Counter-Strike Team Rankings Explained

Valve Regional Standings vs HLTV vs ESL — what each measures, and why they never agree

TL;DR

There is one ranking that decides things and one that describes things. The Valve Regional Standings (VRS) is Valve's own open-source model and is the mandatory basis for direct invites to ranked tournaments. The HLTV world ranking is an independent editorial ranking updated weekly, and it is what most fans mean by "the rankings". They disagree because they measure different things over different time windows — neither is wrong.

"Who is the best CS2 team" has at least two defensible answers at any moment, and the gap between them is not a bug. Counter-Strike has never had a single governing body running a single table. What it has instead is a Valve-published model that controls tournament access, an independent site whose ranking carries enormous cultural weight and no formal power, and a set of organiser rankings that the 2025 rule changes have largely made redundant.

Understanding which is which takes about five minutes and saves a lot of pointless argument. Below is what each system actually does, straight from the published methodology and, in Valve's case, the source code.

The three systems at a glance

Valve Regional StandingsHLTV world rankingESL World Ranking
Published byValveHLTV (independent)ESL (organiser)
MethodologyOpen source on GitHubDescribed publicly, not open sourceDescribed publicly, not open source
RanksRostersTeams with a three-player coreTeams
Update cadenceMonthly, plus a live trackEvery MondayHistoric — see below
Decides invitations?Yes — mandatoryNoNo longer

Valve Regional Standings: the one that decides

Valve introduced the Regional Standings to replace direct invites with something teams could audit. Valve's stated goals for the system are that it be "accurate, not easily gamed, and have a transparent process", and it has gone unusually far on the last of those — the entire model, the standings files and the historical data live in a public GitHub repository, so anyone can read the code that produced a given number.

What the model actually does

The model runs in two stages. First it seeds every roster from a set of weighted modifiers, then it runs a Glicko rating over actual match results from that seeding. The modifiers in the published code are:

  • Bounty offered — the roster's own prize money earned, scaled by how recent each result is.
  • Bounty collected — the prize money earned by the opponents this roster has beaten, so beating rich teams is worth more than beating poor ones.
  • Opponent network — a transitive measure: how many distinct teams your defeated opponents have themselves beaten.
  • LAN factor — a roster's recent record at offline events specifically.
  • Own network — the count of distinct teams the roster has beaten. This one is in the code but currently carries a weight of zero, so it feeds the opponent-network term without directly scoring the roster itself.

A few consequences fall straight out of that design. Event prize pools are capped at $1,000,000 inside the model, so a single enormous event cannot dominate the table. Results are scaled by age, so the standings drift rather than lurch. Rosters with no wins at all are removed entirely. And since the start of 2025 the model only counts matches from Valve-ranked events — results at unranked tournaments are simply not in the dataset.

It ranks rosters, not organisations

This is the single most misread feature of the VRS. Valve's rules define a Roster as "a collection of five specific Athletes, irrespective of their current or later association with a given Team organization", and the model implements that literally: a lineup keeps its ranking identity only while at least three of its five players persist. Change three and the points do not follow the badge.

That is also why an invited team has to actually turn up with the roster that earned the invite. Under Valve's requirements, at least three of the invited athletes must be in the lineup and must play every match of the event, and fielding fewer than three is declared a forfeit.

How it drives invitations

A Tier 1 ranked tournament must issue at least twenty direct VRS invites, starting from the number-one ranked roster and working down in order. For every eight rosters invited from the standings, an organiser may add two wildcard invites — but only to rosters ranked outside the global top twelve, and only if they meet published criteria. Tier 2 events that use VRS invites at all must invite at least four rosters, and may only invite from below global rank twelve.

Practically, this is why the standings matter more than any other table in Counter-Strike: your position on a monthly snapshot determines which events you are allowed into for months afterwards. The consequences for the current cycle show up on our tournaments tracker and the Major hub.

HLTV: the one everyone quotes

HLTV has ranked Counter-Strike teams on and off since 2011; the current system dates from a September 2015 relaunch and has been broadly stable since. It updates every Monday.

The ranking blends three components: achievements over the past year with severe point decay month by month, recent form over roughly the last two months, and performance at offline events over roughly the last three months. Online results carry little weight at the top of the table, though they matter for surfacing emerging teams — the exception was the 2020–21 period, when a lack of LANs forced online results to be weighted more heavily throughout.

Two structural details are worth knowing. HLTV also requires a three-player core for a team to keep its points, so a wholesale roster rebuild resets the badge's standing much as it does in the VRS. And the scale is normalised so that the number-one team sits at 1000 points, with everyone else scaled beneath — which means HLTV point totals are relative to the current leader, not an absolute measure of quality.

HLTV additionally mirrors the Valve standings on its own ranking page by implementing Valve's public formula, so you can compare the two side by side. That is possible precisely because Valve open-sourced the model.

