A Statistical Look at Anti-Mage’s Declining Win Rate

In 6.80, Anti-Mage had a terrible pub winrate of 41.59%. The problem is that in 6.76 he had a considerably higher winrate of 46.58%, and his only version changes since then is a buff to his Mana Void AOE. His skills have not been nerfed, and none of his core items have been either. Manta, Vladmir’s and Butterfly have been untouched, while Vanguard and Battle Fury have been buffed slightly. Despite the hero receiving buffs to his skills and items, his winrate dropped a massive 5 raw percent. There is something serious going on for a hero to fall so far with no direct nerfs.

Scouting the Cause

6.76 ended on December 2012, while 6.80 ended over a year later in April of 2014. My first thought was the changes to the gameplay of Dota during this time frame had a massive negative effect on Anti-Mage. Roshan has been buffed; the Jungle, Buybacks and Offlane were reworked, while Timbersaw, Medusa, Troll Warlord, Tusk, Bristleback, Elder Titan, Skywrath Mage, Abaddon, Earth + Ember Spirit, Legion Commander, Phoenix and Terrorblade were all added into Dota 2. But the change that caught my eye the most, was the changes to how people were itemizing Anti-Mage.

AM Items 676 You will notice that Manta Style was very popular, but was neck and neck with Battle Fury and Vanguard as all 3 were in around 40% of Anti-Mage inventories when the Ancient fell. Quite further down the pack was Vlads, Butterfly and Basher. In the 15 or so months since 6.76, people were building Anti-Mage very differently.

AM Items 680

While AM’s winrate dropped 5 raw percentage points, Battle Fury’s build rate skyrocketed from 40% to 60%. Vanguard more than halved from 39% to 16%. while Manta fell a few points. Butterfly has very interesting stats, while dropping in build rate significantly, the win rate also went up massively from 64% to 82%. This all probably due to perceived build effectiveness by the Dota community. With less players getting Vanguard and more Battle Fury, more players are opting for the late game and leaving their team out to dry in the mean time. I also gather that Butterfly’s changed win rate is a result of people getting later than they did in 6.76, and thus an AM who gets to pickup a Butterfly is having a better game than the one who never gets there in the first place.

Battle Fury Season

I have found that since 6.76 there seems to be an increasingly common belief that Battle Fury is core on Anti-Mage, and that Vanguard is generally a pretty bad item. The data on the changing pickup rates on these items support this. The problem is that Battle Fury’s win rate doesn’t reflect it’s effectiveness. The Battle Fury win rate is higher than the average winrate, but in 6.80 he still loses more games with a Battle Fury than he wins with one. Vanguard is in a similar situation. The difference however is that Vanguard is an item that you will always eventually replace, where as Battle Fury is only occasionally replaced. Once AM gets more and more farmed, he will replace the Vanguard, just like people replace their Drums of Endurance on other heroes. This means that Vanguard’s true win rate is most likely higher than the data gives, since the more successful AM’s who go Vanguard replace it before the game ends. Battle Fury is also replaced, but it is replaced way later into the game, and sometimes is still an item worth keeping. The likelihood that a Battle Fury will be replaced is thus much lower than Vanguard’s.

A 60.45% build rate for Battle Fury is pretty darn high. For comparison the most core item in the game – Boots of Travel on Tinker – is only 85%, where as the very common Blink Dagger on Sand King’s is 75%. There’s also the 3.6% of Anti-Mages who lose the game with a perseverance in their inventory, that is before they even manage to get a cleaving axe.

