Group stage · August 13
We model how the Swiss stage ends: team ratings over a year of matches, recent form, results on patch 7.41, market odds and predictions from well-known analysts. The model fills the bracket for you — edit it freely and watch the expected points move.
Your bracket
Click a team to swap it with another one. Under every card is the probability of that exact slot for that team. The source switcher recalculates the model in place: as long as you have not edited the bracket by hand, it follows along.
Undefeated
One loss
Five out of the elimination round
Five losers of the round
One win
Winless
This is the model bracket with the highest expected points
Team odds
The share of simulations where the team finished the group stage exactly like that. Each row sums to one, each column to the size of its slot: one team undefeated, two with a single loss, five advancing.
TEAM VISION rating 1723 ± 33 5–1 
Team Yandex rating 1698 ± 35 5–1 
BoomBoys rating 1681 ± 31 4–2 
Team Falcons rating 1687 ± 33 4–2 
Team Spirit rating 1662 ± 32 4–2 
Aurora Gaming rating 1666 ± 31 3–3 
Team Liquid rating 1640 ± 31 3–3 
Vici Gaming rating 1639 ± 38 4–2 
Iron Wing rating 1627 ± 31 2–4 
Team Resilience rating 1643 ± 71 5–1 
Xtreme Gaming rating 1619 ± 34 1–5 
LGD Gaming rating 1597 ± 33 2–4 
Nigma Galaxy rating 1581 ± 37 1–5 
OG rating 1551 ± 38 2–4 
GamerLegion rating 1541 ± 39 1–5 
HULIGANI rating 1518 ± 45 1–5 Opinions
Public predictions from players, casters and analysts. Dashed logos are slots the author never named: they are restored from what is left of the bracket or filled in by the model, otherwise an incomplete set cannot be compared by points with a full one.
Professional player
August 1
Professional player
August 2
Professional player
August 4
Dota 2 writerHotspawn
August 1
Team Vision have been the most consistent team on this patch by a considerable margin. Six of GamerLegion seven qualifier games lasted under thirty minutes.
Streamer
August 3
PARIVISION goes through the group stage without a loss. paraphrased from the source
Streamer, analyst
July 31
BetBoom Team goes through undefeated. Falcons and PARIVISION advance at 4-1. paraphrased from the source
Professional player
August 1
Team Spirit analystTeam Spirit
August 1
Professional playerTwo-time International champion
August 1
Editor-in-chiefHotspawn
August 1
Streamer
August 2
The Satanic and No[o]ne duo looks untouchable on LAN right now. paraphrased from the source
No exact slot: OG, Nigma Galaxy — out in the group
Streamer
August 1
Top three of the Int — BB, Spirit, Iron Wing. Out of the group — Yandex and Falcons.
No exact slot: BoomBoys, Team Spirit, Iron Wing — direct to playoffs; Team Yandex, Team Falcons — out in the group
Predictions are collected from public materials and from bracket screenshots filled in by the authors inside the Dota 2 client. The source is listed on every card. Where an author did not fill the whole bracket, the missing slots are dashed — first restored from what the bracket leaves, then filled in by the model.
Sources
Five sources of team strength plus a separate weight for analyst opinion. Their shares are tuned by the switcher next to the bracket, where the result is visible right away.
4,746
maps from August 3, 2025 to August 3, 2026. A game loses half its weight every 120 days, and every 40 for the form rating. Source — OpenDota.
43,220 $
turnover on the “The International 2026: Winner” market. Prices are converted into strength by regression, agreement with the rating is R² 0.86. The 28% spread is removed. Polymarket
+30
Elo added to map strength for Chinese teams: the event is in Shanghai — their own time zone, no flight across half the world and the arena behind them. Everyone else travels far. The weight is adjustable with the “Home LAN” slider.
March 24
A separate rating counts only matches played after the patch, while “hero meta” compares the team hero pool with how those heroes perform in pro play.
Method
Across 4,746 maps of the season we solve a maximum likelihood problem: every team gets the rating that makes its actual results most probable. A game loses weight with age and depends on the tier of the event — qualifiers count half of what majors do.
Rosters beat name plates: a team that played the season under a different brand is merged with the current roster by matching players. Otherwise half the field would have a history one month long.
The rating gap gives the chance to win a map, and the best-of-three follows from it. A favourite is stronger over a series than over a single map: three maps smooth out the variance.
Each of the 20,000 runs plays out by Valve rules: five rounds, pairings inside equal-record groups by Buchholz, leaving the group at four wins or four losses, then the elimination round between the 3-2 and 2-3 fives. The share of runs is the probability of a slot.
Public predictions are merged into one matrix, balanced to the slot sizes and mixed into the model with an adjustable weight. Loose calls such as “they go out in the group” are spread across the fitting slots instead of being pinned to one cell.
There are over two hundred billion possible brackets, so the set is assembled greedily by probability and then improved by swapping teams while the expected points on the Valve table keep growing.
Correct slots out of sixteen → points. The playoff prediction opens after the group stage and is scored separately.
The forecast is for reference. The model knows nothing about last-minute stand-ins, visa trouble or team morale — only played matches, market odds and published opinions. battlepass.ru is not affiliated with Valve Corporation.
Мы используем файлы cookie для улучшения вашего опыта на нашем сайте.
Продолжая использовать сайт, вы соглашаетесь с
Политикой использования файлов cookie в BP FLOWENGINE.
