Hi. You might know me as that one scrub who writes fantasy guides on Reddit sometimes. Well, in the comments to some of the guides I've done, I got a lot of requests to do guides on Dota 2 Compendium Predictions. I've always said no to those requests because I never perceived it as my strong suit and I didn't want to go out of my depth leading people astray on stuff I didn't understand. But I thought about it some more, and realized...most of what I've done for Fantasy Dota has been gathering and reporting data...why couldn't I just do that for something else? I didn't really arrive at a great reason not to try it, so I'm trying it. This is PART ONE, where I cover the Heroes section and the Tournament section (I'll cover Teams and Players, tabs two and three on your predictions pages, in Part 2).

Also, the usual disclaimer applies. The thoughts and conclusions in this piece are my own. I encourage you to read it and take mental notes, but I also encourage you to listen to what others say too. I promise your Merlinis and Purges of the world are going to approach this stuff a lot differently than I will. Gather as much information as you possibly can, and then make your own decisions.

And for those interested, I used Liquipedia, DotaBuff and Open Dota as my primary sources of information to pull from when doing each of these. I had no preconceived notions when I went in, I just filled out my predictions as I filled this out.

Anyway, let's get started.

In TI8 Main Qualifiers, the most picked five heroes were Windranger (101), Warlock (88), Skywrath Mage (74), Beastmaster (72) and Naga Siren (68). Earlier in the China Supermajor, that list was Beastmaster (42), Bane (38), Sand King (37), Death Prophet (37) and Disruptor (36). So the first thing I recognize is Beastmaster is the only dude who appears on both lists. Every one of those Top 5 picked heroes from qualifiers took nerfs in the latest 7.18 patch, so I'm just going to roll with the one who's been there consistently and hope for the best.

Please. You don't need my help with this one. Wisp. Wisp was the most banned hero in every single region during TI8 main qualifiers. Come on.

I'm going to start off by saying this pick is a total crap-shoot because of the minimum only being 5 games. There's a lot of good, logical heroes out there to pick based on what's hot and effective right now. But couldn't you just see a random hero like Nyx Assassin or somebody appearing in like 6 games, randomly winning 5 of them, and winning this category with an 83% win-rate? Guessing this right should be worth 5000 points, not the 800 it's currently listed at. Pro players will probably tell you qualitative things like "Pick a cheese hero that gets last picked that can't be countered like Brood or Huskar". But that seems too obvious, so it CAN'T actually be that, right? So I'm going about it a different way. Once again, I'm going to compare China Supermajor with Main qualifiers.

Heroes from the Supermajor,70%+ winrates, 5+ games played: Chen (100%), Faceless Void (88%), Shadow Fiend (86%), Wisp (83%), Zeus (75%), Bloodseeker (72%), Tinker (71%), Omniknight (70%), Visage (70%)

Heroes from TI8 Main Qualifiers, 70%+ winrates, 5+ games played: Abaddon (100%), Enigma (86%), Juggernaut (83%), Shadow Shaman (80%), Chen (75%), Bristleback (71%), Huskar (70%)

Then I went to look at the 7.18 patch to try and find any heroes on either of those lists who got buffed significantly. Here's who I found: Enigma. That's it, that's the end of the list (unless you consider Abaddon's 5 extra move speed to be a game-changer).

Again, this particular category is really an educated stab in the dark, so your stab in the dark is as good as mine. As far as my compendium is concerned though, in Universe we trust.

Looked at the China Super Major at the heroes who averaged the most kills (and also played at least five games). The highest four: Shadow Fiend, Storm Spirit, Morphling and Bloodseeker. I then looked at individual outliers at both the Super Major and TI8 Qualifiers for kills (games in which said-hero logged 20+). Among those four heroes, the first three managed it just once. BS did it twice. I'm going with BS.

I'd like to acknowledge that the "conventional wisdom" pick here is Zeus because of his ult. I would then like to acknowledge that I'm here for looking at data, not conventional wisdom. If you want conventional wisdom, you're reading the wrong guide (by the way, you SHOULD look at other guides besides mine when making your picks, don't just take my word for it). Here's what the numbers showed me...

