Jakiro: A Dragon Slain
Despite suggestions that a new patch might be coming soon, I've become more and more interested in analyzing the current metagame in competitive Dota 2. To to be clear, I'm a big fan of this version precisely because it's been difficult to establish too many trends in terms of what the patch is all about. Sure, we are generally seeing more activity in the early game than before. And yeah, there's a lot of pushing and ganking happening. But can't we discover anything more specific about 6.79? Is the metagame really so open? I like to think so - I think an open metagame is a good thing, placing great emphasis on the agency of players in deciding their own results. However, it seems quite unlikely that this is the case. More likely, I think, we are dealing with a metagame that is just a lot more difficult than usual to crack.
Thus, I had this idea: to try to establish some modest, but specific, conclusions about 6.79, I would take a look at which heroes are least picked in this version in competitive games and see how those heroes were doing in previous versions. Hopefully this would show that there are least some specific limitations being placed on picks by the current version. Below is a list of the 25 least-picked heroes in this version, as according to datdota. [25 entries but one is some kind of bug, so 24.]
Incredibly enough, only one of these heroes ends up being worth investigating. Sniper, Meepo, Skeleton King, Bloodseeker, Riki, Tusk, Zeus, Axe, Drow Ranger, Omniknight, Ogre Magi, Witchdoctor, Centaur Warrunner, Faceless Void and Brewmaster were all on the same page in the previous patch. Subtracting those 15 from the original 24, we have 9 left. Huskar was not available in CM in the previous version so he is also disqualified from potential significance. Meanwhile, Ursa, Skywrath, Sven, Warlock and Phantom Assassin were all only just barely more picked than the least-picked heroes in 6.78. All of them had around 50 or less picks in that version.
In 6.78, Phantom Lancer was picked 115 times and Spiritbreaker was picked 125 times. While this means they were not entirely fringe picks in the old version, they were still neatly in the bottom 50% of hero popularity overall and cannot compare to the 30 or so most popular picks of that version which all had upward of 300 picks, some even over 1000. The situation with PL and SB probably warrants further analysis – Spiritbreaker, in particular, had a very promising run for a while in 6.78 - but for the purpose of this article it's good enough to recognize that their drop between the two versions is not entirely dramatic. There is, however, one last hero – Yep, you guessed it - Jakiro. And at a glance, his drop in popularity does seem extremely dramatic. After only receiving a buff to Liquid Fire in 6.79, Jakiro has since been picked a mere 17 times compared to its 453 picks in 6.78 [the duration of the two patches is fairly similar at this point]. And with that, I have my first tentative conclusion – that 6.79's metagame is somehow not kind to Jakiro.
Now I'd like to try investigate why this is the case. Jakiro has historically shone as a hero providing good cover as primarily a defensive support, disabling well in teamfights, and spamming down pushes very effectively. One other major advantage of the hero is that it's an unusually tanky intelligence hero and thus can play the role of support without necessarily playing the role of an easy target. If 6.79 is about pushing and ganking, why is a hero that does well against both these things not being picked? Well, first, we should look at how well it was doing in the previous version. And indeed we find something interesting here. At 41,5% Jakiro had the lowest win percentage of any hero in 6.78 which was picked over 300 times. In other words, it did the worst of all the popular heroes. In fact, if you go even further back, Jakiro's win rate has been on a downward trend ever since 6.75. In each consecutive version, the hero has had a lower overall win rate in competitive Dota 2. This leads to a potentially simple explanation. In 6.75, Jakiro began to gain new popularity as its Ice Path was significantly buffed. It gained 100 damage instead of 0, a longer disable, a shorter creation delay and the ability to stun anyone who walks into it and not just those initially hit. Subsequently, in 6.76 and 6.77, the skill was nerfed 3 times. Mana cost was increased, creation delay increased, and damage made to scale instead of dealing 100 from level 1. Since this is probably Jakiro's most important skill, perhaps the story of Jakiro's decreasing win rate is really just the story of Ice Path and its changes. In 6.79, Jakiro's win rate has indeed again gone down, albeit over a tiny sample size. That said, while changes to Ice Path might begin to explain the decreasing success of the hero, it is not yet clear that its decreasing success is what is responsible for its decrease in popularity. Indeed, while its success decreased gradually over a few intervals, its popularity seemed to disappear overnight between 6.78 and 6.79. Why the sudden fall?
One thing that I thought might hold some answers was the distribution of Jakiro's results, especially in 6.78. However, here I found nothing too interesting. Looking at the support players from the top teams of the time, we see that EGM had an impressive 6-2 record with Jakiro in 6.78. Other than that, records vary between break-even, slightly positive and slightly negative. Perhaps this means that the overall win rate of the hero is affected by weaker teams doing badly with it. While there isn't really enough data to conclude this, there is a fairly good explanation for this line of thinking. Jakiro is a difficult hero to play very well. For example, it's pretty hard to hit all your Ice Paths [unless you're Fy]. Possibly, games where Ice Path misses a few times more, the hero's contribution becomes significantly less valuable and thus the overall win rate of the hero is compromised by the fact that only the best players have the ability to play the hero well enough to justify picking it.
But there is another thing that is very difficult to do well with Jakiro - meaningful early rotations around the map. I was unable to find the replay but have a distinct memory of Banana doing this really well in one game back at TongFu - this is literally the only game I can remember this happening in though. Surprisingly, Banana's record on Jakiro is 1-4, the worst record of any player I checked. In any event, in considering the difficulty in being active in the early game as Jakiro – because it's slow and bulky and doesn't have a guaranteed or fast-casting disable – further explanation as to its decreased popularity in 6.79 can be found. It just so happens that one of the only things we thought we knew about 6.79 was that it encouraged more active play early on in the game, especially by supports. It should be no coincidence then, that if Jakiro struggles to be successful with early rotations, the hero would see a fall in its popularity. Indeed, at its height of success and popularity, Jakiro's early game was mostly about stacking camps and covering carries. To back up this line of thought, I checked to see if other heroes with a similar playstyle have had their popularity similarly affected in 6.79. Indeed, Keeper of the Light, who also excelled most at a time where supports spent their time early on stacking camps and covering carries, has only been picked 44 times in 6.79 as opposed to its 386 appearances in the previous version. To a lesser extent, Lina also fits this mould and has also seen a significant shift from 239 picks in 6.78 to only 27 picks in 6.79.
There is much to be said about the fact that the 'best' supports at the moment are just so good that there isn't much room for other supports to be picked. As I've discussed elsewhere, In almost every competitive game, each team has at least one of Crystal Maiden, Visage, Venomancer, Rubick or Alchemist on support. Instinctively, you might conclude that the increased hegemony of these heroes is responsible for the decreased popularity of other supports. But I would argue that rather than one thing causing the other, both are symptoms of a greater cause. A large part of why these supports are so popular at the moment is precisely their ability to do what Jakiro, Kotl or Lina [certainly alone] struggle to do – create kills early in the game.
So in the end, what have I discovered? We already could have stipulated that the early game is more active than it was in the past. I suppose my investigation into Jakiro has merely reinforced this suggestion. Perhaps, the real value here, is to consider treating 'activity in the early game' as defining of the current metagame as opposed to merely something that happens in it. Maybe pushing is not about pushing and ganking is not about ganking but rather both are about finding ways for supports to be active in meaningful ways early on. Certainly, this would explain why both pushing and ganking continue to co-exist as dominant forces within the same metagame.