Group Stage Statistics
The group stages of TI5 are over and despite somewhat stale meta we feel it was a blast. This group stage had it all: great teamfights, all-in rapiers, punished mistakes and great comebacks. It would be really hard to point to the best game in the tournament so far - a lot of them had us sitting on the edge of our chairs, rooting for the favourite team or simply enjoying the best Dota there is.
Many fans' souls were crushed in these group stages, yet the possibility of a comeback through the losers bracket gives a glimmer of hope. Behind-the-scenes stories of players push the engagement levels even higher - we are no longer looking at some people trying to make money and become famous. There is some sort of a personal connection many feel and the realisation that behind every hero there is a human, with his own aspirations, dreams, regrets and fears, is what makes a lot of games more alive and pushes the emotional levels even further.
And then there are numbers - cold, hard facts about every player and every game. These numbers have no concern about the feelings of the players, they can not represent regret and disenchantment for every mistake made or the joy of a turn-around victory. They do not show the full picture, yet they are there and they do tell a story. And it is our responsibility to tell you that story, no matter how disenchanting to the fans this story might be.
Teams
There are many reasons to look at the team statistics - it can provide an interesting perspective on the power levels of the teams in different aspects of the game. Given a relatively small sample size this statistics can be quite unreliable, but there are certain things which definitely stand out.
For example, LGD has a current KDA of 5.45 - an impressive number, especially when compared to an almost similarly successful Team Secret with the KDA of 3.14. In our Team Preview we already talked about how LGD has the lowest Death statistics on two of their main cores and it certainly shows. The team also has the best GPM statistics, but a relatively low XPM once again pointing at a very defensive playstyle.
It is not surprising that Team Secret holds the highest XPM stat - their map control and heavy pressure at all stages of the game provide the necessary farming space for their cores. And unlike other teams which excel early in the game, Team Secret do not lose their potency, partly due to well-made mid- and late-game decisions (something many Western teams struggle with currently), partly because of their somewhat greedy drafts and item choices.
The lowest KDA team who made it into the Winners Bracket is Team Empire. Given 4th highest Assist count and average Kill count, it seems like Team Empire are eager to frequently trade with their opponents, unlike other Winner Bracket teams, which exhibit more caution. They also have some of the more aggressive supports and offlaners.
Kills are the means of winning the game, rather than the objective, and it becomes even more apparent when looking at the statistics for some Winner Bracket teams - Complexity, Cloud9 and Empire all have more average deaths than kills. It does not indicate lower potential of the teams or anything of the sorts, but can be an indication of inconsistency.
On the other side we have Virtus.Pro in the Lower Bracket, who on average have 2.47 more Kills than Deaths in their matches. These players certainly know how to teamfight and trade effectively, yet their GPM is one of the lowest across all teams, which is certainly surprising given the fame of Illidan. Trading in the current patch can be very deceptive, and sometimes it is really hard to tell whether you have won a fight or not from the economic perspective.
VG, Na'Vi, NewBee, IG and Fnatic are all more or less similar when it comes to KDA stats. In fact, almost all of their team stats look very similar. It is going to be a hell of a fight in the lower bracket - aside from Virtus.Pro and the Korean teams, they all look equal.
Both Korean teams MVP.Hot6 and MVP.Phoenix are at the bottom of the leaderboard with 1.72 and 2.22 KDA respectively. It is quite unsurprising that they also have the highest amount of deaths in their games. MVP.Hot6 is also at the bottom of the assist list, potentially pointing at a rather poor support performance and low team coordination. At the same time, MVP.Phoenix has a rather high average GPM - it is higher than the average GPM of other Lower Bracket teams as well as EHOME.
Players
Every International so far has brought new faces into the highlight. For me personally, SoNNeikO and yoky- were such discoveries - they have been amazing in the games I've watched and I am convinced they have a very bright future in their respective roles. Yet from the strictly numerical perspective they do not necessarily stand out, hence the information in this section should be treated carefully - numbers do not tell the whole story and the tournament is far from being over.
Best Position #1
When you think of a carry player, KDA and GPM are stats that come to mind first - historically the job of a carry is to farm and kill. While I (and many others) disagree with the notion, these are the things we get to operate with if we want to be as objective as we can.
From this perspective, there are two players which really stand out - Secret.Arteezy and LGD.Sylar are top Position #1 farmers. Their GPMs are also well ahead of CDEC.Agressif, who takes the third place. It certainly helps that these players are both on the most winning teams, but watching most of the games myself I feel convinced that they are, in fact, very crucial contributors to the Teams' Win Rates in this tournament, rather than free riders on the backs of their teammates.
If we look at the KDA stats, it is LGD.Sylar at the top spot once again, proving my point. He is very closely followed by CDEC.Agressif and they are both well ahead of EG.Fear who takes the third position. It is worth mentioning, however, that CDEC.Agressif has the highest average Kill count of all three, which comes at a price of more frequent Deaths.
