CS2 statistics site updates Rating 3.0 with Cache data and bomb timer

The Counter-Strike news and statistics site behind the Rating 3.0 player rating changed how that rating is calculated on October 1, 2026. In an article by NER0cs, the site said Rating 3.0 had been updated to reflect new maps and the thousands of matches played since its release, with win probabilities recomputed from every Counter-Strike 2 game up to September 29, 2026.

The changes are small in size but wide in reach. They touch the data behind every round, add map-specific numbers for Cache, introduce a bomb-timer rule for kills made after a round is effectively over, and close two loopholes in how credit is shared. The site measured the effect on the maps played between July and September 2026 and published the result alongside the patch notes.

The statistics site's update article of October 1, 2026, measuring the new formula on maps played from July to September 2026.

What Rating 3.0 measures

A player rating condenses a performance on one map, one event or one year into a single number. By design, the average player scores 1.00 over a Counter-Strike 2 event, so a 1.20 is a strong showing and a 0.85 a difficult one.

The site introduced Rating 3.0 in an article published on August 20, 2025, calling it the biggest overhaul of its rating since version 2.0 in 2017. Its central idea was context. The site argued that earlier versions could not tell a kill on an opponent with no armor from a kill on a fully equipped player carrying an AWP, a sniper rifle.

Version 3.0 answered that with two tools. The first is Round Swing, which the site describes as a metric that measures the direct impact of each kill. It estimates how much a kill changes a team's chance of winning the round, using the map, the side, the economy, the bomb's status and the number of players alive, then shares the credit between players according to the final blow, the share of damage, flash assists and whether the kill was a trade, meaning a quick revenge kill on an opponent who has just scored one.

The second is eco-adjustment. A round's economy decides what each player can afford, and a rifle against a pistol is not a fair fight. The site groups duels by the price of armor plus the most expensive weapon carried, then weights kills by how often each category wins against the other. In its launch example, a Terrorist-side kill in a rifle-against-rifle duel, a fight won 48% of the time, was worth about 1.10 kill points, while killing a player holding a starting pistol, a duel won 75% of the time, was worth about 0.54.

Rating 3.0 combines six sub-ratings: kills, damage, survival, KAST (the share of rounds in which a player gets a kill, an assist, survives or is traded), multi-kills and Round Swing. Version 2.1 had five.

An earlier correction in October 2025

The October 2026 update is not the first change to the formula. On October 29, 2025, the site published a first set of adjustments. It reduced the weight of Round Swing and raised the weight of kills, lowered the weight of KAST and multi-kills, and restored the 60-40 balance between output and cost that version 2.0 had used, after the first 3.0 version had settled on 56-44.

That update also changed how leftover swing is shared at the end of a round, splitting it between the player who wins a clutch, the kills that changed the team's chance of winning, the player who defuses the bomb and the players still alive, and added an eco-adjustment toggle to match pages and scoreboards. Matches already rated were overwritten with the corrected version.

The new round of changes is narrower. It does not touch the weights. It updates the data the formula runs on and tightens three rules about when a kill earns swing.

Why the data needed a refresh

Win probabilities are the foundation of Round Swing: the rating needs to know how likely a team is to win from any given situation before it can say how much a kill changed that likelihood. Those probabilities come from rounds already played, and the site said its database now holds 29,030 maps played since Rating 3.0 was released, 9,029 of them on LAN.

Rating 3.0 uses separate averages for each map, because the same situation does not play out the same way everywhere. The site said a player's win probability in the same five-against-five scenario can differ by nearly ten percentage points between Anubis and Overpass, a map that favors the Counter-Terrorists.

Three gaps had opened since launch. Cache had no Counter-Strike 2 data when the formula was built, so its rounds were rated with global averages; the site found it favors the Counter-Terrorists in rounds where both teams buy, with the Terrorist side winning 47.1% of them. Valve's economy changes of July 2025 had made the game more favorable to the Counter-Terrorists, which the older averages did not fully reflect. And rare situations, such as a four-against-four after a bomb plant on Inferno, needed as many rounds as possible before their win rates could be trusted.

The site said Cache now has map-specific numbers, though its sample remains small. Where an exact scenario still lacks enough examples, the formula reverts to global averages, and that choice is made per round, so a round never switches between map-specific and global numbers halfway through.

The six changes in the patch notes

The site listed six items in its patch notes. Two refresh the data, three change how Round Swing is credited, and one fixes the treatment of stand-ins, players who replace a regular member of a lineup for one match or one event.

The averages and standard deviations matter for the scale itself. They are what keeps an average player at 1.00 over a single event, and the site recalculated them from LAN games only, the matches played in person at tournaments rather than online.

The two fixes are small. In rare cases, a kill could lower a player's Round Swing, because the team's chance of winning still fell after it. A kill now earns zero swing at worst. And a stand-in no longer collects swing for rounds they never played. On the second point the site wrote, in the patch notes: "Standins no longer receive round swing for rounds they didn't play."

The bomb timer and the end of a round

The most visible rule change concerns the last seconds of a round. Once the bomb is planted, the Terrorists win if it explodes, and the Counter-Terrorists win if they defuse it in time. A defuse kit shortens the defuse, which is why the site uses two thresholds.

Under the new rule, a round is treated as over once the bomb can no longer be stopped. Kills made when the Counter-Terrorists have no kit and fewer than 10 seconds remain on the bomb, or fewer than 5 seconds when they carry a kit, no longer count toward Round Swing, unless a Counter-Terrorist is defusing. The site said those kills, and the matching Counter-Terrorist deaths, are now treated the same as exit frags, the scene's term for kills made once the outcome of a round is already settled.

The site gave one example: a clip of the player aliStair, left alone at the end of a round, which under the new rule no longer gives a disproportionate amount of swing because the round was decided before he was the last player alive.

A related change applies to rounds with a bomb plant. When the round ends, any swing left over is shared among the players who contributed. In bomb-plant rounds, that excess now goes only to kills made after the plant.

How much the numbers moved

To show the size of the change, the site ran the new formula over the 6,687 Counter-Strike 2 maps played from July to September 2026. A single-map rating moved by 0.021 on average, 9.4% of player-maps moved by 0.05 or more, 1% by 0.10 or more, and the largest single move was 0.31.

For most players over a season, those are differences in the second decimal place. For a single map, they can be larger, and they are concentrated where the changes bite hardest. The site said Cache was the most affected map, as its rounds moved from global averages to their own numbers. Inferno, a map known for long fights after the bomb is planted, was the most affected by the bomb-timer rule.

What comes next for the rating

The site tied the update to its year-end calendar. It said it believed the trade-off of small changes to old ratings was worth it for greater accuracy before the 2026 edition of its annual awards and its Top 20 Players of the Year list, its year-end ranking of individual players. The update article gives no date for either.

Version 2.0 of the rating arrived in 2017, and the site's article introducing version 3.0 was published in August 2025. The October 2025 correction changed the weights of the formula; the October 2026 update leaves them as they are and changes the data and the round-end rules beneath them.

The .esports registry publishes Esports Observer, so the whole scene can read its news for free.