Tournament predictions are always interesting to analyze. We will be using the previous TI, the last three Majors and The Summit 9 as our reference, however keep in mind that the best we can do is give you an educated guess, so feel free to change things up if you disagree.
China Supermajor had 56 games in its main event, but it also had bo3 instead of bo1 in the lower bracket first round.
Considering that bo3 adds on average 1.5 games, compared to bo1, we can subtract 6 games from the China Supermajor for our best guess, leaving us with 50 games. Hence, for this category, we are putting down 50 to 54 games as our prediction.
There were 102 unique picks at the China Supermajor, 86 at the Summit 9 and 112 across qualifiers, meaning that the meta is actually quite diverse, despite most teams defaulting to lane-domination playstyle.
Given how short The Summit was and the fact that it still had 86 unique picks, we can safely assume that there are going to be at least 91+ unique picks at the International. The question to answer then is whether the TI meta is going to be at least as diverse as it was in June.
Given how there are now teams like Fnatic, experimenting with different playstyles and teams that didn’t play too much internationally, like iG, we expect the group stage to be even more diverse than previously and it alone will probably give us 101+ unique heroes.
Looking at stats for TI8 qualifiers, TI7, TI6 and China Supermajor, on average professional players ban roughly 86% of the total hero pool of the tournament. It means that if we expect 101+ unique hero picks, we expect at least 86+ unique bans.
Depending on how diverse you expect meta to be, both 81-90 and 91-100 are solid choices, however in recent tournaments, namely ESL One Birmingham, MDL Changsha, China Supermajor and The Summit 9, the percentage of banned heroes was slightly lower, closer to 80% and that is despite having an extra ban in the first phase, compared to previous TIs.
As such, we are going with the 81-90 bracket for this prediction. It fits in both historical and recent ban percentage for a hero pool of less than 105 contested heroes and if looking at recent data only, covers bans for hero pools of up to 111 heroes.
There is always that one game that goes on for a very long time and has a ton of kills, but historically, occasions where the combined kill score would exceed 110 are extremely rare, and are virtually non-existent in high-profile tournaments, such as the International.
There were no games with 110+ kills in TI7, China Supermajor, The Summit 9 or ESL One Birmingham, so we think there won’t be such a game in TI8 either. Even in the recent Fnatic vs. Pain Gaming match the kill score only reached 105 kills, hence going for 101-110 combined kills is probably the safest bet.
Last year there was a 2 hour game at the International, so it isn’t unheard of for a game to go over a 100 minute mark. But the game is different now.
Even when looking for an outlier, we have to take into account that games are shorter in the current meta. There were no 90+ minute games in China Supermajor, no 80+ minute games during the Summit 9, no 70+ minute games in MDL Changsha and even no 60+ minute games in ESL One Birmingham.
TI8 will have almost double the amount of games compared to the Supermajor, so it is for you to decide whether you believe the amount of tries will allow for an extreme outlier. We don’t, so we will go for the 90:00 to 99:59 prediction bracket.
Across the three last Majors, previous TI and the Summit 9, there is always at least one game that doesn’t reach the 20 minute mark, but there are no games that end before 15.
15:00-19:59 prediction is the most rational choice, according to stats and we will go with it.
Professional matches rarely have too many kills—pro-players are better at ending the games and won’t risk giving the enemy a chance of comeback to increase their stats.
We still managed to find several recent high profile games that had a hero with 26+ kills: Vici Gaming vs. Mineski had 30 kills on 哎哟’s Storm Spirit, Vici Gaming vs. Evil Geniuses had 26 kills on XG.Paparazi灬's Gyrocopter and TNC Predator vs. VGJ.Storm had 28 kills on Aurora.Armel's Mirana. This is more than enough to convince us that in a tournament with roughly twice the amount of games, there is going to be a at least one game with 26+ kills on a single hero.
Not a single instance of 21+ deaths with several exact 20-death games lead us to believe that the 18-20 bracket is the most optimal choice.
Moreover, of all TI paticipants, pieliedie has the highest death average and his record is 19 deaths in a game vs. Pain Gaming.
There were several 36+ assist games in our sample, so we are confident that 36+ is going to be the correct answer.
Once again, it is a question of whether you believe Alchemist or Anti-Mage will have at least one very good game. Alchemist is one of the least popular heroes in the current meta, but we believe new OG with Ana will make him work at least once, so we bet on GPM exceeding 1000 in at least one game.
Predictions are never guaranteed to be correct. Even with all our data, the best we can realistically do is give you several options and be 30% sure one of them is going to be correct. It is not exact science, but it still gives you much better odds, compared to predicting randomly.
The same rules apply to real life betting as well. Buff88.com can give you a chance to test your Dota knowledge and if you can keep your cool and bet smart, you will surely prevail.
I think all predictions have just a little bit too high numbers
except assists and shorter game.
Hero predictions where ?
Hero predictions soon)
@stars so you think 15-20min shortest game is too high? Are you high?
in the last TI compendium, if I am not mistaken, for tournament predictions they only considered games in the main event. is this true this time as well?
I think 36+ assist is logical, since there will be several game with more than 70+ minutes. Longer games means more teamfight, thus more assist. I am still hesitant on longest game and heroes ban tho
Is your summer boring? Do you want to make it interesting by loose all your money? Or would you just taste how does it feel to make us a Winner?
Go and make your bets on Buff88.com, and make YOUR summer into something exciting! Buff88.com - with the mindset of a lame gambler.
There's an error: "bo3 adds on average 1.5 games" This is not true. If every matchup were exactly even teams (with 50-50 chance of winning), then 1.5 would be correct. But of course in every matchup there is a team which statistically would win more matchups, shifting the average games in each bo3 BELOW 2.5. Doing the math above would shift the number even higher than 50, but China Super Major was a statistical anomaly. I didn't do all the math now, but considering how competitive TI is, I guess around 2.3 or 2.4 per bo3, so a better prediction of the total would be 45-49.
Just looking at the last TIs: 2017=47, 2016=47, 2015=48.
I'm guessing the above statistical error (winrate <> 50%) is why TobiWan, Merlini, BananaSlamJamma, Synderen, and others all got it wrong.