nightstalker and pushed early as 5 they could deny QPAD their only avenue of victory.  Treant was a total non-factor because he had no one to amplify.

mouz vs RoX.KISVOD link

Mid:

mouz Puck vs RoX.KIS Magnus

Top:

mouz Lifestealer, (support) Gyrocopter, Lina

vs RoX.KIS Anti-Mage, Shadow Demon, Treant Protector

RoX.KIS in the Dire safe lane

Bottom:

mouz Lone Druid vs RoX.KIS Batrider

mouz once again goes for early aggression and pushing tower.  It’s interesting to note that in the first game mouz was not drafting with Treant Protector in mind.  QPAD took Treant Protector with the final pick of the draft, so it’s more accurate to say that mouz was drafting against Anti-Mage.  QPAD selected Anti-Mage with the final selection of the first round, which likely prompted the Beastmaster mid.  RoX.KIS is a slightly different story.  RoX.KIS also selected Treant with their 5th pick, but in this case, mouz had one pick remaining.  Their decision to take Lone Druid may have been prompted by the Treant selection.  Prior to that point they could have been planning to take Gyrocopter solo bot, but then recognized that support Gyrocopter would be quite effective at quickly chewing through Living Armor charges.  Having a Lone Druid bottom would give them even greater mid-game carry potential and pushing power, both of which we’ve seen to be quite effective against a team running both Anti-Mage and Treant Protector.

RoX.KIS’s draft has a similar problem to the one faced by QPAD.  They have to stall for 25 minutes, but they don’t have the tools to do it and mouz does not intend on giving them and breathing room.  Magnus and Batrider are both great initiators, but with Anti-Mage needing to be in full farm mode no one is there to capitalize on the initiation.  Treant is forced to use Living Armor entirely for defensive purposes, which isn’t a great place to be when you need to stall against a lineup with the kind of midgame force that mouz is packing.

But where this game is even worse for the Treant/Anti-Mage combo is that at least QPAD won the laning phase.  Kill-wise the game is a slow start.  No first blood until 7:30 when a good Puck initiation in the bot lane catches out Batrider, which then gets turned into a double kill (and wasted RP) on the Magnus who teleports in.  Meanwhile, and perhaps more importantly, the top lane is an even bigger disaster for RoX.KIS.  Why?  Not because of kills.  Neither trilane manages a kill before the laning phase ends.  No, the entire disaster for RoX.KIS is that Anti-Mage only has 4.5 creeps per minute by 10:00.

From a slightly above average pub perspective, 4.5 cs/min on an Anti-Mage is a godsend, but here it is not going to cut it.  Naix can do just fine on 4.5 (and for the record was leading the game at 10 minutes with 4.8),  Lone Druid and Gyrocopter can impact a game on much less, but Anti-Mage desperately needs gold to get the items he needs to survive teamfights, and he simply cannot find enough of it with Lifestealer, Gyro, and Lina constantly pressuring him.  Every CS they deny him just extends the amount of time RoX.KIS needs to stall further, and they simply do not have the tools for it.  Treant, Shadow Demon, and to a lesser extent Batrider make their impacts on fights by amplifying other heroes, but with Anti-Mage in full farm mode, the closest thing to a carry they have is an early game Magnus who absolutely is not going to cut it.

Unlike the QPAD game this one never even looks close.  By 10 minutes mouz is up 5 kills to 1.  The CS might look relatively even, but RoX.KIS has two heroes who aren’t in position to take much of an advantage with any amount of CS (Magnus and Batrider), and one hero who will need at least another 15 minutes before they can do anything with their CS (Anti-Mage).  mouz is in the dominating position and by the final GGs extend their kill lead to 24 to 2.  It gets so bad that Anti-Mage gives up on a half completed Battlefury.

So what do we take from all this?

I can’t say that Treant + Anti-Mage is a strategy that will never work, but I’m certainly skeptical.  It’s a question of power curves.  Treant can be extremely relevant for the first 20 minutes.  Anti-Mage is potentially at his most potent by around 30 minutes depending on his farm rate.  I feel that Living Armor, as amazing as it is, isn’t good enough to bridge that gap.  What you ideally want from a Treant strategy is a team that peaks around 15-25 minutes.  This is the point where you use your early carry potential and Treant’s Living Armor to completely control the game and choke out your opponent.

