posted by Nahaz,
In the last days before TI4, joinDOTA statsman Nahaz once again turns to Martin Decoud of datdota.com and Noxville, Dota stats writer for 2p.com, to help you make your Compendium Predictions for the TI4 Group Phase!


In this edition of Stats Corner, I’ve enlisted help from Martin Decoud, project director at datdota.com, and Noxville, Dota stats writer for 2p.com, to help you make your Compendium predictions for the TI4 Playoffs. 'Hero Specific Predictions' on the first two pages are below. For 'Other Predictions', see the datdota.com blog. Noxville provides his 'Top Six Prediction' in an article on 2p.com. First of all, my apologies to them for the delay in posting this article - I did not predict that the first hurricane of the season would come through during my beach vacation!

We assume that these predictions cover the Group Phase games played on July 9-12 (aka Phase 2). We've compiled the data used to make these predictions in an Excel spreadsheet. As with any statistical approaches, our predictions implicitly assume that what will happen in the 120 Group Phase matches looks about like what has happened in the data.

Predicting single game events is much more difficult than predicting averages or totals across games. In a change-up to prior compendium questions, many of these questions are asking for ‘most average’ rather than the most over the total event or the most in a single game. The addition of ‘average’ changes things significantly. By looking at the ‘most average’ by tournament (see the spreadsheet) a few patterns emerge.


  • Generally you want to pick a hero that plays one and only one game.
  • The hero has some innate skill or play-style that makes him good at that category.
  • The hero needs to perform very well (or poorly) in that one game.


In other words, predicting the hero with the `Highest Average GPM' is, from a statistical standpoint, as or even slightly more difficult than predicting which hero will have the highest GPM in any single game. Feel free to play your hunches!



Hero Specific Predictions, Page 1



Most Picked Hero



datdota: Picking Mirana. She’s the most picked hero in 6.81b, the most picked in 6.81b games between two TI4 teams, and every team has picked her multiple times in 6.81b. Mirana fits into multiple roles and is flexible enough to be picked early. Doom could also take this one but appears to be banned too often. Batrider possibly if his Ban rate drops sharply or maybe Tide or Brew if their popularity continues to increase. Io and Shadow Shaman are very popular among some teams but not played at all by others.

noxville: Mirana and Doom have this awkward dynamic where one or the other starts trending as a ban in a tournament on one day, and then on the next day it’ll flip back over. Given the data it seems like Doom gets banned slightly more while Mirana gets picked more often, and with the huge Eastern-team respect for the hero, I feel it’ll be banned more leading the way for Mirana to win this. If there’s a ‘Treant Protector’ move of TI4, I’ll be upset.

Nahaz: It’s just too hard to justify going against the odds here. Mirana has appearaned in almost 50% of pro Dota 2 games in 6.81 (click on the 'Pick %' column to sort). No other hero is even close.


Most Banned Hero



datdota: Lycan. Lycan has passed Batrider as the most banned hero in 6.81b and is the most banned in games between two TI4 teams. Doom and Batrider are also worth considering.

noxville: Also going with Lycan. Although we saw more Batrider banned in the EU qualifiers, we’ve subsequently seen games where Batrider is just ignored completely in the draft.

Nahaz: This is certainly down to Lycan, Bat, and Doom. All three have been banned in over 60% of pro matches (sort by ban %) in 6.81b; only one other hero (Brewmaster) is over 50%. The decider for me is that Batrider has a sub-45% win rate since 6.81b, while Doom is at 55% and Lycan is over 60%, including a dominant 30-16 record on the Dire side.


Heroes Never Picked or Banned



datdota: I like the four intervals that cover the 10-21 range as possible choices. I took a list of all the heroes and put them into four groups: will be picked/banned, probably, maybe, and unlikely. By multiplying an arbitrary percentage by the size of each group, I got a rough indicator of how many heroes I felt would not be picked or banned: 15.8. I’m going to go with 13-15 over 16-18.

Nahaz: Tough call. In 424 pro matches on record during 6.81b, only 11 heroes have been picked or banned 5 times or fewer. Only five heroes (Necro, PL, Huskar, Pudge, and Ogre Magi) have zero bans. I’ll go with 16-18 here, but it’s very close.


