Hero Win Rates on Radiant and Dire

September 30, 2013

Received a question asking whether any heroes had a noticeable win rate advantage on either Radiant or Dire, so I threw this together.  There’s not a lot that’s terribly interesting here to me, so I’ll present it mostly without comment.  Keep in mind that the percentage differences we’re looking at here are fairly small, so most are likely well within the margin of error.


(Quick explanation on the expectation column: since Radiant has the higher win rate overall, it tends to have the higher win rate for most heroes.  I just did a quick regression to see which heroes grew the most/least.  There were a couple other ways I considered expressing this, and none of them had results that were much different than the others.)

Ursa is, unsurprisingly, the largest standout given his predilection towards Roshan.  Beyond that who knows.  I’m sure you’re all capable of coming up with your own theories, but I wouldn’t read too much into this and instantly demand Bloodseeker mid every time you’re on Radiant.

Also I looked at what heroes had the largest percentage of their games either on Radiant or Dire.


Medusa is the big mover here for Radiant ancient farming shenanigans.  On the opposite end but to a much less degree we have Beastmaster for his Dire ancient stacks.  Also of note, despite being the most successful hero on Dire, Ursa was one of the most popular Radiant picks.

But yeah, overall I’m not enthralled by any of this.  However, if you are interested and want to look through the numbers yourself you can find them here.


Dota 2 Hero Winrates by Game Duration

September 24, 2013

Today we’re going to look at the relationship between the success rates of different heroes based on the length of the game.  Some heroes perform distinctly better in games that end early, and others prefer games that drag on, and while most of the placements aren’t terribly surprising, it’s still something worth establishing.

My method here is pretty straightforward.  I took the +35k game Very High sample that I’ve been recently, and created a list of game durations.  As it turned out, 1/3 of games were shorter than 33:20, and 1/3 of games were longer than 41:46, so I used those as my break points to define short, medium, and long games (the overall average was 37:50).  I found each heroes win rate in all 3 types of games, and sorted them by how skewed their win rates were towards either end of the spectrum.  To keep things simple, I’ll just be expressing this skew in the tables by the difference in win rate between the short and long games or vice versa.  So let’s kick things off with the top 15 heroes that prefer short games and the top 15 that prefer long games.


The biggest surprises here for me were Nightstalker and Axe both doing significantly better in longer games.  You might also notice that the Long list is very carry oriented while the short list is more support oriented.  Part of this is likely that every support you have on a team is a position that isn’t filled by a carry and will automatically bias your team strategy towards quicker games.  What this means is that your core farmers likely exert the stronger pull on the idea team timing, so if you have both a Chen and Anti-Mage on a team, Anti-Mage’s duration preference probably trumps Chen’s.

So given this tendency, I thought it’d be helpful to point out the supports that skew the most towards late wins and the carries that skew the most towards early wins.


One word of warning I feel I should add here is that while it can make sense to pair a bunch of early oriented heroes together (say, Power Ranger’s lineup with Chaos Knight and Weaver as their carries with Treant as a support), the reverse doesn’t necessarily hold.  If you’re running a late-game oriented lineup you might need some strong early heroes just to make sure that you actually make it to the late game.  This is particularly true when pubbing, and to illustrate why that is, let’s look at the heroes with the highest observed win rate in Very High games.


You can see pretty clearly that this list pretty heavily biased towards early game heroes (though the big exception in Spectre is pretty noteworthy).  Being able to put the game away early is a huge advantage in a disorganized pub environment, so if you’re trying to bump up your win rate make it a point to pick heroes that you’re capable of making a sub-20m impact with.

For those of you who want to see the entire list here’s the table with all the data (Click to enlarge)


As well as the entire dataset for both Very High and Normal in Google Drive.

Finally, have the overall match duration for every hero in both Very High and Normal


Drafting an Upset: Power Rangers vs Alliance

September 21, 2013

[Obviously spoilers.  If you like, you can watch the VOD before reading further]

I’ve talked a lot about Alliance here, first about their G1 performance and then recently about their usage of Naga Siren at TI3, so when I found out about this game I knew I had to talk about it.  Power Rangers did their homework for the draft phase, and it earned them a victory over the current International champs.