Now, it’s definitely true that Treant allows you to run lanes you might not normally get away with, but that only solves half of Anti-Mage’s problems; the other half (Anti-Mage’s extremely late power spike) Treant only exacerbates.  In my next post we’ll look at what DD did differently when crafting their lineups around early Treant Protector picks, why it worked for them, and perhaps why it stopped working for them when mouz took the finals 3 games to 1.


The Shadow Blade Surge

May 15, 2013

A couple of you noticed a conspicuous absence in Dota 2 Item Trends in 6.77.  In truth it was intentional, because the shift in Shadow Blade usage since the large changes the item received in 6.75 deserved a post all to itself.  Let’s go straight to the top 6.74->6.77 item usage shifts in Very High.

[Shadowblade]Intro

We already covered how Dagon and Diffusal were covered by the rise of Nyx and Phantom Lancer respectively, and Refresher is likely attributable to a combination of Magnus and Warlock.  Tranquil Boots and Rod of Atos were recently introduced items that have seen increases acceptance as people have become more accustomed to them.  But at the top of the list is the Shadow Blade everyone was expecting to see.  Truth is, the only reason Shadow Blade didn’t show up in my previous post was that I chose to look exclusively at the shift between Normal->Very High.  Shadow Blade didn’t show up there because it’s increased usage was relatively uniform across all three skill brackets.

I’ve mentioned how often a shift in item usage can be attributable to underlying hero shifts, but with Shadow Blade this is not the case.  I isolated 32 heroes with what I consider a significant amount of Shadow Blade usage (a >3.8% usage rate in any bracket in either patch).  Three of these heroes were not in the game in 6.74.  All of the remaining 29 saw their Shadow Blade usage increase between 6.74 and 6.77 in both the Normal and the Very High brackets.

(Click for Usage by Hero in Very High Bracket)

[Shadowblade]byHeroVH

(Click for Usage by Hero in Normal Bracket)

[Shadowblade]byHeroN

Focusing exclusively on 6.77, let’s look at the trends in usage between Normal and Very High:

[Shadowblade]SkillShifts

The distinct trend here is that Very High is much more likely to build Shadow Blade on melee heroes for the speed boost and initiation, with Alchemist being the poster boy for this trend.  Normal is more likely to build Shadow Blade on ranged carries, but this gap has lessened significantly since 6.74.  Normal also sees a much higher presence of supports buying Shadow Blade.  Aside from both these trends we have Nature’s Prophet as a much more common carrier in Very High and Death Prophet as a more common carrier in Normal.

Of course the question on everyone’s mind is whether the post-6.75 Shadow Blade is better, and unfortunately I don’t have a good answer for you.  Win rate comparisons on items through the API is a messy and often misleading endeavor.  I have some ideas about new ways to approach it, but they’re not ready yet.

You might be tempted to just look at the base win rate of Shadow Blade alone, but this is a bad way to go about it because of compositional issues.  To illustrate this, let’s look at the 6.77 item data.  As you can see at the bottom of the chart, Shadow Blade has a 61.6% win rate in Normal and a 59.02% win rate in Very High.  Must mean something right?

Not so fast.  Remember that while we’ve already looked at the usage rates of Shadow Blade on the most popular heroes, the usage rates of the heroes themselves vary significantly from bracket-to-bracket.  To show what’s going on, I counted up the total observed carriers of Shadow Blades in the two brackets (i.e. of all the Shadow Blades built in Normal, Drow Ranger is the carrier for slightly over 1 out of 5).

[Shadowblade]PieN

[Shadowblade]PieVH

The composition between the two brackets looks completely different, largely due to how less often ranged carries like Drow, Sniper, and Viper get picked in Very High.  Drow in particular is a big deal because her win rate in 6.77 was quite high and her increased influence on the Normal win rate of Shadow Blade could easily account for its better performance in the bracket.

To really look at item performance, we need to look at each hero in isolation while controlling for factors like GPM and (in an ideal world) purchase time.  In the meanwhile, let me say that while I believe Shadow Blade is definitely a stronger item now, I suspect it’s not strong enough to justify all of the extra purchases it’s receiving in 6.77.  Just because it’s a decent purchase now doesn’t mean that there aren’t times and situations where you have more pressing needs, and it’s probably still a bit dumb to build 2 or 3 on a single team.