Best Mid
It seems there is simply no match to LGD.Maybe. He simply has it all and very few mid players even come close. He is cautious and patient, much like the rest of LGD. Is it coordination from xiao8 or the whole team chemistry? We don't know, but the next contender for the spot - EG.Sumail has less than half of LGD.Maybe's KDA.
To EG.Sumail's credit, though, his average Kill count is actually slightly higher AND it was not uncommon for him to play natural 
Bloodstone carriers. As I said - numbers don't tell the whole story. I honestly can't wait to see them battle it out. I really wish this game will also feature an early 
Bounty Hunter ban to make it fairer.
Best Offlane
CDEC.Xz and Secret.Zai are the two players who managed to have the highest impact on the game in their offlane role. Secret.Zai has already been a very well known offlane player and it is not surprising to see him do well, but CDEC.Xz and his 
Clockwerk are definitely worth the highlight.
Apart from these two, I think EG.Universe also deserves an honourable mention - he is the third offlaner in terms of KDA, but he has shown the flashiest plays of all three, especially on 
Sand King.
Best Supports
Typical statistics does not really translate well into game contribution for supports. Supports range from heavy damage dealers to healers and disablers. As I said, for me personally Na'Vi.SoNNeikO was one of the best supports in these group stage matches. He was flashy, on point and really entertaining to watch, despite how hard it was to watch Na'Vi games in general, especially as a long time Na'Vi fan.
CDEC.garder comes as a close second in terms of how unexpectedly good he was. Yes, I do think that Secret.Kuroky, Secret.Puppey, EG.Aui_200, LGD.MMY!, C9.BigDaddy and, probably, VP.fng are better support players, but the two I mentioned previously deserve your attention. Attention many support players really lack.
Heroes
As been mentioned previously, the Group Stage meta was rather stale with two heroes being picked or banned in every single game - 
Leshrac and 
Gyrocopter. Both of them have a really high Win Rate of 61.5% and 57.1% respectively, however given a relatively small sample size of 115 games I do think it is a bit premature to cry for nerfs. And it certainly did not feel like they were that dominant. Gyrocopter generally only worked out in the games where the supports were doing overtime with ancient stacking, while Leshrac has only actually been played 13 times.
Templar Assassin, 
Crystal Maiden, 
Dragon Knight, 
Storm Spirit and Bounty Hunter are the heroes which have all been picked more than 15 times and have a Win Rate of 60% or higher.
The bane of public matchmaking - 
Bloodseeker has been played 34 times with a Win Rate of 38.2% - certainly not something we expected for the hero. Another surprise came in the form of 
Winter Wyvern who was picked 29 times and has a Win Rate of 34.5%. In this case, however, it is less of a surprise given a meta revolving around magic/pure AOE bursts, rather than hard hitting carries. Her ultimate simply does not do enough in these popular scenarios.
24 heroes still remain unplayed in this tournament.
Trivia
Longest Game: IG vs. Fnatic Game 2 (1:37:47) Link
Shortest Game: Fnatic vs. MVP.Phoenix Game 2 (21:17) Link
Bloodiest Game: MVP.Hot6 vs. CDEC Game 1 (83 Kills) Link
Most Peaceful Game: Cloud9 vs. IG Game 1 (16 kills, all by C9) Link
Highest Hero Damage: MVP.Phoenix vs. IG (35.4k by MVP.QO) Link
Highest Pitch "Chuan" Shout: IG vs. Fnatic Game 2 (@1:37:05 by TobiWan) Link
Most Kills: MVP.Hot6 vs. CDEC Game 1 (26 Kills by CDEC.Agressif) Link
Most Deaths: Cloud9 vs. LGD Game 1 (15 Deaths by LGD.Yao) Link
Most Last Hits: IG vs. Fnatic Game 2 (1071 by Fnatic.Mushi) Link
Most Denies: VG vs. CDEC Game 2 (51 by CDEC.Agressif) Link
good post m8, although shouldnt the bloodiest/most peaceful stat be based on kills per minute rather than total kills?
nice stats
i think bone7 deserves a shoutout for being like lvl 13 at 60 min as clock in a game
Where is Tusk?
Since when is sonneiko only a ti discovery ?
He is (wthout anydoubt) better than 80% of the participants of this ti
which 24 remain unplayed? I doubt they will ever be played once things get serious :(
http://i.imgur.com/lwV82ul.png (Oracle and Earth Spirit don't count)
Why don't Earth Spirit an Oracle count?
Why no mention of Techies' high win rate?
@TheReal - they are not yet in Captains Mode, so they couldn't have been picked
@Dsrt - He was picked seven times so far. Too small of a sample size to actually say something. (Blog Post) - I did an analysis on the hero back in December marking his addition to CM as very important. The meta was a bit slow to pick up.