Hero with the Most Average Kills



TI 2011 Correct Pick: Sniper (1 game played)
TI 2012 Correct Pick: Faceless Void (1 game played)
TI 2013 Correct Pick: Nightstalker (1 game played)

datdota: For the purposes of the ‘most average’ questions, I made a list of heroes that I think are likely to be picked only a few times (or none at all). Looking at the winner in this category in prior tournaments with over 100 games,10/24 of the winners appeared in only one game. In comparison, 17/24 ‘most average deaths’ winners appeared in exactly 1 game. That means that unlike the other questions there’s a bit more flexibility. The hero you select can appear in more than one game and still win this category (though fewer games is better). I considered Templar Assassin, Shadow Fiend, Clinkz and Axe but think they will be picked too often. I considered Zeus, Necrophos, and Sniper (he did it at TI1 in his only game) but they might not be picked at all. Going with Ursa.

Nahaz: For reasons stated above, these ‘most average’ questions are basically the same as single-game predictions, so there is a lot of uncertainty here. Tinker and Morphling are the only heroes to reach 30 kills in 6.81b but both are being picked up in about 13% of matches, which is too many. Void is the only hero to break 20 kills twice, but he’s being picked up far too frequently in the recent meta. Based on recent stats, Spectre strikes a decent balance (sort by kills) between averaging a large number of kills and being infrequently picked. I seriously considered Meepo for both this answer and the next, which should give you an idea of how much of a toss up I consider these questions.


Hero with the Most Average Deaths



TI 2011 Correct Pick: Pudge (1 game played)
TI 2012 Correct Pick: Axe (1 game played)
TI 2013 Correct Pick: Slark (2 games played)

datdota: Please see the notes on ‘most average kills’ above. This is incredibly difficult. Even more strongly than in the other ‘most average’ categories you want a hero that is only going to appear once. 17 out of the 24 tournaments with over 100 games had a winner in this category that only appeared once and all of them had a winner that appeared 3 or less times. I’m going to guess a hero that I don’t think is even going to appear at TI4. Crystal Maiden. I considered Pudge and Vengeful Spirit but think they will be picked too often. Ogre Magi, Abaddon, Sniper, Warlock, and Huskar, are also possible but might not get picked at all.

Nahaz: Four of the top 7 death totals in 6.81b (15+ deaths in a game) belong to Io/Wisp. As Martin’s analysis suggests, this is really a craps shoot: we’re trying to pick a hero who will only have 1 or 2 appearances and do really badly when he is picked. I’ll go out on a limb here and pick Meepo. With Era competing for FNATIC, I don’t think we’ll see a lot of this hero, and he tends to die a lot (6.4 deaths/game in 6.81b). I’d toss in Drow Ranger as an option as well in addition to the heroes Martin listed.


Hero with the Most Average Last Hits


TI 2011 Correct Pick: Spectre (8 games played)
TI 2012 Correct Pick: Faceless Void (1 game played)
TI 2013 Correct Pick: Anti-Mage (16 games played)

datdota: Though a carry picked multiple times can sometimes overcome this, it’s still easier to be picked less frequently. 9/24 previous winners were only picked once and 15/24 previous meetings were picked 3 or less times (in 24 tournaments with 100 more more games). The best carries at last hits like Anti-Mage, Shadow Fiend, Tiny, Morphling, and Luna are likely to be picked too often. Alchemist, Meepo, and Naga suffer from sometimes being run as supports. Medusa, Phantom Lancer, and Necrophos might not be picked at all. While it’s likely he’s picked too often to win this, I don’t see a good contender for the being picked only once group. Anti-Mage.

noxville: I think that Anti-mage is a reasonable pick, but for some reason I think that Gyrocopter is going to sneak this one in. He’s got a bunch of records for most last hits in games, and isn’t likely to get picked a lot.


Hero with the Most Average Assists



TI 2011 Correct Pick: Spectre (8 games played)
TI 2012 Correct Pick: Axe (1 game played)
TI 2013 Correct Pick: Sven (1 game played)

datdota: Split on this one. 17/24 heroes that did this in tournaments with 100 or more games were only picked once. I’d pick Venomancer but get the feeling he’s going to play in too many games (though his assist grabbing abilities may allow him to overcome being picked more than a few times). KotL, Magnus, Earthshaker, Elder Titan, Spectre, Undying, and Alchemist are also likely to be picked too often. Zeus, Necro, Tusk, and Cystal Maiden might not be picked at all. Going with Zeus, though once again I think it’s unlikely that he’s even picked.

Nahaz: Time for another toss of the dice. Undying actually has the highest assist average in 6.81b. Both FNATIC’s Fly and Aui_2000 have posted double digit assist totals on the hero in the last couple of weeks.



