HLTV updates Rating 3.0 with map-specific data

HLTV updated its Rating 3.0 algorithm to include map-specific win probabilities, accuracy fixes and revised statistical baselines using all CS2 games through September 29, 2026. The changes add data for the map Cache, alter how round swing is calculated around bomb timers, and update averages and standard deviations for sub-ratings.

HLTV’s database now contains 29,030 CS2 maps, including 9,029 contested on LAN. Rating 3.0 uses per-map averages for win probability, so identical 5v5 situations can show different odds depending on the map. HLTV cites a near 10 percentage-point gap between an attack-favored map such as Anubis and a CT-sided map like Overpass. When a specific map scenario lacks sufficient samples, the system reverts to global averages; that decision is applied per round so values do not switch mid-round.

Cache had no CS2 history when Rating 3.0 first launched and previously used global averages. With added Cache data, HLTV reports the map is CT-sided in buy rounds, with Terrorists winning 47.1% of those 5v5 buy situations. The map-specific numbers have produced larger rating shifts on Cache than on many other maps.

HLTV changed round-swing rules tied to bomb timers. A round is now treated as decided when the bomb cannot be stopped from exploding. Kills that occur with fewer than 10 seconds on the bomb, or fewer than five seconds when CTs have a defuse kit, do not add to round swing unless a CT is actively defusing. When distributing excess swing at round end after a bomb plant, only kills that occur after the bomb was planted are credited. Stand-ins will not receive round swing for rounds they did not play.

Several bugs were fixed. In some low-sample scenarios a team’s win probability could previously decrease after a kill, producing negative swing for the fragger; HLTV adjusted the logic so negative round swing after a kill is impossible. In the rare cases where win probability still appears to drop after a kill, the system now records zero swing rather than a negative value. HLTV also fixed a separate bug that caused players’ round swing to fall after making a frag.

Averages and standard deviations for Rating 3.0 sub-ratings were updated using LAN CS2 games through September 29, 2026, and values were adjusted so the mean rating across a single event is 1.00. The update produced small historical changes: across 6,687 CS2 maps played between July and September 2026, the average single-map Rating 3.0 moved by 0.021. About 9.4% of player-maps shifted by 0.05 or more, 1% shifted by 0.10 or more, and the largest single-map change was 0.31. Cache showed the largest map-specific effects, while Inferno was most affected by the new bomb-timer rules because of its frequency of post-plant scenarios.

HLTV noted the update incorporates a longer span of data following Valve’s July 2025 economy changes, which increased CT advantage. Earlier Rating 3.0 averages covered only a few months under the revised economy; adding more games increases sample sizes for rare scenarios such as a 4v4 post-plant on Inferno.

HLTV published full patch notes that detail the accuracy improvements, data sources and the specific rule changes to round swing and swing distribution. The site wrote the refinements were intended to produce fairer, more map-aware ratings ahead of its 2026 Awards and Top 20 Players list.

Stay informed about CS2 updates and check current skin prices on our marketplace.