TI5 Retrospective Part 1: The Meta of the Absent

September 24, 2015

TI5 Retrospective Part 2: CDEC, Masters of the Meta now available at LiquidDota.com

(Part 1 is also available here in what is likely a more mobile-friendly form)

When we talk about the hero meta at the conclusion of a major tournament like TI5, the discussion typically revolves around the heroes that dominated the Picks and Bans (Lina and Leshrac respectively), the heroes that put up absurdly high win rates over a large selection of games (for instance, Bounty Hunter’s 69% win rate over 48 games), or heroes that dominated all three of these metrics (Gyrocopter). But for TI5 a huge part of the story is the heroes that are missing, either as part of the 6.83->6.84 transition or just between the International and all of the other games of the 6.84 patch. This is because there’s a conspicuous absence that helps us to identify the connecting factor that led to the success of heroes like Lina, Leshrac, Bounty Hunter and Gyrocopter. And it also helps inform us on what we might expect from the inevitable 6.85.

With that said, I’d like to draw your attention to an unusual statistic: the ultimate cooldowns of the top 15 picks at TI5.


Without context, this doesn’t mean much, so let’s divide it into two groups: heroes with roughly minute long cooldowns or less, and heroes with nearly two minute cooldowns or more.

On the shorter end of things, we have every hero that has gone from having virtually no draft presence to being some of the hottest picks in 6.84.


The longest ultimate of this group is Undying at 75 seconds, but given that it’s a transformation ultimate with a 30 second duration, you’re looking at an effective downtime of 45 seconds. With that adjustment made, we have a group of composed entirely of heroes with sub 1 minute ults.

(I did, admittedly, leave off two heroes that fit my less than 5% P/B in 6.83 to greater than 30% P/B in 6.84 criteria but did not have as big of an influence on TI5 itself. These heroes are Dragon Knight and Visage. Dragon Knight is another transformation ult with an effective downtime of 55 seconds. Visage is the most complicated case of all given the nature of the familiars, but their expected downtime should be significantly less than their 180 second cooldown at level 6.)

“But what about heroes like Lina, Dazzle, and Storm Spirit?” you might ask. Well, let’s examine a second trend: the heroes whose TI5 selection rate outperformed their 6.84 expectations.


Once again you have a list completely dominated by short cooldown ultimates, albeit with the echoed exceptions from before of Naga Siren, Dark Seer, and Winter Wyvern.

Let’s flip the script now and look at the other end of things, starting with the three heroes I highlighted initially: Queen of Pain, Shadow Fiend, and Winter Wyvern. Of the top 15, they had the three lowest win rates, but you can’t really fault the drafters for expecting a better performance out of these heroes given their solid performances in the months prior to TI5.


But even if these heroes struggled, at least they got TI5 representation. The broader trend is that the 6.83->6.84 transition has coincided with a broad reduction in usage for nearly every hero whose gameplay largely depends on long cooldown ultimates.


If there’s a long ultimate cooldown hero that you’ve noticed missing from this list (not named Naga, Dark Seer, or Earthshaker), then it’s likely they were already a non-factor in 6.83.

Now, it’s certainly true that many of these heroes were already on the decline going into 6.83 or bore the brunt of a specific nerf in 6.84, but let’s examine a few specific cases of conspicuous TI5 absences.


Lion has had a very respectable 6.84 patch period. While he experienced a modest popularity decline in 6.84, from the second highest P/B in 6.83 down to eleventh in 6.84, his 6.84 win rate has been an acceptable .493, down less than a percent from his 6.83 performance.

On top of this, teams were highly valuing Lina’s ability to erase an opponent’s most valuable core through Laguna Blade. While there’s certainly a great degree of difference between the characters, it would be reasonable to expect Lion to see some play in situations where Lina wasn’t available due to their similar ultimates.

Despite all this, Lion’s went largely ignored by TI5 drafters, and when he did play he put up a dismal 7-12 (.368) record.