Hero Specific Predictions, Page 2



Hero with the Most First Bloods



datdota: This spreadsheet has the number of first bloods by hero since 6.80 and the hero’s first blood %. Using that first blood % and the likely number of games each hero will play Shadow Demon, Treant Protector, Invoker, Rubick, Enchantress, and Mirana are my top choices. I’m going to go with Mirana.

Nahaz: Looking at 6.81b stats, Mirana, Rubick, Shadow Demon, and Treant Protector seem to strike the best balance between getting picked up a lot and having a high first blood percentage. Bane is my sleeper pick here, but I’ll stick with Treant Protector.


Hero with the Most Tower Damage



datdota: This question appears to ask which hero will have the most total tower damage in all games across the event. This spreadsheet has the average tower damage and total tower damage by hero since 6.80. Unsurprisingly, Lycan comes out on top. While Lycan may be banned too much to take this category, I’m still going to go with Lycan. If Lycan can make it into 20% of games (being banned in the others), he’ll likely take the category unless some other carry emerges unexpectedly (like Invoker being picked in 40% of games or Morphling being picked in 25% of games). Honorable mentions to Invoker, Tiny, Nature’s Prophet, Morphling, Faceless Void, Mirana, and Shadow Shaman.

noxville: I think Lycan is going to get banned a lot, and not make it into many games. A core hero like Tiny or Pugna, or a support hero like Shadow Shaman could sneak this record, although Lycan is also very reasonable (he always has huge tower damage, so only needs to make it into a few games).

Nahaz: To me this is really a question of whether we’ll see more Lycan or Tiny-Wisp, or whether both get banned in which case this defaults to Mirana. Lycan has done the most total tower damage in 6.81b (and in only 85 appearances) while Mirana leads 6.81 overall simply by virtue of her massive lead in total times picked. I'll go with the recent data and take Lycan here. If you're worried about a repeat of ESL One (where Lycan was banned in 14/18 matches), go PotM.


Hero with the Highest Average GPM



TI 2011 Correct Pick: Sniper (1 game played)
TI 2012 Correct Pick: Faceless Void (1 game played)
TI 2013 Correct Pick: Anti-Mage (16 games played)

datdota: Unlike most of the other ‘average’ questions, this one actually has a hero who has consistently won this while actually being picked multiple times. It’s still more likely that a hero will be picked only once and win this, but I don’t have strong feelings about who that hero might be. Despite likely being picked too often, I’m going with Anti-Mage. Naga and Alchemist suffer from being potentially picked too often and being run as supports. Honorable mentions to Usra, Sven, Sniper, and Medusa.

Nahaz: AM, Shadow Fiend, and Meepo are the only heroes to average over 600 GPM in 6.81b. I’d favor AM over SF as I think we’ll see more of the latter at TI4. No Excalibur makes Meepo no longer an option, so Anti-Mage by default.


Hero with the Lowest Average GPM



TI 2011 Correct Pick: Dazzle (3 games played)
TI 2012 Correct Pick: Nyx Assassin (1 game played)
TI 2013 Correct Pick: Lion (4 games played)

datdota: Looking for a support here that will be picked at most once, farms like crap, and hopefully is on a team that gets crushed. Undying, Ogre Magi, Abaddon, and Spirit Breaker come to mind. Lich often has a crap GPM but I think he’ll get picked too often. Lina and Crystal Maiden are supports unlikely to get picked much or at all but have farming mechanisms. Going with Abaddon, though it’s possible he might not even make an appearance at TI4.

Nahaz: Ogre Magi has the lowest GPM in competitive Dota 2, just over 200, and only 85 all-time appearances. In Akke I trust?


Hero with the Most Wins



datdota: If you are torn between two or three heroes for most picked hero, you can hedge your bet and pick a different hero for most wins and most losses. Going with my most picked hero here. Mirana.

Nahaz: 2ez4PotM. See my notes on most picked above. Doom has gained on her recently but the data still say it's Mirana, easily.


Hero with the Most Losses



datdota: Mirana. It’s unlikely that the same hero has the most wins and losses unless that hero is picked substantially more than every other hero. However the alternative is to choose a hero that I think will be picked less often and have a crappy win rate which seems
like a difficult thing to predict. If Mirana is the most picked she’ll likely win one of these categories too.

noxville: I’m going to agree with Mirana, but also suggest Tidehunter as another very commonly picked hero that could also lose a reasonable amount.


Thanks for reading, and good luck! Check back in a few days for our main event precitions, and be sure to follow @NahazDota, @NoxvilleZA, and @datDota on Twitter for updates!

Comments