Ice Frog Knows Best

In the 6.81 New Bloom patch, all the default suggested builds for heroes were changed. This removed a lot of things people had gripes with, such as the effectiveness of Vanguard on Batrider, Razor, and Viper, and Battle Fury on Bounty Hunter. These heroes now all now have different items in the same spot. However Anti-Mage didn’t get his Vanguard removed; it’s still listed as core while Battle Fury is situational. Despite the growing sentiment that it should be the other way around, the build remained. This suggests to me that the reasoning as to why Vanguard is not core on Razor was more effective to Valve and Ice Frog than the reasoning as to why Vanguard is not core on Anti-Mage. I’ve heard a lot of people both in game and on sites like Reddit call Vanguard on Anti-Mage trash. And yet Razor, Bat and Viper got theirs removed, while AM didn’t?

I agree that Mekanism is better than Vanguard on Razor and Viper and that Blink is better on Bat. But I disagree on how people feel about Vanguard on Anti-Mage. The problem I have with Anti-Mages is that they commit your team to things they often are ill equipped for. The majority of Anti-Mages tunnel vision the late game, no matter what. They are likely of the mantra that AM is only effective if he can get to the late game with an advantage, for which he requires a Battle Fury. Often before he upgrades his boots. This suggests to me that AMs aren’t building depending on the game. If I’m a Phantom Lancer or Spectre I don’t tunnel vision Radiance, I go for one if the game permits me to. Sometimes it doesn’t and I settle for other items. AMs don’t seem to do this. This results in Anti-Mages that get poor farm, still buying the same items as a good farm Anti-Mage. If you’re losing hard as Lone Druid, you generally settle for Maelstrom or something else, and skip Radiance. Because it is of utmost importance that you are able to fight earlier, because the game is likely to end before the late game even comes.

The New Heroes and the Meta

Perhaps this isn’t the problem though, there are a host of variables that change a heroes’ win rate, and itemization is just one of them.

The offlane and jungle were changed during this time period too. But if these factors were decreasing AM’s win rate, I would also expect to see similar declines across the board for heroes of a similiar role.I found no correlation between the offlane/jungle changes impact on Anti-Mage when comparing him to the win rate variance of heroes such as Medusa, Faceless Void, Phantom Assassin/Lancer and Spectre.

The most played heroes are somewhat similar to 6.76. Heroes like Drow, Invoker, Pudge, Riki, Axe and Sniper are all still popular. However Pheonix, Terrorblade and Bloodseeker were all hot picks in 6.80. Bloodseeker is especially of note, since he is statistically the best hero against Anti-Mage, while Pheonix is statistically pretty bad against AM. This creates a constant balancing act, and an incredibly difficult factor to gauge. That is how the most played heroes in the game match up against a certain hero, since the hero is more likely to come up against the most popular heroes. This means that if a hero is bad against the top heroes, your win rate is likely to be unfairly low.

An Act of Balance

Finding the exact reasons a hero’s win rate is declining is almost impossible. But I can find statistical differences between two points in time, and then attempt to draw conclusions from them. Statistically, Anti-Mages are becoming less and less flexible with their itemization and strategy, opting to play the hero the same way with the same items more than before. Analytically, I think this constant tunnel visioning, leaving your team out to dry for 30 minutes, and a lack of flexibility are having a negative impact on Anti-Mages winrate.

However I accept that the heroes that are picked vary substantially during these two time periods, and it’s immensely difficult to discover the exact impact the changing playrates of over 100 heroes have. Anti-Mage had a 40% win rate is 6.80, where as Bloodseeker had 53%, and yet AM wins a sad 33% of games against Bloodseeker. Increasingly greedy Anti-Mage builds don’t help him weather the storm that BS is bound to create. Even if Anti-Mage counters are now more popular than before, going greedier than before won’t help alleviate these pains. After all, you can only get to the late game, if you and your team are able to survive every thing that comes before it.

Overall, Anti-Mage’s constant farming tunnel visioning has harvested a lot of food, a lot of food for thought.

Statistical Source: Dotamax


2 thoughts on “A Statistical Look at Anti-Mage’s Declining Win Rate

  1. Pingback: Bounty Hunter’s Massive New Bloom Shift | ziedota

  2. Pingback: Anti-Mage Win Rate Hits All Time Low | Zota

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