Stats provided by datdota.com

Stats provided by datdota.com

Since TI3, there’s been a lot of talk about how to ban against Alliance, particularly the sentiment that Na`Vi had the right idea in ignoring AdmiralBulldog’s heroes to focus on the star supports in Naga Siren and Chen.  Of course the “right” bans depend a lot on the context of the game and what your team is best prepared to deal with, but I actually think Power Rangers have a far better plan here.  My top three draft threats on Alliance are Wisp/Io, Naga Siren, and Nature’s Prophet in that order.  Power Rangers opens with bans against Naga and Prophet and leaves Wisp in the pool as they have first pick.

The lesson of TI3 wasn’t that you don’t target AdmiralBulldog with first round bans; it was that you don’t target Lone Druid.  While Lone Druid is Bulldog’s signature hero, Prophet is a far stronger complement to Alliance’s typical gameplan.  I also feel that Alliance is far more willing to grab Prophet with a first round pick than they are Lone Druid.  Power Rangers capitalized on this tendency by using their second pick to take Bounty Hunter and then immediately followed that with a Lone Druid ban.  This allowed them to take out Bulldog’s trinity while only giving up a Chen in exchange.

As for Chen, yes, he’s a strong hero, and Alliance uses him well.  But he’s also predictable.  If you give up Naga Siren, you don’t know what to expect.  She can fit into a defensive trilane, offensive trilane, and 2v3 lane.  Chen on the other hand will be in the jungle, and likely his jungle (Alliance seems to prefer Enchantress for aggressive jungling).  Combined with the early CC-less Abaddon pick, Alliance’s selections revealed that their laning phase would likely be relatively passive and protective.  Chen will occasionally gank the safe lane and mid, but their overall safe lane pressure on the suicide lane Bounty Hunter would be weak and they will not be able to offer Bulldog any support in the Radiant bot lane.

Taking advantage of Alliance’s early commitment to Chen, Power Rangers goes greedy with their final three selections: Weaver, Treant Protector, and Chaos Knight.  Treant Protector in particular is a hero I’ve been down on since his recent nerfs, but Power Rangers know that Alliance’s lineup will not be able to contest his pull camp farm effectively.  They pair his selection with Weaver, a hero that he synergizes well with and that doesn’t require a babysitter against Batrider, freeing Treant to find all the farm he needs for quick ranks in Living Armor.  Meanwhile, Wisp can find those relocate levels from mid lane pulls and through babysitting Chaos Knight.  It’s pretty much the greediest support draft imaginable, but Alliance doesn’t have the tools to contest it.

The other aspect here is that not only does Living Armor tilt the Weaver vs Batrider matchup in Weaver’s favor, it also turns suicide lane Bounty Hunter into a credible threat against Alliance’s relatively weak safe lane.  If my memory serves me, Living Armor allows Bounty Hunter to evade an early Chen gank, and then immediately destroy both of Alliance’s sentry wards with one of his own.  One of the keys to a good Treant draft is using him to get more out of the laning phase than your opponents, and Power Rangers successfully uses Living Armor to successfully lane three heroes with some degree of right-click carry potential.

The final thing Power Rangers’ draft has going for it is that it’s very unified in its timing goals.  As I’ve pointed out previously, Treant Protector as a hero peaks very early and falls off as the game goes on.  Wisp also has its greatest influence around the same point of the game immediately after it picks up Relocate.  Given the early-game skew of their supports, Weaver and Chaos Knight make perfect sense as carry pickups.  The big liability to Chaos Knight as a hero is that he does not scale in in the late game nearly as well as most other carries, but this isn’t a liability if your entire lineup is oriented to dominate the game between levels six and eleven.  Bounty Hunter also fits in perfectly to scout out these early ganks and turn them into a massive gold advantage through track kills.

