One of the neat features in the International Compendium (IC) is the option to make various predictions on the Western and Eastern qualifiers. Let’s try to make some educated guesses on the Western Qualifiers.

A few things to consider before beginning the analysis. The qualifier will consist of 33 to 50 games. A team can play as few as 4 games and as many as 23 games. I’ll examine data for all games, recent games, recent tournaments, and for each of the teams that qualified. Keep in mind though, that one or two teams will likely play an disproportionately large amount of the games in the tournament and their play style will heavily influence the results. If you feel strongly that one or two teams will dominate the tournament, adjust your picks accordingly.

1.) Most Picked Hero (3 Points)

The three main things I used for this:

  • Which heroes where picked most overall since March 1st
  • Which heroes where picked most in the recent, major western tournaments (G1 Western Qualifiers, DreamHack Invitational, the EMS Finals, The Premiere League IV, WePlay, and the Curse Invitational)
  • Which heroes each of the qualified teams pick

I relied heavily on the Draft Stats page to get the data for this category. Here is much of the relevant data collected into a spreadsheet. Feel free to explore the Draft Stats page if you disagree with some of my date ranges and tournaments and come up with your own data set.

The overall data suggests Magnus, Rubick, and Shadow Demon as the three most likely winners in this category. In the last couple of months, those are the only heroes picked over 50% of the time. The individual Western tournaments suggest Lifestealer, Gyro, Lone Druid, Nyx, Dark Seer, and Furion are also reasonable choices. I was leaning towards selecting Rubick until I saw how infrequently Rubick is picked in Empire and QPAD games. I am not sure how the newly configured Empire is going to regard Rubick but if they play like the old Empire at all, Rubick will not see much play in their games.

Magnus is a popular pick by all eight teams in the tournament and Reverse Polarity is too good to ignore. If I had any strong feelings on which two teams would advance to the finals, I’d alter my selection to pick a hero that is picked often in their drafts.

My Choice: Magnus

Honorable Mentions: Rubick, Shadow Demon, Lifestealer, Furion, Dark Seer

2.) Most Banned Hero (3 Points)

Overall data suggests Batrider. However, eastern teams largely ignore Wisp while Wisp is  almost always picked or banned in western tournaments. All the recent western tournaments have Wisp as a more banned hero than Batrider, with the exception being the  Curse Invitational (the most recent western tourney). Seven of the Eight teams in the tourney have Wisp as their most banned and more banned than Batrider. However many of these eight teams have had a lot of roster changes recently. The past draft data may not be completely reliable.

The data says Wisp. I would not be surprised though if there is a shift in the meta soon and Wisp begins to fall out of favor as a first round ban. However that could just be me watching too much eastern Dota.

My Choice: Wisp

Honorable Mention: Batrider

3.) Hero with the Most First Bloods (1 Point)

I’d never thought about this until I saw this question. I did a custom query to pull the data. It’s mostly accurate with the exception that it omits games where the first hero to die was from a deny or from being ‘killed by Radiant/Dire.’ The top first Blood Heroes Overall (roughly 2100 games):

Hero First Bloods
Shadow Demon 119
Rubick 95
Leshrac 92
Queen of Pain 80
Lifestealer 70
Chen 67
Dark Seer 63
Gyrocopter 63
Chaos Knight 61
Jakiro 61

The top First Blood Heroes since March 1st (roughly 775 games):

Hero First Bloods
Shadow Demon 61
Queen of Pain 37
Magnus 37
Gyrocopter 35
Rubick 33
Keeper of the Light 29
Leshrac 28
Lifestealer 28
Chen 26
Dark Seer 25

Shadow Demon is the clear leader. With Shadow Demon remaining such a popular pick and being an early game ganker and tri-lane hero, I’ll stick with Shadow Demon. Rubick, Magnus, or QoP all seem like good choices, particularly if you think a team that prefers those heroes will go far in the tournament.

My Choice: Shadow Demon

Honorable Mentions: Rubick, Magnus, Queen of Pain

4.) Hero with the Most Deaths (1 Point)

Stats since March 1st:

Hero Games Deaths Per Game Total Deaths
Rubick 557 5.2 2896
Shadow Demon 510 5.6 2856
Magnus 530 4.6 2438
Keeper of the Light 376 4.4 1654
Nyx Assassin 292 5.2 1518
Dark Seer 370 4.1 1517
Gyrocopter 358 4.1 1468
Queen of Pain 335 4.2 1407
Furion 301 4.6 1385

In my opinion, it comes down to whether you think Rubick or Shadow Demon will be picked more often. I’m leaning towards Rubick being picked slightly more.

