Is CS2 dead?

519,801 players were in Counter-Strike 2 at the last hourly sample, 35% below the seven-day average of 803,749. The busiest hour of the last 30 days drew 1,429,047, so the game is sitting about 64% below its monthly high-water mark.

The last 24 hours have averaged 855,706 concurrent players, roughly 6% busier than the week behind them. Short bursts like that usually trace back to something specific: an update, a Steam sale, or a big event playing out live.

Averaged over the last two weeks, CS2 is at its fullest around 14:00 UTC (1,275,939 players) and at its emptiest around 01:00 UTC (462,548) — a swing of roughly 176% across an ordinary day. A screenshot taken at the wrong hour can "prove" almost anything, which is why every figure on this page is shown next to an average.

The week has a shape too: weekends have averaged 871,468 concurrent players against 768,331 midweek, about 13% busier. Comparing a Sunday evening with a Wednesday morning tells you about the calendar, not about the game.

Spirit leads the world ranking on 2,031 points, 129 clear of MOUZ. From first to 10th the table covers 432 points, so a single deep run at a big event is enough to reorder most of it.

In the most recent Valve Regional Standings update, Liquid made the biggest move inside the top 20, climbing 15 places to #17.

49 professional matches landed in the results feed over the past seven days across 2 events. The most recent was Luminosity 2–3 3DMAX at Logitech G Play Connect 2026.

Fragcharts is a small independent stats site for Counter-Strike 2, and the idea behind it is narrow on purpose. The numbers that describe this game live in a dozen different places — Steam's API, Valve's own ranking files, tournament brackets, roster announcements — and most of them are published without any context at all. We collect them, keep a history so the present can be compared with the past, and write down what each number does and does not mean. No account, no paywall, and a link back to the source wherever one exists.

Most of this site reports numbers that someone else publishes. Two things on it are ours: an Elo strength rating for every team, and the match predictions built on top of that rating. Because those are models rather than measurements, they are published with a scorecard attached — you should be able to see how often the thing has been right before you decide what it is worth.

The rating itself is deliberately simple. Every team starts at 1500 points. After each match, points move from the loser to the winner, and the size of the move depends on how surprising the result was: a 1700-rated side beating a 1300-rated side collects very little, while the same scoreline the other way around is a large swing. Nothing else feeds it — no reputation, no tournament seeding, no prize money, no roster hype. That is the appeal: a rating is a summary of who a team has actually beaten, and you can follow the arithmetic yourself on the strength ratings page.

There are two versions of it, kept apart on purpose. The ladder you can browse uses a six-month window, because as a "who is playing well right now" list, results from two years ago are noise. The version used for predictions does the opposite: it keeps roughly two years of history and weights each result by its map margin, so a 2–0 counts for more than a 2–1. That split is not a hunch. Backtesting the model against a completed Major showed that the short window — not the learning rate we had been tuning — was what damaged the forecasts, and that adding margin of victory bought a further small but consistent improvement. The reasoning is written out in how we rate CS2 teams.

The accuracy figure above is a replay rather than a claim. Every match in the window is graded against the prediction the model would have made from only the games played before it, so it is never marked on a result it has already seen. Matches where either side has fewer than three prior games are skipped, because a 1500-versus-1500 coin flip proves nothing either way. Hit rate on its own is a weak measure, though — a model that called every favourite correctly at 51% confidence would be useless — which is why the prediction accuracy page also carries a Brier score and a full calibration table: when the model says 70%, the favourite should win roughly 70% of the time, no more and no less.

Once a team has a rating, the same number drives everything downstream. It sets the win probability on any head-to-head — pick two sides from the teams page to see the record, the map splits and the model's call side by side. It seeds the simulated seasons behind the race to the Major and the race into the top tier, where the remaining calendar is played out thousands of times to see how often each team lands where it needs to. And it underpins the season outlook, which sorts teams into the ones effectively already qualified and the ones still needing results. Players are scored separately and by a different method — fantasy hot picks does not use the team Elo at all, but weights a player's own performance rating by how much they have been playing and which way their recent form is pointing.