In short it was a smart draft.  They eliminated Alliance’s most problematic heroes, forced Bulldog out of his comfort zone, took advantage of the expected Chen pick, and put together a draft with a unified timing window.  It’s not a strategy that can work all the time because it’s very dependent on having first pick for Wisp, but it’s still one of the best examples of reading and countering a team’s tendencies I’ve seen this year.

Mid Only, Capture Points, and Cognitive Load in PvP

September 14, 2013

I’m a bit burned out on the TI3 stuff and too many people have been around this week to get any new scripting done, so instead I’m going to do one of those self-indulgent game design philosophy posts.  Next week we’ll get back to the topics you actually care about.

Anyway, one of the biggest trends in the Dotalike genre has been the rise of the all-mid mode.  I don’t know the order of inception off-hand, but both League of Legends and Heroes of Newerth introduced All-Random, All Mid and Mid Wars respectively and given them their own separate matchmaking queue.  Dota 2 has also recently introduced Mid Only as a game mode in private lobbies.

The other trend is an emphasis on capture points, largely driven by League of Legends’ Dominion mode, which in turn was pretty influenced by World of Warcraft’s Arathi Basin.  Many of the genre newcomers have adopted capture points under the perception that this mechanic allows them to remove (or at least de-emphasize) the laning stage.  The hope here is that by making the game more about action and less about last hitting, capture point maps will have a greater appeal to the more casual demographic of the genre.

When it comes to making a low-investment, “casual” game, Mid Only modes have been incredibly successful and capture point maps have not.  This is driven by the fact that while both options simplify the game compared to the standard 3-lane map, Mid Only modes effectively reduce the complexity of the strategic thought necessary to participate in the game, which reduces the cognitive load for a newer player.  Capture point maps simply replace the broad collection of strategic decisions in a 3-lane map with one constant and actually rather complex strategic decision.  As a result,  capture point games are also simpler but not actually less cognitively taxing.


Before getting into the details, I’d like to point out that the evolution of a streamlined, linear map design is actually a much broader phenomena that shows up in a lot of competitive genres.  One of the biggest examples is Alterac Valley in World of Warcraft.  Back in vanilla, AV was the most mindless, Zerg vs Zerg form of PvP possible, with individual games going on for over 24 hours with no actual victor and virtually none of the original players remaining in the game.  And yet, AV was by far the most popular PvP map in the game (at least in part due to the honor mechanics).  Also of note here is that one other popular form of vanilla world PvP, the emergent Hillsbrad war between Tarren Mill and Southshore also resembles the tug-of-war that you see in Alterac Valley.

The other big example that comes to mind is 2fort in Team Fortress 2.  Go into the server listing and you’ll find no end of 24/7 2fort servers with entirely too many people for either side to successfully capture the intelligence in a reasonable amount of time.  Instead, you have just a constant flow of aggression into the two main choke points of the map.  It’s effectively old school Alterac Valley recreated in an FPS.  On top of that, the basic concept of the standard Payload Map is the same sort of linear push refined into a more standard attack/defend game mode.

Part of what drives these modes is the absolute chaos created by putting too many people into too little space.  I personally hate these modes and find them both pointless and mindless, but I think a lot of people like the chaos because it turns PvP into a relative safe space.  As long as you’re spamming your abilities, you’ll eventually get kills; if you die, you can rationalize it as being impossible to avoid, and given the environment, you’ll probably be right; and finally, if your team loses it’s impossible for it to be your fault, and in many cases either team losing is effectively impossible.  It can be a reasonably nice environment to learn about the mechanics of PvP without being overwhelmed.  Of course, it can also be an environment that players go to exclusively so that they never actually have to learn about PvP.  But with dotalikes and All Mid, we’re not really talking quite so much about the overcrowded aspects of linear, streamlined PvP modes, so let’s put this aside for now.