My Choice: Rubick

Honorable Mentions: Shadow Demon, Magnus

5.) Highest GPM in a Single Match (5 Points) — ‘Less than 800 GPM’ versus ‘More than 800 GPM’

I have no idea how they decided which questions were worth which amount of points. Why  is a question with only two options worth 5 points (out of a total of 30 points)? Every other question is worth 3 points or less, including some of the questions that have more than two choices.

Oddly enough though I find this the hardest of the two choice questions. There is data that supports either choice. Heroes have gotten over 800 GPM 117 times in roughly 2600 games or 4.47% of the time. What are the odds that something that happens 4.47% of the time will occur at least once in 33 games? There is a 78% of that event occurring at least once. In 50 games? 90% chance of occurring. That suggests that over 800 GPM is the clear choice. However, many of those over 800 GPM games happened in lesser competitions where there is a greater difference in quality between the two teams. There were no over 800 GPM games in the TI2, none in the 2012 West Qualifier and only one in the 2012 East Qualifier.

In offline games, there have been only 13 occurrences of over 800 GPM in 470 games, or 2.77% of the time. Out of those 13 games, 5 occurred in the ASUS FBotY tourney which had some unknown teams playing the likes of NaVi and Empire. Exclude those games and  the odds are basically 50-50, with an over 800 GPM game having an under 50% chance of occurring if you think the tourney will only be 33 games and an over 50% chance of occurring if the tourney has 50 games.

Here’s a breakdown of the amount of 800+ GPM games by tournament.

Tournament Total Games 800+ GPM Games
StarSeries II Finals 56 1
Gosu MM Asia 35 1
Prodota2 Pro 34 0
ECAL Americas 20 0
The Defense II 111 0
BTS World Tour 78 0
The International 2012 153 0
StarSeries III Prelims 91 0
StarSeries III Finals 34 1
G-1 IV Prelims 53 0
G-1 IV Finals 9 0
Dreamhack 2012 Winter 39 1
GosuLeague IV 49 0
RaidCall I 76 2
GEST D2 Tourney 36 2
GEST D2: The Challenge 15 0
TPL III 119 4
GosuLeague V 54 7
The Defense III 173 14
RaidCall II 145 6
StarSeries IV Prelims 93 7
StarSeries IV Finals 14 1
ASUS FBotY 39 5
The Asia 33 1
G-League II 47 0
Killing Spree 12 1
SEA League 35 2
TPL IV 119 2
EMS Spring 2013 Prelims 79 5
StarSeries V Prelims 146 7
The International 2012 West Qualifiers 35 0
The International 2012 East Qualifiers 34 1
GEST: March 2013 87 8
Dreamhack Invtl. 42 2
WePlay 122 6
Armageddon GS 16 3
G-1 2013 Asian Qualifiers 95 8
Netolic II: East 39 2
StarSeries V Finals 15 2
G-1 2013 Western Qualifiers 24 0
AMD Premier League 64 4
GEST: April 2013 19 1
EMS Season 1 Finals 16 1
G-1 2013 Finals 15 0
GEST: May 2013 18 3
Curse Invtl. 14 1
Netolic II: West 38 5

A few other things to keep in mind. Black of mouz has 12, 800+ GPM games including one this past weekend against RoX.KiS and another in the EMS Season 1 Offline Finals last month against the might Alliance. blowyourbrain of Empire has had 4 such games. However both teams have had major roster reconstruction recently and may not make it far in the tournament. I guess I believe in Black and his farming Alchemist to get the job done, but I am not confident in my answer at all.

My Choice: 800 GPM or More (tepidly)

6.) Number of Heroes Never Picked or Banned (2 Points)

The choices for this question are in groups of five. Ie select 1-5, 6-10 etc. I believe there are 98 heroes currently in the hero pool and eligible for the Curse Invitational tourney. I believe the other tournaments in the following table had 94 heroes eligible.

Tournament Not Picked or Banned Picked or Banned
Curse 55 43
EMS Finals 49 45
G1 35 59
WePlay 13 81
TP IV 15 79
Dreamhack 31 63

The WePlay and TP IV tournaments are not very good comps as they had more than eight teams and took place over a much longer time period than this tournament will. Curse and EMS Finals had much fewer games and took place in a very condensed time frame. I’m leaning towards splitting the difference and somewhere in the 36-40 or 41-45 range. The importance of this tourney and the new roster construction may lead teams to be cautious or it could spur a bunch of pocket strats and unusual hero pickups. I wouldn’t mind seeing someone go hero by hero and giving their opinions on whether the hero will be picked or banned.