Due to his 6th highest P+B rate and very strong .567 win rate in 6.83, Juggs saw some nerfs in 6.84. While this definitely drove down his popularity in the current patch, he nonetheless maintained an even better .602 win rate going into TI5. Given the tournament’s emphasis on carries that lane well, come online quickly, and have some sort of escape, there was a decent chance that Juggernaut would see at least some niche selection. Instead, he went 2-4 (.333) and with only a single game outside of group stages.


With Thundergod’s Wrath at only a 90 second cooldown, he pushes the boundaries of my working definition for “long,” but he certainly does qualify as a hero largely defined by his ultimate. Magical burst has been very popular in 6.84, and so it might be reasonable to look at Zeus as a poor man’s replacement for the hugely successful and eternally banned Leshrac. Instead, Zeus’ win rate has cratered from .524 in 6.83 to .393 in 6.84. At TI5, Zeus’ record was an abysmal 3-8 (.273).


Ok, Warlock has been mostly a competitive non-presence in almost every Dota 2 patch. So why include him?

Well, MVP Phoenix’s March has the most recorded games with the hero at 28, a total almost triple the next closest player, and a lifetime record of 20-8 (.714). compLexity gaming also surprised a lot of teams by winning with Warlock twice in the regional TI5 Qualifiers.

Given that both teams had surprisingly successful performances, an unexpected pick like Warlock could theoretically helped them steal an extra series and push their playoff runs even further. However, neither team had confidence in Warlock working, and he saw zero picks and bans throughout the entire tournament.

I’ve thrown a bunch of stats out in an attempt to convince you that for whatever the reason these ultimate-based heroes have struggled or been outright ignored in 6.84 and that, with a handful of exceptions, this trend was amplified dramatically at TI5. The question then remains, what is it about 6.84 has caused this? If you can answer that question, then the inversion likely explains what it is that made Lina, Leshrac, Bounty Hunter and Gyrocopter the central players at TI5.

The first piece of the puzzle is the infamous rubber-band change introduced in 6.82. This change dramatically increased the gold and XP bounty per kill, and while this comeback mechanic has been toned down both shortly after release and in 6.84, it’s likely that there’s more of a networth shift currently at stake during teamfights than there was back in the 6.81 days. It’s a complicated subject and difficult to evaluate statistically, but even the watered down bounties of 6.84 put a greater emphasis on winning teamfights (or at least not losing them).

The second piece is that the 6.84 reduced the value of both lane creeps and many neutrals.

The third and final piece is that 6.84 boosted the “non-net worth portions” of hero kills by 10%, but also made changes so that a greater portion of the typical kill bounty goes to supports.

With all of these factors in play, it’s extremely likely that the ratio of kill bounty income to creep income was higher for this tournament than its been in any recent major. This first leads teams to emphasize heroes that fight well in the early and mid game, which then has a reinforcing feedback effect where teams need to draft lineups that can survive against expected aggression.

For an example in this shift of aggression, look at the approach to the hard lane by teams at the last three majors:

laneUsageTI5 was both the height in popularity for dual hard lanes as well as the most successful tournament for aggressive trilanes out of all Dota 2 majors. The contrast is particularly striking to DAC where both of these laning styles bombed in comparison to the more standard solo hard lane.

The implications of this laning shift were most pronounced for carries. Facing a triple whammy of reduced income rates, elevated safelane pressure, and increased emphasis on early and midgame teamfighting, hard carries died off almost completely, resulting in an approach to core investment similar in some ways to the one that dominated TI4.

One response, particularly in the group stages, was an increased reliance on semi-carries (heroes that trade raw right-click scaling for increased utility or burst damage), often paired up in dual or tri-core lineups. Lina, Queen of Pain, and Storm Spirit were the most popular choices for this, but Dragon Knight, Ember Spirit, and especially Templar Assassin were more niche examples that still saw a good deal of success. Leshrac would be included here had he not achieved what was essentially perma-ban status.

