Final Thoughts on Kiev Major to TI7's conclusion: A Predictable Group Stage and an Upsetting Main Event

The International 7 is over, the Dota Season concluded, and Awkward Mixture articles about Defense of the Ancients will be few until next spring.  But before concluding the series on The Swapper, I wanted to record some thoughts while they were fresh.  

If you are interested in any of the series of the International, check out NoobfromUA's YouTube channel, whose creator does an excellent job condensing a three hour long best of three into the most relevant fifteen minutes.  

After the Kiev Major, I began a little project, keeping a strength ranking for each team which participated in a Premier Dota event, using a simple system.  I did a few articles about where each of the eighteen teams ranked before the Group Stage of TI7.  According to my system, six teams entered the Group Stage with a good chance of claiming the Aegis of the Immortal. VP, EG, NB, and Liquid had won major tournaments. TNC and Secret shown strong performances throughout the season even though they secure a win. But of these Liquid was clearly the strongest. They'd won all three events they attended post-Kiev including Epicenter, the event with the most skilled teams.

During the group stage, the public commentary focused on LFY and LGD, but Liquid was even stronger afterwards compared to the competition as NB, VP, TNC, and Secret lost ground to LFY and LGD.  Overall, the results of the the group stage were expected.   One of the only surprises was how poorly C9 did. They seemed a possible dark horse going into the group stage. But though they took a game off of VP, all their other victories in the group stage were against teams in the bottom quarter.

Of the eight teams who earned a spot in the upper bracket, seven of them were my top eight. Only Secret failed to record this achievement (IG took their spot).  But this wasn't surprising, because when I created my group stage, I split the top eight evenly, placing four in each group.  But Valve, put five of my top eight in one group, and three in another.  With Valve's groups, it was clear the best I could do was 7 of 8  (Another oddity of Valve's groups, they placed four of their six direct invites into group B, and two into group A)  It would be enlightening if they explained their process for group creation, as they have explained how they award direct invites.

One of the debates throughout the tournament was of group strength.  Was A or B superior?  PPD thought group A was the more difficult group, and my model agreed (even after the group stage).  As I said, they had 5 of my top 8.  But as the four winners of Group B (LFY, VP, NB, and IG) vanquished their group A opponents, this prior sentiment made its adherents a joke. Yet, in the end, Liquid (the victor of group A) vanquished the top 3 teams of group B (LFY, NB, and VP). Even if group B contained better teams, they were all beaten, one after the other.

But this is getting ahead of ourselves.  If the Group Stage was relatively predictable, the Main Stage was were it got weird.  Liquid's lost to IG in the first round of the Upper Bracket was the biggest upset of the entire tournament.  Ranked 1^st and 11^th by model at the time, it was an astonishing dismemberment of an incredible team. Watching all three games, it seemed as if Kuroky was desperate for Mind Control to make Bristleback work, while Miracle looked totally lost. Yet, even after their loss, my model still had them tied for first with LFY and Newbee who had just scored resounding victories. Liquid continued to build up strength after this defeat, and by their victory against LGD reclaimed their singular title in ability. At their peak, all five players were unstoppable, and their games were amazing to watch. Unfortunately, after Liquid's hard won battles with VP and LFY, Newbee failed to score a single victory in the best of five grand finals.  If one was hoping to find the best games to watch, Liquid's series vs VP or LFY are the best choice.  

A few other thoughts. In spite of the disagreement Reddit displayed when I disparaging OG's chances, labeling them as a middle of the pack team, I was right. They were able to score a 7^th/8^th place finish. (I thought they were anywhere in the 9-12^th spot).  Empire surprised by outperforming its season record. They were carried by a resurgent Resolution. He will be the most sought after free-agent carry, now his team Planet Odd has disbanded. 

And a final recommendation for Valve.  For those who follow the International, Valve releases a program.  Part of the program is a fantasy draft, and for each day, observers (those playing the fantasy draft) can pick two core players, an offlaner, and two support players.  Depending on how the players perform, the observer scores points.  Anyone who has participated in a fantasy draft for any sport will understand how enjoyable this can be.  But there is one issue.  On any given day of the nine days of the International, a player may be participating in two to six games.  But players are worth more points based on the number of games they play.  For any observer playing for as many points as possible, this system limits the options and skill.  
 
If (during the group stage), one player is in six games and another is in four, the only logical choice is the one playing six.  This error becomes especially egregious for the last three days of the Main Event.  During these last three days of the playoffs, anyone playing to win the fantasy draft has only two teams to pick from.  This is because, on the last three days, the winner of one set has to play another series later in the day, making them twice as valuable as the other players.  On Thursday one had to go all in on either Empire or Liquid, Friday it had to be either IG or LGD, and on the final day, one had to double down on Liquid or LFY (whichever team would win the lower bracket final). This is because players earn points for each game they play during the day, and some will play two best of threes.  On Thursday, there were five teams playing, but only two were acceptable choices.

Though the event is concluded, there are two simple solutions for next year.  One, Valve could cap each player's points at their best 2 or 3 games per day. Two, Fantasy draft could award points by adding up their score for each game during the day, and dividing by the number of games: an average.  Doing either of these would make every player viable.  It would still require observers to think about who is playing who, but so one could consider who is going to have a good day, not who has the most games.

For those interested, I have included three charts from my ranking system.  The first is a graph of the TI participating teams through the premier events.  The next is their game by game ranking through the Group Stage and Main Event.  Finally, there is a small graphic about where every team seemed to be ranked for me, according to the method of the International (1, 2, 3, 4, 5-6, 7-8, 9-12, 13-16, 17-18).


From Kiev to the End:

 

From the Group Stage through the Finals:

General Predictions Using TI Result Brackets:

 

My model didn't produce a perfect result.  But it did some things right.  Liquid came out on top, even though it had to pass through the entire lower bracket.  Of the pre-predict, seven of them reached the upper bracket, and the post-group predicted five of the final six correct, with Liquid, LFY, LGD, and VP even in the correct position.  On the other hand, IG overachieved on EG's failure, while Secret, TNC, and C9 failed to perform to their expected level.  

That's it for Dota for awhile.  The Swapper concludes next Monday.

Recent:

An Adaptation of a Conversation