Resources / Esports News
Aug 6, 2019

Looking for what to predict for your TI9 Compendiums? Noxville is here with the stats to help you out!

With The Summit 10 (and Bot TI 2 obviously) ending so recently, it seems like an appropriate time to do predictions for The International. I wanted to delay this a bit so I could see the 7.22f Immortal pub meta in case there was some interesting trends (since Alliance is the only team to play at The Summit who is also scheduled to play at TI).

TI9 Compendium Tournament Predictions

Many of the predictions rely on a significant tournament-wide number: how many games will be played at TI9, so we’ll start with the Tournament Predictions. These are actually the only predictions which are very reliable, the rest become crapshoots simply because the number of possible answers increases so rapidly (and all can be offset by just a single game), so the analysis will be shorter for them.

There are 3 categories of matches: group stage (fixed at 112 games), tiebreakers (could be a large number, but is likely 0-6 with the most likely being 0-1), and playoffs (4 bo1 + 17 bo3 + 1 bo5). Epicenter averaged 2.35 per bo3 and MDL averaged 2.53, so assuming ~2.41 is a reasonable average this puts it at 48-50 games depending on the finals - an annoying number given the prediction categories. I feel that this season has seen some clear stratification in skills of teams, and thus there’ll be slightly fewer matches than average - so I’m going with 45-49 for the main event. This also means a total of ~ 160 games in the entire event.

So far in 7.22 we’ve seen every hero picked, and every hero except Vengeful Spirit receive a ban. The patch is a sample of 662 drafts however, but even reducing to just Premium matches the story is similar. For 101+ not to be correct it would mean that the set of heroes including Clinkz, Slardar, Riki, Tinker, Viper, Drow would not have any element getting picked even once - something I think unlikely. Bans have historically been just a slight less bit behind (fewer teams playing in TI than in a normal sample of ~160 games), so this seems like a 91-100 / 101+ split. I’ll be a bit conservative here and go 91-100, but both are reasonable.

compendium hero predictions

Looking at all professional vs just premium matches shows that are mega-blood baths are truly rare: just 57 occurrences in 4255 pro games since TI8 ended have had > 90 kills (i.e. 1.34%) and just 19 have had > 100 kills (0.45%). Given that the event is ~160 games, a per-game roll of 0.45% translates into ~51.1% chance to occur at least once, and a 1.34% per-game roll translates into 88.4%. I’m going to veer towards the slightly longer, bloodier games and go with 101-110. 

ti9 compendium game length

Just one game this patch (of 665 games) has gone longer than 90 minutes (Newbee vs EHOME in the TI9 Qualifiers), but as I said above I expect some slightly longer games than average - it’s just a question of how much above average we’re talking. As seen in the picture above, 71-80 seems the best category, and even a moderate shift up would just possibly bring in 81-90, which I will go for. 

In terms of serious competitive matches there’ve been only two this patch shorter than 15 minutes - Chaos Esports Club vs six eight two (TI9 qualifiers), and OG vs Forward Gaming (ESL Birmingham). With more up for grabs, and closer team skills here than other events, I think that such a short game is less likely and thus 15:00 - 19:59 is much more safe.

TI9 Compendium predictions

TI9 Compendium Hero Predictions

For Hero Predictions, I used a spreadsheet for drafting which incorporated a mixture of 7% Immortal matches from this week, 30% for 7.22[a-d] matches, 50% for 7.22[d-f] matches and 3% from 7.21 matches involving any of the 18 teams. This is because I feel like drafters sometimes fall back on good strategies and ideas from their past - especially in ever-changing metagames that we see develop in longer events like TI. This suggests Grimstroke and Chen as likely most picked/banned, although Crystal Maiden seems like a good alternative pick, and Io a good alternative ban. This spreadsheet defined the set of heroes from which to evaluate highest/lowest averages, so those are simply below.

TI9 compendium hero predictions

TI9 Compendium Team Predictions

I think it’s accepted that Team Secret are the current favourites to win TI9. Although it’s obviously way more likely that they don’t win than they do, they are still the safest bet here in my eyes. That said, the team advancing through the winner’s bracket plays fewer games than one which drops down and plays a bit in the losers bracket. As a result, I think the other strong teams such as LGD, Liquid, Vici & EG also have a good chance of playing a lot of games (which affects most of the categories below). Both Alliance and TNC have a habit of playing longer games which they can’t close out easily, or making insane comebacks - but I think TNC will advance further and grind it out.

TI9 compendium team predictions

TI9 Compendium Player Predictions

TI9 compendium player predictions

Agree with his predictions? They could also be a good guide for making your first bet here at Rivalry! Your first one is free, so what are you waiting for? Let's make the games a little more interesting.

0 Bet Slip