What I want to focus on the most basic decision a player has to make in a PvP game: “Where do I go?”  It’s actually a pretty complicated question in general.  3-lane maps have evolved this basic expectation of 2-1-2 that then gets altered by higher level concepts like jungling and trilanes.  But the expectation of splitting up relatively evenly is established by the importance of finding farm, which simplifies the initial strategic decision of where you’re ‘supposed’ to be.  Eventually one team brings 5 people somewhere, and then the other brings 5 and you get teamfights.  Weaker players get confused about the part between the laning and the teamfights.  This is why if you watch lower level games you’ll either see a ton of people wandering around aimlessly in a single lane for way too long waiting for a teamfight that will never come.  Alternatively, you’ll see a bunch of static laning for way too long instead of ever attempting to gank anything or even jungle.  Strategically, both laning and teamfights are pretty easy; it’s the stuff in-between that requires complicated decision making.

The problem then with capture point maps is that they’re entirely take space in that strategic grey area.  An effective Dominion or Arathi Basin team is effective because they divide their forces intelligently.  They read the map and can make a good estimation of the minimum number of forces they need to send to capture a new point or safely delay the enemy’s attempts.  But the average teams playing these games are not effective.  They’re a group of strangers with weak to non-existent communication.  As an individual, your movement decisions are basically gambles on not only the positioning of the enemy team, but also on where the rest of your team will choose to go.    You could hypothetically be in a situation where your group has a choice between two options and either option will work so long as everyone in the group makes the same decision, and then when you fail because you split up, both side will blame the other for not going with their plan.  But if you never split up your team is guaranteed to lose.


So coming back to the idea of cognitive load, basically it’s the theory that when learning a new subject you can only handle so much incoming information at once.  As you grow more familiar, you develop schemas that free up your memory for other uses.  To take the idea to Dota, a new player doesn’t know anything about heroes or items, but eventually they develop schemas like “Queen of Pain is a mid hero that ganks heavily between 7-11,” or “Mekansm is an item I should build on anything that’s not a hard carry if my team doesn’t have one.”  When you have this heuristic understanding on what different heroes need, what they can accomplish at any given time, and what item build you need to go, you then can spend more time thinking about other more complicated strategic or tactical decisions.

And if we follow this line of thought, the most successful introductory modes would be the ones that reduce cognitive load while having a structural similarity to the more complicated game.  Capture Point modes fail this test because outside of the initial move of the game, there’s never a moment where the decision on where to go is simple and straightforward.  Conversely, in all mid games the decision on where to go is always simple and straightforward in that you can only go in one direction and you will eventually run into the single fight on the map.  So if I’m a new player unfamiliar with Dota or LoL or HoN I can just get into an all mid game and learn about the heroes and items in relative isolation.  Capture Point maps do not provide this environment.


I think eventually Dota 2 should look into some kind of a mid only queue, possibly with a relatively limited hero pool and all random.  In some ways it competes for attention with the main mode, but I think it serves as a better casual gateway into real Dota than, say, Easy Mode which just provides instant gratification for the worst strategic tendencies of the typical player.  A well designed mid only mode would do a much better job of emphasizing clever play using supports and semicarries, which could honestly make the main queue more enjoyable if it has enough influence.

And the takeaway from all this for any of these new games trying to evolve away from the basic 3-lane construct is that you really need to step back and evaluate the strategic decisions a player needs to make to play your game successfully.  Most importantly, you should:

  1. give players a limited number of objectives at the start of the game that split the players’ attentions in a relatively predictable way
  2. gradually expand these objectives as the game goes on in a way that will shake up the players’ behavior
  3. add incentives to the later portion of the game that periodically focus both sides on a particular objective in order to provoke large teamfights before one team has a large advantage

This lets you create a game that has strategic depth without ever feeling overwhelming.  The initial decisions are simple and straightforward, and the later decisions are bounded in a certain way so that they’re punctuated (and to some degree telegraphed) events that eventually culminate to a climactic fight.

The Alliance’s Naga Siren

September 11, 2013

Originally published on Team Liquid (Gfx: riptide, Heyoka // Editors: TheEmulator, Firebolt145 // Photography and art via Valve and R1CH)

Last week we talked about Naga Siren at TI3, but today I have a confession to make: it wasn’t so much about Naga Siren as it was about Alliance’s Naga Siren. Why would I write a hero profile but focus almost exclusively on a single team?