For this one, I made a list of all the heroes who averaged 17+ assists in both the Supermajor and Qualifiers. That list goes as follows: Wraith King (19.6, 20.0), Silencer (19.5, 17.0) and Spirit Breaker (19.0, 22.0) (Bounty Hunter was close, but no cigar). The consideration that needs to then be made is will said-hero reach 5 games played at TI? Wraith King and Spirit Breaker were the two strongest...can they get to FIVE games though? Personally, I THINK Wraith King can get there. I'm going with him.

Trying to sift through Open Dota for low-death games would be a colossal pain in the butt (it's not immediately obvious what the nature of a low-death game is), so I didn't bother with outliers for this. All I did was look at death averages per game for heroes who played at least five games in the China Supermajor. The lowest were, in order, Broodmother, Medusa, and Lifestealer. None of them got affected by 7.18, so I'm just grabbing the one whose name came up first. If you're leaning more toward one of the other two? You do you.

At the China Super Major, one hero played in plenty of games and far, far exceeded everybody else in last hit average (he was at 556, while next closest hero was Storm Spirit at 464). This one should be easy.

Three heroes averaged over 700 in the XPM category at the China Supermajor: Juggernaut, Timbersaw and Zeus. Jugg being there weirds me out because he doesn't have an XP talent like the other two. For me, I lean toward Timber over Zeus because I feel like Zeus can get crapped on in lane which can screw up his game. It's hard to crap on Timber though, it seems to me like he kind of just does what he does regardless of what enemies have to say about it. Plus, I think the hero will appear in a lower number of games than Zeus (but still over the 5 minimum threshold), which minimizes my chances at their being some ultra-poor performance outlier game to prove me wrong.

The highest kill count on the pro scene in the last couple of months was when Moonn got 30 kills on Storm Spirit at the Supermajor. But there hasn't been another Storm game like that in that span. Meanwhile, what if I told you that, in the Main Qualifiers alone, there were FIVE instances of Mirana getting 18 kills or more? I don't know if she's going to be THE hero that scores THE most points in the one game that yields the most kills...but she should give you lots of chances at the very least because she SHOULD have lots of high-kill performances.

You can make a case for lots of heroes here, but I think all I need to say is this: in the Main Qualifiers, three of the Top 6 last-hitting performances submitted (I will remind you there were 241 matches in main qualifiers, so this is a legit sample size) were by players using Templar Assassin. Sold.

I'll just lay out the facts for this and let you decide. The TI main event formula consists of 4 BO1s, 17 BO3s, and 1 BO5. That means the MINIMUM number of games played is 41, and the MAXIMUM number of games played is 60. The probability is that it gets split somewhere down the middle at 50.5. Luckily, we have an option that spans that choice in the 50 to 54 , and that's my choice. If you think there's going to be a lot of 2-match sweeps this year, I don't have a problem with you picking 45-49 (last year's correct answer was 47 if you're looking for historical precedent to do that...but last year was strange in the sense that only six of the 18 series didn't result in sweeps). Basically, if you pick one of the options in the middle, you've got a good shot. You do you.

Per Liquipedia, 112 of the 115 heroes in Dota 2 were picked over the course of the 241 matches that made up the TI8 Main Qualifiers. At the China Supermajor, 100 of the 116 heroes were picked over the course of 103 matches. TI8 will last nearly (if not, more than) 200 matches. And this compendium prediction tops out at 101+. Easy.

This one is a little bit tough. My methodology might be slightly controversial, but here's what I did. I ignored South America qualifiers because A.) The South American meta factors into TI8 the least of the six regions, B.) Doing so drops the number of matches accounted for from 241 to around 200, making it a more similar amount of games to what TI8 will be, and C.) South American qualifiers didn't come up in google immediately on DotaBuff like the other five regions did, and I was feeling slightly lazy. Anyway, I made a tally of all the heroes banned in the five regions I looked at, and arrived at a total of 98. The choices are either 91-100 or 101+ for heroes banned. It's gonna be close.

