Eight games settled the Summit.
Learn why a match point needs thirteen.

Overview

What the Summit was

The first LAN of the 2026 competitive year and the marquee event of the season’s first Major: Battle Royale duos, no respawns, a 75-duo field cut to a 50-duo Grand Final over three days in Düsseldorf. The final was eight games of cumulative points — play all eight, add them up, highest total is champion. No match point, no conversion requirement, no sudden death. That matters for the rest of this page: the Summit is the control a match-point final gets measured against.

Venue, dates, field size and schedule are Osirion session timestamps and the event’s published format; the prize figures are the event’s published prize breakdown. We compiled them into an internal stats pack (draft, 2026-08-22) and are repeating them from there — they are third-party facts we have not re-derived, which is why they are flagged as sourced rather than measured in Method. Everything counted below this line is recomputed from the leaderboard capture itself.

Where the wins went

Eight Victory Royales, 7 different duos, nobody with more than two. Four of the 7 finished outside the top ten.

A spread that even is the whole reason a match-point model can assume no duo has a lock on winning a game. The working is in Match Point.

How it was watched

Broadcast figures are Esports Charts’ (escharts.com), event page pulled 2026-08-22, quoted here with attribution and not independently verified. For scale: the FNCS 2025 Global Championship peaked at 954,441. Against the 2026 events we have a figure for, the Summit is the largest Fortnite audience of the year; we have not surveyed every 2026 broadcast, so read that as the biggest we can show, not a record.

Twelve weeks later, the same duo won the World Cup leg

Vico & Malibuca — BIG vic0 + BIG Malibuca 7 in the EWC replay parse — won the Reload Elite Series Championship in Paris that August, on a different mode, a different scoring system and a match-point format. Per Insider Gaming that makes them the first team in Fortnite history to win back-to-back major LANs; we are repeating that claim with attribution, not verifying it.

Comparison rows compiled in our internal stats pack from Esports Charts event pages and the two events’ published formats and prize breakdowns. Summit elimination rates recomputed here from the leaderboard: 788 ÷ 8 games ÷ 100 players. The EWC room is here: EWC 2026 →

Standings

The complete Finals Day 3 board, all 50 duos, as captured on 2026-06-11. Points and eliminations are cumulative over the 8 games. “Pts/game” is the only derived column: points ÷ 8.

The top four are a different event from the other forty-six. 4 duos finished on or above 350 points — the number the EWC would later use as its match-point line — and fifth place is 29 points off fourth. From tenth (270) the board roughly halves by thirtieth (124) and collapses to 4 at fiftieth. The gold bars are the 7 duos that took a Victory Royale; four of them sit in the flat part of the tail.

Reading the board

Three point ties sit on the board (274, 210 and 68, each shared by two duos). The published leaderboard resolves them in the order shown; the tiebreak rule itself is not recorded in our capture, so we do not state one.

Match Point

A cumulative final ends when the games run out. A match-point final ends when somebody wins a game after crossing a points line, and that is a completely different piece of arithmetic. This is the model that used the Summit board to work out how long that takes, what it predicted for the EWC Grand Final in Paris, and — now that the real final has been decoded — exactly which part of it was wrong.

1 · The rule, in one paragraph

Under match point you play a normal points final until you cross a threshold — 350 points at the EWC, in a 20-duo lobby. Crossing it puts you “at match point”. From that moment your points stop mattering: the only thing that ends the event is you winning a game outright. Everybody else keeps scoring, keeps crossing the line and keeps joining you. The event does not end when someone is clearly the best team. It ends when one of the qualified duos happens to take a Victory Royale.

2 · Why the Summit is the calibration set

To model a match-point final you need one number above all others: how likely is any given duo to win any given game? If one team wins half the games, a match point converts almost immediately. If the wins scatter, it drags. The Summit final is the cleanest recent read on that question — a full LAN Grand Final, the best duos alive, a complete win record, and no match-point rule distorting how anyone played for the win.