All stats provided by http://www.datdota.com

Naga had the 12th highest draft representation at TI3 with a 63.9% Pick/Ban rate, but in Alliance’s games that rate shot up to 100%, almost a full 20% higher than the next closest team. To break things down even further, Alliance also had the highest draft rate of Naga Siren. Looking at the chart you might think that Alliance would be third after LGD.int and Orange, but those rates include both the team and all of the team’s opponents. Roughly 33% of the Naga appearances in Orange games were picks by their opponent, and for LGD.int, opponents accounted for over 50% of Naga’s usage. Alliance’s pick rate is 100% theirs. As in they never allowed their opponents access to Naga, even going so far as to ban her immediately in five of their games. Long story short, the emphasis Alliance put on Naga during TI3 was entirely unparalleled.

So what makes Alliance’s Naga so different from the average Naga? Part of it stems from the fact that Alliance was one of the first teams to experiment with using her as a support, with the first attempts being a pair of games back in January when they were still No Tidehunter. Conventionally, Naga is seen as a frustrating hero that can reset teamfights almost indefinitely in the late game with her ultimate. The problem is that if you’re playing Naga as a support, it’s going to take you a while to get the 120 second cooldown Song at level 11, and you may never reach the 60 second cooldown at level 16. When you play Naga as a support there is this pressure to make the most out of each cooldown, and I feel that this has made them less complacent about their Song of the Siren usage. Ideally, Alliance doesn’t use Song to merely reset fights; they use Song to win them.

This also complements Alliance’s broader gameplan, which I feel revolves around objective control. Alliance loves to use tower pushes and Roshan attempts to provoke teamfights that they feel confident in winning. For example, their emphasis on having Dire side and Nature’s Prophet is based on using Prophet’s Treants to do Roshan early and safely, and if the enemy contests Roshan, Prophet can play extremely aggressively and then buyback and teleport to turn the teamfight into a 6 vs 5 contest. Similarly, Alliance loves Song of the Siren because it allows them to dictate the initiation of any teamfight, either by using Song to catch their opponents in a positional disadvantage or by using it to completely negate their opponents initiating ultimates while setting up their own initiation combos. But enough talking about the idea of what Alliance wants to accomplish; let’s move on to some actual examples.

[Prelims] Alliance vs LGD.int Game 1 — stats // VOD

We’ll start with Alliance’s very first match of the tournament against LGD.int, and jump ahead to 27:00. Alliance catches Dragon Knight under his mid tier 2 tower with Batrider’s Lasso, and immediately go to Roshan knowing that his buyback is on cooldown. As they’re leaving the mid lane, Akke on Venomancer lays down a very important ward.


LGD.int unknowingly marshals their forces under Alliance vision.

This ward reveals that LGD.int is massing their forces at their mid tower. If they go south, Alliance will know their exact positioning, and if they disappear from ward vision, Alliance will know to expect them to be wrapping around the other side of Roshan. As it turns out, LGD.int sends Rubick, Chen, and the teleporting Beastmaster directly south through Alliance’s vision, and it leads to this.

Spectre’s illusions are neutralized, and the rest of her team is trapped as the Aegis is secured

Thanks to their warding, Alliance knows precisely what to expect and immediately responds to Spectre’s Haunt with Song of the Siren while also catching the rest of the team in the sleep radius. Alliance calmly finish up Roshan before surrounding the sleeping members of LGD.int. In the end, Song of the Siren combined with good warding allows them to safely convert a single Dragon Knight kill into a Roshan and three more kills.

Moving a bit ahead in the game to 31:00, Alliance is looking to turn their now substantial lead into a final Barracks push. Batrider attempts a blink+Lasso initiation, but eats a Beastmaster Roar into Telekinesis for his trouble.

Batrider limps out as Gyrocopter prepares a 5-man Call Down.