The other response was to build around a better scaling hero that could still somehow survive the pressures of TI5. By far the most popular and successful example of this is Gyrocopter. While Gyro’s scaling barely qualifies as a true carry, in a tournament of the blind, the one-eyed fighter pilot is king. More importantly, he possesses possibly the greatest deterrence to (as well as initiator of) laning aggression in the game in Rocket Barrage, and that ability combined with the low cooldown Call Down made him an early teamfight force in the way very few actual carries could hope to compare to.

Shadowfiend was the second most popular selection for this role, but we’ve already talked about his struggles. The next two most noteworthy selections were Phantom Lancer and Anti-Mage. Mobility in the form of Doppelganger and Blink allowed these heroes to survive early aggression, but of the two, Phantom Lancer was the far more stable pick. Anti-Mage was great at punishing passive teams or teams that bungled their aggression, but he took too long to come online against competent aggressors. Luna also deserves a footnote, but not much more than that and we’ll get to it later.

With all these factors in place, it becomes clear that the defining factor of the TI5 meta was early to midgame fighting. In order to win reliably you either needed to outright win these fights or to draw just long enough for your superior scaling hero to win out, and while many teams preferred that second, investment-driven strategy, the only safe centerpieces to run with it were Gyrocopter and Phantom Lancer.

And in this land of eternal war, ult-centric heroes had a tendency to be a liability. Their laning contribution tends to be weak, or at the very least, only strong in short bursts, and this is a greater than usual liability when aggressive duo and trilanes are at their historic peak. Past the laning phase, the teamfight potential of these heroes is strong, but only while their ultimate is available. When teams were as investment focused as they were at DAC this is not a problem, but maintaining a measured tempo against heroes like Storm Spirit, Tusk, and Undying, heroes that can blow everything in a fight and be ready to go again thirty seconds later…it’s not impossible, but it’s significantly riskier than it used to be, and I’m not surprised that the teams acted in a way consistent with the belief that the risk/reward payoff just wasn’t there for the majority of these heroes.

So what made Earthshaker, Dark Seer, and Naga Siren exceptions to this trend. I suggest three factors:

1. A long cooldown ultimate is less of a detriment if the ultimate is not the centerpiece of your kit.

You can imagine every hero having a ratio in teamfight value between their Ultimate and the rest of their abilities. A hero like Tidehunter, for example, is extremely skewed towards the ultimate side of things. Earthshaker and Dark Seer are likely more balanced with Fissure and Vacuum accounting for a greater proportion of their net utility.

2. A long cooldown ultimate is more prohibitive if it accounts for a larger portion of your net teamfight output. This means that not having a long CD utility ultimates or support ultimate is less of an expected loss than a damage ultimate or core ultimate.

Song of the Siren is a gamechanging ability, but a team with a support Naga Siren is still reasonably capable of winning fights with it on cooldown. Teams that include Juggernaut(Omnislash), Zeus (Thundergod’s Wrath), Chen (Hand of God), or Lion (Finger of Death) are less capable on average of winning fights without the full use of their ultimates.

3. Support Naga is simply back to being good

She was amazing in TI3, and fell out of favor when the post-TI3 changes made it more difficult to run greedy supports. 6.84’s increased emphasis on kill bounties allows you to run greedy supports again and farm them up through teamfights. This is a large part of why aggressive Duo and Trilanes worked as well as they did.

Additionally, Naga has always been very bulky for a support, and Ensnare has always been a great form of reliable CC. This allows her to contribute enough to fights even when Song isn’t available, and get enough experience out of these fights to carry her to the big cooldown reductions she receives at levels 11 and 16.

That covers the major points of the TI5 meta. In part two I’ll look at the impact of the meta on individual teams: how CDEC over-performed by finding their comfort zone, what EG did to take them out of it, and why Secret stumbled in a patch they seemed destined to dominate.

Editor: TheEmulator
Graphics: Ninjan, FO-nTTaX