Fragcharts follows one question: which CS2 teams are on their way to the next Major, and what can still change it. Valve's Regional Standings decide who is invited; this time to Singapore, from the standings of 2 November 2026. Everything with a percentage is ours: a simulation of what is left to play before that cut.
Valve invites teams from its Regional Standings (VRS) as they are published on the cut date. Each region has its seats: Europe 18, the Americas 9 and Asia-Pacific 5. The best placed start in a later stage: in Europe the top 6 go straight to Stage 3, the next 6 to Stage 2, the 6 after that to Stage 1. The seats are not fixed: Valve sets them for each Major from how each region's teams placed at the Major before, for Singapore at IEM Cologne Major 2026.
A team's VRS points are a starting value and a head-to-head adjustment. The starting value comes from four factors: the prize money a team earned, the prize money of the teams it beat, the network of teams those teams beat, and its wins at LAN events. Each counts only the ten results worth the most to it, once weighed: prize money by how recent it is; a win over a team by what that team is worth, how recent the win is and how much the event paid; a LAN win by how recent it is. A result counts in full for a month, then a little less each day, and is gone after about six months. The head-to-head part moves a team's points after every match, by the strength of the opponent.
Valve's: the standings, the points, the four factors and the weight Valve gave each event already played. We read them as Valve publishes them and never change them.
Ours: every chance of the Major and of each stage, the place at the cut, race impact, our Elo rating, and the weight of events still to come, from their prize pool through Valve's formula or the VRS weight HLTV lists for them. They are a forecast, not a promise, and nothing on the site is betting advice.
Every night we take Valve's latest standings and play every event left before the cut, thousands of times over: until each team's chance moves less than half a percentage point over the last thousand runs, usually about 9,000 runs. All teams share each run, so two teams never win the same event.
Who plays where comes first from what is known: published team lists and qualifiers already decided. The biggest events must invite teams straight from Valve's invitation list; in each run those invitations go down the list, and a team may decline, as about half of those invited did over the last two years. The seats still open go to teams by where they have played in the last six months: the tiers, regions and LAN or online events they usually enter. Small events that are not announced yet are stood in for by placeholder events, filling each week up to as many as were played in the same weeks before earlier cuts.
At the cut, Valve's standings are worked out once, from every result in the run, as Valve does it. A team's chance is the share of runs in which it lands inside its region's seats.
Each simulated match is decided by our Elo rating, trained map by map on every played match we hold (walkovers aside) and aware of the format: a best of one is less certain than a best of three. Each team also gets a form for the whole run, drawn from how much top teams' ratings have really moved over the same number of weeks.
On 5,389 Ranked matches it had not been fitted on (1 Jan to 24 Sep 2026), the favourite by our Elo won 64.6% of the time, with a Brier score of 0.2185 against 0.2272 for a guess from Valve's points alone. Every team by it.
An event's race impact is how many points of Major chance it is expected to move, added up over the teams in the race: for each team, how likely it is to be there, how likely a top-half or a bottom-half finish is in our runs, and how far apart its chance of the Major is after each. An event that ends after the cut has none.
A roster change since Valve's latest standings: lineups are held as they are there until the cut. A team moving to another region. Events nobody has announced yet, beyond the placeholders. Prize money at events with no published prize table. And it seats top teams at fewer LAN events than they really play, about a quarter fewer in our tests.
It is also too sure close to the lines: in Europe, teams just below the Stage 1 line qualified more often than it said, and teams just above missed more often. Its near-certainties have been a little too sure as well: a few more of the teams it put under 10% made it than it said, and a few more over 90% missed.
Valve's Regional Standings, from its public repository, usually released on the first Monday of the month. bo3.gg for the calendar, results and prize pools. Liquipedia for team lists and prize tables (CC BY-SA). Some event details, whether Valve counts an event and the VRS weight it carries, are entered by hand from HLTV. Everything is fetched every night, and the site moves to the new forecast only when every step has worked.
No percentage went on the site before the model had been run on the last Majors as if it were the day of each forecast, and scored against who really qualified. Against a plain guess from the standings on the day, it was clearly better at Stage 1, better at Stage 2 though within what chance could give, and level at Stage 3, where the top eight barely move. The table below shows, for each chance the model gave, how many of those teams really made it.
Our evaluation on the last two Majors (Budapest 2025 and Cologne 2026): the model's forecasts for each region's top 40, made before each cut, against who actually qualified. Of the teams the model gave a chance in each band, the share that actually made it.
| The model said | Stage 1+ | Stage 2+ | Stage 3+ |
|---|---|---|---|
| 0–10% | 3% of 255 | 2% of 337 | 0% of 394 |
| 10–20% | 19% of 26 | 10% of 20 | 44% of 9 |
| 20–30% | 35% of 17 | 33% of 15 | 13% of 8 |
| 30–40% | 30% of 10 | 18% of 11 | 57% of 7 |
| 40–50% | 50% of 16 | 50% of 6 | 100% of 1 |
| 50–60% | 30% of 10 | 50% of 6 | 33% of 6 |
| 60–70% | 69% of 16 | 38% of 8 | — |
| 70–80% | 56% of 16 | 88% of 8 | 50% of 4 |
| 80–90% | 89% of 9 | 67% of 9 | 100% of 4 |
| 90–100% | 94% of 72 | 100% of 27 | 100% of 14 |