But no problem because Naga’s Song can get Batrider out safely, while also setting up all five LGD members to eat a Call Down. Not only is this another (and one of many) example of Naga nullifying the opponents initiation ults, but it also serves as an example of how Alliance prefer to combo with her. In TI2 she was known for setting up massive teamfight combos, but Alliance prefer to instead draft a balanced team that happens to have just one or two heroes that can capitalize off of Song. This gives them a strong teamfight presence without making them overly dependent on their ult cooldowns. For example, Alchemist with Unstable Concoction is another hero Alliance likes to use to combo out of Song.

Ultimately, Alliance do not get the Barracks thanks to buybacks, but they do get the tower and win the teamfight 4-0, which sets them up for an eventual victory.

[Prelims] iG vs Alliance Game 1 — stats // VOD

Another instance of the +4 man Call Down under the cover of Song.

Just to prove that the setup we saw in LGD.int wasn’t a fluke, here’s almost an exact replication. Naga again uses her ult to both cancel a Spectre Haunt while also setting up a 4-man Call Down. And this is actually the second time in the match that Song shut down Spectre during a Rosh attempt, with the first coming at 36:11 on the in-game timer.

[UB Quarterfinals] LGD.cn vs Alliance Game 2 — stats // VOD

Changing gears a bit, here’s an example of Alliance using Song to turn an enemy tower push into a winning teamfight in a situation where they are in a losing position. It’s no secret that the early part of this game did not go well for Alliance’s aggressive trilane. At the beginning of the clip, Alliance is down 8-3 in kills, and it becomes 10-3 once Puck and Nature’s Prophet get caught trying to push mid. LGD immediately seeks to capitalize off of their two man advantage by pushing mid.

Naga wandering around while massively outnumbered.

As the tower is nearing death, Naga uses an ult that looks like nothing more than a denial attempt, but they don’t actually back down from a 3v4 with their two most leveled heroes dead. The important thing to notice is that with half of the Song left to go, Puck and Prophet have only seconds left on their respawn, and Puck has a TP scroll and Blink Dagger. Prophet is, of course, Prophet.

Puck’s blink initiation lights up the screen as the typical Prophet contribution mops up the creep kills.

The Puck initiation combined with Wrath of Nature melts most of LGD, and Prophet teleports in to help clean up Magnus. Gyrocopter, Alliance’s only death buys back, and they immediately turn their attention to Roshan. It’s a slow attempt, and LGD contests, but Alliance barely pulls it off, thanks in a large part to one of those Prophet buybacks I talked about earlier. At the end of everything, LGD’s 10-3 lead turns into a narrow 13-10 game, and Alliance takes Roshan but loses the Aegis. Most importantly, LGD’s 4k exp lead evaporates into a 1k deficit and their gold lead is cut by 2.5k.

And then there are the BKBs

We’ve now looked at a small sample of the winning teamfights set up through Alliance’s excellent Song of the Siren use. They have a clear understanding both of what enemy abilities they’re attempting to nullify as well as what they need to do during the seven seconds of sleep to win the ensuing teamfight after the Song ends. But there’s another distinct use of Song of the Siren that I want to draw your attention to: the way Alliance uses it to turn a Black King Bar into a liability.

[Prelims] Rattlesnake vs Alliance Game 1 — stats // VOD

The earliest example of this takes place towards the end of Alliance’s first group stage match against Rattlesnake. Alchemist buys a BKB and immediately attempts to use it to stun Naga with a concoction, but EGM gets song off as the stun travels towards him. Alliance quickly converge on Alchemist, and the rest of the team can only watch helplessly as their carry is isolated before the fight even begins.

BKB doesn’t do much when you’re netted and your entire team is asleep.

Aside from this game, the BKB isolation play doesn’t happen during the group stages. This is largely because in the five games that Alliance played Naga Siren, their opponents only bought two BKBs. That level of deterrence alone speaks for itself, but in the Upper Bracket Finals we get yet another demonstration of the havoc Song of the Siren wreaks on carry item builds.