My Choice: 36 – 40 

7.) Number of Big Comebacks (1 Point) —  ‘3 or Less’ versus ‘4 or More’ 

I’m in the process of revamping my win probability model. Hopefully next week for the Eastern Qualifiers I’ll have better data and can make a better prediction. The data that I do have (982 games) says that a team has come back from being down 10,000 or more gold 32 times (17 Dire, 15 Radiant) or in 3.25% of the time. This is too rare of an event for me to predict it will happen 4 or more times in a 33-50 game tournament. Using the Binomial Distribution, the range of probability for ‘3 or Less’ comebacks is between 92% and 97% depending on how many games you think will happen in the tournament. Clear choice for the under.

My Choice: ‘3 or Less’

8.) Fastest Game Time (3 Points) —  ‘Under 18 Minutes’ versus ‘Over 18 Minutes’ 

I am not super confident in my data on this as occasionally short games that were not official due to bad replays, bad API data, DDOS attacks, etc. get mixed in with the good data. I believe I have eliminated all such games from my database. Under 18 minute games have occurred 9 times in 470 offline games, 35 times in 910 games since March 1st and 97 times in 2,691 games overall. Those percentages suggest the smart bet is on ‘Under 18 Minutes’ with the data from offline games suggesting it is almost a toss-up.

Here is how often under 18 minute games have occurred per tournament:

Tournament Games Under 18 Minute Games
StarSeries II Finals 56 1
Gosu MM Asia 35 1
Prodota2 Pro 34 2
ECAL Americas 20 0
The Defense II 111 1
BTS World Tour 78 4
The International 2012 153 3
StarSeries III Prelims 91 4
StarSeries III Finals 34 0
G-1 IV Prelims 53 1
G-1 IV Finals 9 0
Dreamhack 2012 Winter 39 1
GosuLeague IV 49 1
RaidCall I 76 9
GEST D2 Tourney 36 2
GEST D2: The Challenge 15 0
TPL III 119 6
GosuLeague V 54 4
The Defense III 173 3
RaidCall II 145 6
StarSeries IV Prelims 93 4
StarSeries IV Finals 14 0
ASUS FBotY 39 2
The Asia 33 1
G-League II 47 1
Killing Spree 12 1
SEA League 35 0
TPL IV 119 7
EMS Spring 2013 Prelims 79 5
StarSeries V Prelims 146 8
The International 2012 West Qualifiers 35 0
The International 2012 East Qualifiers 34 0
GEST: March 2013 87 2
Dreamhack Invtl. 42 3
WePlay 122 4
Armageddon GS 16 0
G-1 2013 Asian Qualifiers 95 2
Netolic II: East 39 3
StarSeries V Finals 15 0
G-1 2013 Western Qualifiers 24 2
AMD Premier League 64 1
GEST: April 2013 19 1
EMS Season 1 Finals 16 0
G-1 2013 Finals 15 0
GEST: May 2013 18 0
Curse Invtl. 14 1
Netolic II: West 38 0

The data says ‘Under 18′ but I have a gut feeling that it is going to be ‘Over 18.’ I feel that most of these teams are very evenly matched, have much to prove, and will be playing not only for this qualifier but a chance to be one of the wild cards.

My Choice: Over 18 Minutes

9.) Number of Greater than 60 Minute Games (3 Points) —  ‘5 or Less’ versus ‘6 or More’ 

Tournament Total Games 60 Minute or More Games
StarSeries II Finals 56 4
Gosu MM Asia 35 0
Prodota2 Pro 34 4
ECAL Americas 20 0
The Defense II 111 7
BTS World Tour 78 2
The International 2012 153 5
StarSeries III Prelims 91 1
StarSeries III Finals 34 1
G-1 IV Prelims 53 3
G-1 IV Finals 9 0
Dreamhack 2012 Winter 39 0
GosuLeague IV 49 0
RaidCall I 76 1
GEST D2 Tourney 36 2
GEST D2: The Challenge 15 0
TPL III 119 2
GosuLeague V 54 0
The Defense III 173 1
RaidCall II 145 2
StarSeries IV Prelims 93 0
StarSeries IV Finals 14 0
ASUS FBotY 39 0
The Asia 33 0
G-League II 47 4
Killing Spree 12 0
SEA League 35 1
TPL IV 119 2
EMS Spring 2013 Prelims 79 1
StarSeries V Prelims 146 1
The International 2012 West Qualifiers 35 4
The International 2012 East Qualifiers 34 3
GEST: March 2013 87 6
Dreamhack Invtl. 42 0
WePlay 122 1
Armageddon GS 16 0
G-1 2013 Asian Qualifiers 95 2
Netolic II: East 39 1
StarSeries V Finals 15 0
G-1 2013 Western Qualifiers 24 1
AMD Premier League 64 1
GEST: April 2013 19 0
EMS Season 1 Finals 16 0
G-1 2013 Finals 15 1
GEST: May 2013 18 0
Curse Invtl. 14 0
Netolic II: West 38 0

Seems like ‘5 or Less’ is the smart bet. Only two tournaments have had over 5 games last 60 or more minutes. Out of 2,691 games 60 minute plus games have occurred 64 times. The binomial distribution puts the odds of six or more 60 minute games happening in this tournament (33 to 50 games) occurring less than 1% of the time.