ESL: a ranking whose job has been taken

ESL ran its own World Ranking through the CS:GO era, built from results at ESL and partner events. The published outline of the system tiered tournaments by the quality and number of participating teams, applied modifiers for regional events so that a strong regional field was not undervalued, and decayed points over time so that old results faded out.

Its purpose was to give ESL a defensible basis for inviting teams to its own tournaments. That purpose no longer exists. Valve's Tournament Operating Requirements make the VRS the mandatory basis for direct invites to any ranked tournament — which includes IEM and ESL Pro League — so an organiser ranking cannot legitimately decide who plays. As of this writing, ESL's public world-ranking URLs no longer resolve to a ranking page.

If you find an ESL ranking table quoted somewhere, check its date before using it. Much of what circulates is a CS:GO-era or 2024 snapshot presented as current.

Why they disagree — five concrete reasons

  1. They rank different objects. The VRS ranks a set of five athletes. HLTV ranks a team badge that carries its points forward as long as three players remain. After a roster shuffle the two systems can be describing genuinely different entities.
  2. Money vs form.The VRS is built on prize money — yours and your opponents' — which rewards a long record of deep runs at big events. HLTV weights recent form and offline results much more heavily. A team on a hot two-month streak climbs HLTV faster than it climbs the VRS.
  3. Different eligible events. Since 2025 the VRS only ingests Valve-ranked events. HLTV counts a much wider set of results, with online play discounted rather than excluded.
  4. Different cadence.HLTV moves weekly. The VRS invitation standings are a monthly snapshot, and it is the snapshot in force on an organiser's invite date that binds — so the table that determines who plays can be weeks out of date by the time the event starts.
  5. The numbers are not comparable.HLTV normalises the leader to 1000. VRS point values come out of a different scale entirely. Comparing a team's HLTV points to its VRS points is a category error.

Which one should you use?

  • Predicting who gets invited to an event: the VRS, specifically the monthly invitation snapshot in Valve's repository. Nothing else is relevant.
  • Arguing about who is currently the best: HLTV, which is more responsive to form and is what the scene actually talks about.
  • Assessing a specific matchup: neither, really. A ranking is a season-long average; a best-of-three is decided by map pool, form and matchup history. Our live match tracker and player form analysis get closer to that question than any table does.

What csdb publishes, and what it does not

To be straight about it: we do not publish a CS2 team ranking, and we are not going to mirror one badly. Two of our pages get mistaken for a team ranking, so here is what they actually are.

  • Rankings is an individual player form leaderboard. It is rebuilt from recent S- and A-tier match data, ranking players by their average per-map rating over a stated window, with the map count shown on every row so you can judge the sample yourself. It is a form board, not an all-time or team table.
  • Teams is a settings and gear database — sensitivities, crosshairs, resolutions and peripherals grouped by roster. It has no standings column, because it is not that kind of page.

For live match data and the competitive calendar, the esports hub, match tracker and tournaments pages are the right entry points, and pro player pagescover the individuals. For team standings themselves, go to the primary sources: Valve's Regional Standings repository and HLTV's ranking page.

Frequently Asked Questions

Which CS2 team ranking is the official one?
The Valve Regional Standings. It is published by Valve, its full model code is open source on GitHub, and under Valve’s Tournament Operating Requirements it is the mandatory basis for direct invitations to ranked tournaments. HLTV’s ranking is influential and widely quoted, but it is editorial — it decides nothing.
Why do HLTV and Valve rankings disagree?
Because they measure different things over different windows. The VRS weights prize money won, the prize money of the opponents you beat, your network of distinct opponents beaten and a LAN factor, then runs a Glicko rating over match results. HLTV blends a year of achievements with heavy decay, roughly two months of recent form, and offline results from the last three months. Different inputs, different answers.
Does the Valve ranking rank teams or rosters?
Rosters. Valve’s rules define a Roster as five specific athletes “irrespective of their current or later association with a given Team organization”, and the model treats a lineup as the same roster only while at least three players overlap. Swap three or more players and you are, for ranking purposes, a new entity.
How often are the rankings updated?
Valve publishes the invitation standings monthly, and also maintains a continuously updated live track in the same repository. HLTV updates every Monday. The monthly VRS snapshot is the one that matters, because it is the version organisers must use on their invite date.
Is there still an ESL World Ranking?
ESL built one in the CS:GO era to drive invitations to its own events, tiering tournaments by field quality and size with points decaying over time. Since Valve’s 2025 rules require direct invites to ranked events to come from the VRS, that function no longer exists, and as of this writing ESL’s public world-ranking pages no longer resolve.
Does csdb have a live CS2 team ranking?
No — and we would rather say so than fake one. Our live leaderboard ranks individual players by current form, and our teams pages are a settings and gear database rather than a standings table. For team standings, the Valve Regional Standings on GitHub and HLTV’s world ranking are the primary sources.
JL

Director at Bettor Media. CS player since 2013 with experience in skin trading, marketplace analysis, and competitive play.