6.77c’s Treant Protector is Secretly Really Good

May 4, 2013

Technically this is another Skill Build Analysis, but on one hand I wanted a title that was a bit more inflammatory, and on the other, there’s not a lot to analyze.

Ever since the 6.75 nerf to Treant, people have been really down on him.  In 6.74, he was quietly a pubstomper extraordinaire at the lower levels of the skill bracket, thanks in a large part to his completely passive global aura.  6.75 revamped Living Armor into an active global ability with the potential to make him more viable at higher skill levels, but for various reasons it never really worked out, and couldn’t offset the removal of all ultimate damage that he received in the same patch.  Every patch since has come with a litany of minor buffs for Treant to the point where anticipating Treant buffs in every new patch is a running gag.  But 6.77c’s buff was different in that it finally plugged the final weakness of Living Armor.  Relatively speaking, Living Armor is a silent contributor, but it’s also one of the most potent abilities in the game right now and I have the stats to prove it.

Let’s start by looking at how players adjusted their skill builds in light of the new patch.  If you need a refresher on the methodology I’m using to generate these charts, you can find it at the top of my Silencer Skill Build Analysis.  And for a quick reminder:

Q: Nature’s Guise — single target tree stealth.

W: Leech Seed — single target DoT and AoE lifesteal

E: Living Armor — global single target HoT and damage block

[Treant]Usage

Pre-patch Leech Seed and Living Armor builds were neck and neck at all skill levels.  Post-patch, we see an unsurprising surge in Living Armor builds that gets stronger the further up in the brackets we go, likely reflecting that a higher percentage of Very High players are aware of patch notes.  You’ll also note that the vast majority builds are either W->E or E->W.  Having looked through the stats, I don’t feel any other build has a statistically significant representation, so we’re basically going to be doing a head-to-head comparison.

And for those of you who like this kind of thing, that .08% of Q builds in Very High 6.77c represents 4 games.  They were not very successful.

Moving on to win rates, I mentioned a few weeks ago that Dotabuff was showing a 6.5% surge in his win rates between patches, but if we wait a bit and break things down it turns out that 6.5% was actually quite an understatement.

[Treant]winRates

That’s an 8.4% bump in Normal and a rather amazing 12.3% bump in Very High.  Living Armor builds outperform its nearest competitor by a minimum of 4.5%, a gap not seen in any SBA I’ve done before.  Treant’s aggregate performance now scales greatly with skill level, capping out at a startling 62% win rate for Living Armor builds in Very High (out of a sub-sample of ~2700 games).  For reference, if we were including these stats in The Heroes With the Greatest Improvement in High Skill Games, Treant would now come in very close to Nyx Assassin who placed 4th.

To be fair, we only get 62% after we filter out his less performing builds, which could also have the side effect of filtering out less informed players.  For reference, Crystal Maiden’s top performing builds in her second SBA came in at around 63%, but they were minority builds and all of them combined didn’t have close to the 50% representation that the Living Armor -> Leech Seed build has.

It’s also the case that there’s no guarantee that any of this means Treant is competitively viable.  There have been plenty of heroes with high public win rates that don’t translate to competitive success for whatever reason.  That being said, he’s being tried more often, and I’m not convinced that the teams trying him out have him completely figured out.  There’s still a lot of room for team comp improvements, and it could be that the first team to put him in the right place and surround him with the right heroes will surprise a lot of people.

But for the average pub-player the message here is simple.  4 ranks of Living Armor by 7.  No exceptions.  0-2-4-1 appears to be the strongest performing level 7 build, but maybe a relatively early point in Nature’s Guise is ok situationally.  You exist to cast Living Armor on things.  Is someone possibly in a fight?  Cast Living Armor on them.  Is your carry missing HP in lane?  Cast Living Armor on him.  Are you going in for CS?  Cast Living Armor on yourself.  Has a tower taken damage?  Cast Living Armor on it over and over.  Your initial item goal is to find enough mana that you can cast Living Armor essentially on cooldown while also being able to Leech Seed and Ult.

And maybe you hate this playstyle; that’s fine and totally understandable.  But if you do, I advise avoiding the character entirely.  His kit is balanced around the idea that Living Armor is insanely good for the first 20 to 30 minutes, and you absolutely have to take advantage of that if you want consistent success out of him.