[UB Finals] Na`Vi vs Alliance Game 1 — stats // VOD

Towards the end of the first game of Upper Bracket Finals, Na`Vi gets drawn into a losing teamfight. Na`Vi’s Gyrocopter looks to create some kills, but with two Na`Vi players already dead, he quickly finds himself overextended. He attempts to use BKB to retreat, and almost instantly Alliance responds with the Naga ult they have been saving, using it to isolate and focus Gyrocopter down.

For an idea of the timing, Naga is already in the Song cast animation while Gyro’s BKB duration is mostly full.

Minutes later, Na`Vi is forced into a teamfight at their bottom Barracks. Gyrocopter uses his BKB, and again, Alliance responds with an immediate Song of the Siren. This time Gyrocopter barely manages to escape to the fountain. It’s a minor point as the fight was likely already in Alliance’s favor, but it reinforces the pattern that they’re using Song reactively in response to that golden BKB shine.

[UB Finals] Na`Vi vs Alliance Game 2 — stats // VOD

Before moving into the final game, I want to remind you that Alliance did ban Naga Siren five times during the entirety of TI3 and that all five of those bans came in the opening round. Why would Alliance deny themselves the opportunity to pick one of their favorite heroes? Because all five games were games where Alliance’s opponents allowed them to have Wisp. Alliance clearly believes in Naga as a counterpick to Wisp, and in this game we get a demonstration why.

Song gets used a half second too late to prevent BKB.

The end result of this play is a 2-0 kill advantage to Na`Vi, but if Luna does not have a BKB here, Naga would have had a good chance to completely negate Na`Vi’s Relocate+Eclipse combo. It’s true that Relocate has half the cooldown of Song at level 6, but if you can counter a Relocate gank and then play safe for the remaining 90 seconds you’re neutralizing a huge portion of Wisp’s influence over the game. Wisp might also sit on Relocate longer looking for an opportunity that won’t leave his carry slept and potentially untethered at the end of the timer. If this is the case, then Song is giving you extra coverage as a deterrent.

But in this case, all of that is moot. Luna uses her BKB before Song and gets two kills off of the successful Relocate. Naga’s presence did necessitate purchasing the BKB, and the problem for Luna is that Naga’s presence demands that she buy a BKB if she wants successful Relocate ganks but then makes that BKB useless in teamfights.

Note the unused BKB in Luna’s inventory.

Towards the end of the game, we have one final teamfight. Upon rewatching it, what stood out to me is that Luna never uses her BKB despite taking plenty of Radiance burn and Urn charge damage. However this makes sense when you remember the way that Naga isolated the Gyrocopter in the last game. XBOCT knows that he’s dead here if he pops his BKB, so Naga has effectively turned a 3900 gold item into an expensive stat stick and has removed one of the strongest active item effects in the game entirely from the equation.

Ultimately, I believe that it’s this BKB deterrence that really makes Alliance’s Naga so threatening. Every other big teamfight ultimate in the game can either be neutralized by BKB (like Ravage and Dream Coil) or has strict positional requirements or other downsides (like Black Hole and Chronosphere). Song of the Siren is an incredibly versatile ultimate in the hands of a team with strategic planning on the level of Alliance. It has the capability of shutting down so many strong teamfight abilities with no obvious downsides, and it comes on a hero with a strong laning presence as a trilane support and other lategame perks such as the vision from Mirror Images. It’s no surprise that Naga was an immediate ban from Na`Vi in every single game of the Grand Finals, because right now she is easily one of the most threatening heroes in Alliance’s arsenal.

The Alliance’s Naga Siren [Link]

September 10, 2013

Available on Team Liquid.  I’ll archive it here in a day or two.

Google Drive Link for 6.78 Farm Dependency

September 9, 2013

Couple people asked for it in the comments, so here it is.  I’ve kinda drifted away from using Drive extensively, so it might be a bit sloppy.  I also included my trend estimate for the N->VH comparison as the final tab, but don’t read too much into it.

Actual content should be coming tomorrow, provided everything goes according to plan.