The limits matter as much as the numbers. The model knows match results and nothing else: it cannot see a roster change until the new lineup has played, it holds no opinion on map vetoes or on whether a match is online or on LAN, and it does not read the news. Expect it to lag for a few weeks after a transfer window, and to be least useful exactly where you would most like a second opinion — a reshuffled roster's first event. Every percentage on this site describes what past results imply, not what will happen, and none of it is betting advice.

Counter-Strike has been "dying" in forum threads for over two decades, and the concurrent player count keeps telling a different story. The number at the top of this page is the live count pulled straight from Steam — the players in CS2 right now — and the averages beside it put that figure in context. A quiet Tuesday afternoon is not the same as a genuine decline, which is why we compare against the 24-hour and 7-day baselines rather than a single dramatic screenshot. Player counts swing with the time of day, the day of the week, patch drops and the tournament calendar; it is the trend over weeks that actually matters.

It is worth being concrete about what a real decline would look like, because that is what makes the question answerable instead of endless. You would see the weekly baseline stepping down month after month rather than hour to hour; you would see each new peak coming in lower than the last one, even when an update or a Major should have lifted it; and you would see it echoed away from Steam, in a thinner tournament calendar, smaller prize pools and rosters disbanding mid-season. One of those signals on its own is noise. Three of them moving together for a couple of months is a trend, and that is the bar we hold the numbers to here.

The argument itself also has a rhythm. Every patch cycle produces its own wave of it — anti-cheat frustration, movement and subtick changes, the state of the skin market, a controversial map rotation. A loud community is not a shrinking one, and the two frequently move in opposite directions: the periods with the angriest threads tend to be the periods with the most people playing, simply because more players means more people with something to be annoyed about. That is exactly why a chart is more useful than a mood.

Beyond the raw count, Fragcharts tracks the competitive side of the game: the live CS2 world ranking built on Valve's official Regional Standings, pro player stats, recent results, and our own model-based match predictions and Elo strength ratings. Whether you came to settle the "is CS2 dead" argument or to follow the race to the next Major, the aim is the same: turn the scattered numbers of the Counter-Strike scene into something you can read at a glance.

The figure at the top of this page is a concurrency count: the number of people in Counter-Strike 2 at the moment we asked Steam. That is not the number of people who play the game. The population that plays CS2 in a given month is many times larger than the number online in any single hour, because people play at different times, in different regions, for different lengths of time. Comparing a concurrency figure with a "monthly active players" figure from somewhere else is the single most common mistake in these discussions, and it can make a healthy game look tiny or a small one look enormous.

Concurrency also has at least three cycles running on top of each other. There is the day — European evenings stack on top of the Americas coming online, and the small hours empty out. There is the week, where weekends carry more casual play than midweek. And there is the season, where holidays, a major update, a Steam sale or a Major final all leave a visible mark on the chart. Any two points you pick out of that will "prove" something. The averages exist so you do not have to pick.

Finally, sampling explains most of the disagreement between trackers. We take one reading an hour from the official Steam Web API and keep every one of them — that is what the 24-hour, 7-day, 30-day and 90-day views on the player-count history page are built from. A tracker that polls every minute will always report a higher peak than we do, and one that averages across a whole day will always report a lower one. None of them is lying; peaks are simply only comparable when the sampling behind them is.

Nothing on this site is hand-entered, and every section is fed by a source you can go and check yourself:

• Player counts — the official Steam Web API, sampled once an hour and stored, which is what makes the averages and the history chart possible.
• World and regional rankings — Valve's published Regional Standings, mirrored each time Valve issues an update, with the previous update kept alongside so movement can be shown.
• Matches, events and rosters — the professional circuit's results and fixture feeds, restricted to the top tiers we can cover properly rather than every open qualifier.
• Player statistics — a third-party stats API (csapi.de), refreshed daily. It is the one source we do not control, so when it goes quiet the player pages age while everything else keeps updating.
• Ratings and projections — our own models, run on the data above. The method behind the Elo numbers is documented in how we rate CS2 teams.

The limits are worth stating as plainly as the sources. We cover the top of the circuit, so a regional team's early qualifier may not appear. Team and player names follow the ranking data, which can lag a day or two behind a roster move. Projections and predictions are models: they describe what the current numbers imply, not what will happen, and they are published for reading rather than for betting. If something here looks wrong, the contact page reaches a person, and corrections get made.