My Choice: 5 or Less

10.) Earliest Roshan Kill Time (1 Point) —  ‘Less Than 1 Minute’ versus ‘Over 1 Minute’ 

Early Rosh kills are rare, occurring 19 times in 2,118 games. Using that probability, the odds of one happening this tourney range from 27% to 36%, depending on how many games this tournament has.

My Choice: Over 1 Minute

11.) Earliest Tower Kill Time (1 Point) —  ‘Less Than 2 Minutes’ versus ‘Over 2 Minutes’ 

This is even more rare than early Rosh kills, happening only 9 times in 2,118 games. Using that probability, the odds of an early tower kill happening range from 15% to 19%, depending on how many games this tournament has.

My Choice: Over 2 Minutes

12.) Rapiers Purchased (1 Point) —  ‘2 or Less’ versus ‘3 or More’ 

My data on this is not completely accurate as it is difficult to distinguish if more than one Rapier was bought or if a Rapier was lost and someone else picked it up. I am going to assume that my data does not have any instances of games with more than one rapier in it.

There have been 109 games with a Rapier in 2,118 total games. Using the Binomial Distribution and those numbers the chance of ‘2 or Less’ Rapiers ranges from 76% to 52% depending on how many games this tournament has. Keep in mind though, that the Gyro Rapier combo is a valid strat on the competitive scene and may make a few appearances this tournament.

My Choice: 2 or Less

13.) Refreshers Purchased (1 Point) —  ‘3 or Less’ versus ‘4 or More’ 

Refresher was purchased 181 times in 2,118 games overall and 84 times in the last 768 games. Both numbers (181/2118 and 84/768) and the Binomial Distribution put ‘4 or More’ as the more likely occurrence. There is a 50% to 82% chance of ‘4 or More’ (since March 1st numbers) and 31% to 63% chance of occurring (overall numbers) depending on how many games the tournament has. I’m predicting Magnus as the most picked hero and a lot of Mag + Refresher combos.

My Choice: 4 or More

14.) Largest Single Crit Hit Damage (1 Point)

I’m not sure what this question is asking. Are they asking for the largest crit actual damage or largest crit number that appears on the screen? Are they only counting crit hits from items and passives or are they counting some of the hits that are labelled as crits in the replay parser but do not fall into one of those two categories? If anyone has any clarification on this please let me know.

Edit 5/11/13: see this post for my guess.

15.) Most Roshan Kills in a Single Game — ‘3 or Less’ versus ‘4 or More’

Roshan has been killed four or more times in 66 games out of a total of 2,118 games or 3.12% of the time. Using that percentage, the chance of ‘4 or More’ kills happening this tournament ranged from 65% to 80% depending on the number of games this tournament has. Here’s a breakdown by tournament:

Tournament 4+ Rosh Kill Games
ECAL Americas 0
The International 2012 7
StarSeries III Prelims 7
StarSeries III Finals 0
G-1 IV Prelims 2
G-1 IV Finals 0
Dreamhack 2012 Winter 6
GosuLeague IV 2
RaidCall I 1
GEST D2 Tourney 0
GEST D2: The Challenge 0
TPL III 3
GosuLeague V 0
The Defense III 6
RaidCall II 5
StarSeries IV Prelims 2
StarSeries IV Finals 1
ASUS FBotY 0
The Asia 0
G-League II 4
Killing Spree 0
SEA League 0
TPL IV 3
EMS Spring 2013 Prelims 2
StarSeries V Prelims 1
GEST: March 2013 5
Dreamhack Invtl. 1
WePlay 2
Armageddon GS 0
G-1 2013 Asian Qualifiers 2
Netolic II: East 0
StarSeries V Finals 1
G-1 2013 Western Qualifiers 1
AMD Premier League 0
GEST: April 2013 0
EMS Season 1 Finals 0
G-1 2013 Finals 0
GEST: May 2013 1
Curse Invtl. 1
Netolic II: West 0

My Choice: 4 or More