A model with no favouritism — every duo equally likely to win each game — produces 6.74 different winners over 8 games. The Summit produced 7. The assumption survives contact with the real board, so no skew was added. Worth saying plainly: this model does not believe peterbot is more likely to win a game than anybody else in the lobby, because eight games of the best duos in the world produced no evidence that he is.

3 · The engine: a coin only part of the lobby is allowed to flip

Once k duos are at match point in a 20-duo lobby, the chance that this game ends the event is simply k out of 20 — the chance the Victory Royale lands on somebody qualified to convert. Flip that over and you have the expected wait: 20 ÷ k more games. One duo at the line means expecting to sit through twenty more games. That is the whole reason these finals run long, and none of it requires anybody to be playing scared.

Straight from match-point-model.json. The relationship is exact: chance = k ÷ 20, expected wait = 20 ÷ k. It is a hyperbola, which is why the cure for a dragging final is not patience — it is more duos reaching the line.

4 · What the model predicted

200,000 simulated finals on those assumptions, run on 2026-08-21, the day before the Paris final:

Mean 8.7 games. Chance of running past 10: 10.5%. Past 15: 0%. The simulation never hit its own game cap.

5 · What actually happened — and where the model broke

Paris ran 13 games against a median call of 9. That is past the 90th percentile and four games longer than the call. The interesting question is not that it missed, but which assumption missed — and the decoded final answers it, because we can rebuild every duo’s running total from the replay parse and watch the line fill up in real time.

The real final, game by game

Rebuilt from data/analysis/ewc/Day4.json — every duo’s per-game placement and eliminations from the replay parse — scored with the fitted model in data/analysis/ewc-farming.json (3 points an elimination, placement 60/50/45/40/35/30 then −3 a rank to fifteenth). That model reproduces 19 of the 20 posted Day 4 totals exactly; the twentieth is the champion, whose posted total carries a +1,000 championship marker on top of the 499 points they actually scored.

The model’s published sample run, for comparison

From match-point-model.json → sampleFinal. This is one representative run of the simulation, not the real event — the file was written on 2026-08-21, before the final was played. It is here because comparing it against the real trajectory is what exposes the faulty assumption.

6 · The model already told you what slow pace does

The sensitivity table shipped with the model, and it is the part that turns out to matter most. Pace is how fast the lobby banks points. It decides how quickly duos reach the line and therefore how fast k grows — and k is the only thing that ends events. A cagey lobby is not just slower per game; it keeps k small for longer.

Pace 1.00× is the calibrated baseline. Rows from match-point-model.json → sensitivity.

Scoring 30% slower stretches the median from 9 games to 12 and the worst case from 14 to 18. Scoring 30% faster pulls the median down to only 7. That asymmetry is the point: slowing the lobby costs 3 median games, speeding it up saves 2. Match point punishes a passive lobby much harder than it rewards an aggressive one. By the two anchors in section 5, Paris played at roughly 0.65–0.75× the assumed pace — effectively the 0.70× row, whose median is 12 games and whose 90th percentile is 14. It took 13. Fed the pace the lobby actually played at, the model was close all along.

7 · Where this model is still weak

Stated plainly, because a model you cannot argue with is not worth reading.

Method

Every number on this page and the file it came out of. Nothing here is inferred from Summit footage, because no Summit replays have been decoded.

Measured, derived, modelled, missing

Two numbers that are ours, not the files’

Only two figures on this page are arithmetic we performed on top of the sources, and both are labelled where they appear. First, the 42% chance that games 9, 10, 11, 12 all went to unqualified duos: the product of (1 − k ÷ 20) over those four games, using the model’s own uniform assumption. Second, the pace comparison — 516 against 788 points and game 8 against game 6 for the first crossing — which measures the real final against one published sample run rather than against the simulation’s mean, and is therefore a direction rather than a measurement.

Names

Duo names are the display names on the captured leaderboard, left exactly as recorded — tags, spacing and symbols included. They do not always match the same players’ names at other events: the Summit champions appear here as Vico & Malibuca and in the EWC replay parse as BIG vic0 + BIG Malibuca 7. The head-to-head section of the stats pack is what links the two; we have not matched them on account IDs, because the Summit capture has none.