Finding pertinent data for this was a challenge: Open Dota is a great resource for tracking INDIVIDUAL performances, but team performances aren't its forte, and it certainly isn't great for finding what two teams combined to do. So the way I went and approached it is, in Open Dota, I went to the China Supermajor and had it sort by Duration (it may be a false premise, but I'm operating under the assumption that there are likely to be more kills the longer a match progresses). I set my threshold at 60 minutes, and checked the match IDs of each of the _ matches that went longer than that (normally I throw out anything from the qualifiers in April, but I made an exception this time to maximize the amount of data I was receiving). The totals I got back mostly ranged from the mid 70s  to the low 90s, the one exception being this 105-kill 88-minute bad boy here.

So what does that tell us? Spoiler alert, but in the next section, I'm going to make the claim that we're going to see longer matches than 88 minutes at TI8 (the longest one at TI7, for reference, was 128 minutes, but then again there were only 100 kills in that match). I think we'll see something crazier than 105 kills (the craziest stuff usually happens at TI, right?), but not TOO much crazier.

At the China Supermajor, there were two games in the 80-minute range which stood as the longest of that tournament. But here's a funny bit of data for you. At TI7, there were three games that eclipsed the 100-minute threshold, and not only that, there were ZERO that fell in the 90-minute range. Weird, right? Anyway, my feeling is that, having watched a few TIs, I think EVERY TI has that one or two games that just go for ultra marathon sessions, and there's no avoiding it. TI8 will span nearly 200 games, which is a huge sample size. I feel like a 100+ minute game is bound to happen.

At the China Supermajor, there was just one game that went under 20 minutes, an 18:11 game between Liquid and Spirit. At Birmingham, the shortest two games were in the 19-minute range. And the very shortest match at TI7 was a little over 15 minutes. LUCKY FOR US, 15:00 to 19:59 is an option! That's really convenient!

26+ is the way to go, because we saw it happen three times in the China Supermajor. Paparazi stacked up 26 as Gyro, Armel pulled off 28 as Mirana, and Moonn picked up 30 on Storm (in a loss!). Thank you and please drive through.

In my opinion, the only two viable picks here are 15 to 17 or 18 to 20. Allow me to demonstrate why. At TI7, we had seven instances of a hero dying 15 times, two instances of heroes dying 16 times in a game (JerAx did it on Nyx, Yao did it on AA), we had one instance of a hero dying 17 times (Accel did it on Disruptor), and then we had friggin' N0tail die 20 times on Pugna. At the much more recent China Super Major, there were 18 instances of somebody dying in the 15-17 range, while on just two occasions, somebody died in the 18-20 range (Jabz died 18 times on Leshrac, Fly 19 times on Venge). With such a massive amount of examples in the 15 to 17 range, anything below that is foolish. Between China and TI7, we have three examples of somebody falling in to the 18 to 20 range. Nobody in the above sample has died 21+ times...it's possible, but it just hasn't happened.

Valve being a little lazy makes this one a simple pick. The highest range available to pick is 36+, and at the China Supermajor, there were 3 games in which heroes finished games with 36+ assists (Tims and kuku had 37 and 36 respectively in one game, eLeVeN and LanM had 39 and 38 respectively in another game, and YapzOr had friggin' 45 on Elder Titan in a game). Take the over with confidence.

You can go one of two ways by this. If you think SOMEBODY will play a Greevil's Greed Alchemist competently, take 1000+. SumaiL was the only one to do it in Qualifiers, and he got 1061 GPM, which is a super absurd amount of coin. So if you think THAT will happen again, there's your answer. If you think that particular thing WON'T happen, keep reading.

At the China Supermajor, there were 6 examples of heroes finishing in the 800s in this category, but just one (Paparazi playing Monkey King) of somebody surpassing 900 (he got to 919 GPM, btw). Let's get some more data. At Birmingham, there were just four examples of 800, and none at 900. That's pretty much where we're at. If you choose 800 GPM, that's a very conservative estimate, but it's possible. Where I come down on it is this: there's going to be many more games played at TI8 than Shanghai and Birmingham combined. TI8's group stage will span 144 games at least (there may be some tiebreakers, too), and you're looking at another ~50 from the main event. In those nearly 200 games, you're telling me NObody is going to reach Paparazi's 900 GPM threshold? I think it'll happen

And that's it for Part One, keep an eye out for Part Two when I come back and cover Team and Player predictions. Remember, I'm not telling you what to do, I'm just showing you how I think and giving you the information to decide for yourselves. Happy Gambling.